The title was: Why Is No One Building the Software Stack for Energy in Africa?

I had watched a brilliant Nigerian developer demonstrate an AI coding assistant.

Beautiful interface. Impressive context awareness. Already had users. I asked whether they had considered building software for energy instead. They looked at me like I had suggested they abandon technology and begin rearing goats.

So I did what any reasonable energy professional would do.

I went home, opened several dozen browser tabs, ignored more urgent work, and wrote several thousand words accusing an entire continent of neglecting its digital energy infrastructure.

The argument was simple. Africa was preparing to deploy enormous amounts of distributed energy infrastructure, and much of the industry was coordinating it through WhatsApp groups, Excel models, scanned PDFs, screenshots of bank transfers, and Google Drive folders named NEW FINAL DATA ROOM V3 USE THIS ONE. The technology generating the electricity was becoming cheaper and more sophisticated. The technology coordinating everything around it was frequently WhatsApp.

I decided this was absurd. That part, I still believe.

The title was where I got into trouble. Because it turned out people were building. A lot of people.

I simply could not see them.

The article escaped

I expected the post to do what most of my writing did at the time: a few people read it, someone points out a wrong number, someone else asks whether I have considered blockchain, and we all continue with our lives.

Instead, founders sent it to founders. Investors forwarded it to portfolio companies. People I had never met began introducing themselves with some variation of:

“Hello. We are actually building one of the things you said nobody was building.”

Which was slightly embarrassing. It was also wonderful.

Over the following year I spoke with people building utility software, metering and billing systems, mini-grid platforms, payment infrastructure, asset-monitoring products, financing tools, carbon platforms and data products. Some were established companies operating across several countries. Some were small teams with one deployment and a dangerous amount of optimism. Some had excellent products and almost no public presence. Others had excellent public presence and a product I am still trying to understand.

That article genuinely changed my life. Not because it gave me the answer. Because it found the people.

An annoying correction

After enough of these conversations, I had to admit something. My original question was wrong.

Africa did not have a zero-builder problem. It had a visibility problem.

The companies existed, but the market could not see itself. A founder in Nairobi might be solving a problem an operator in Lagos had spent eighteen months trying to solve internally. An investor searching for African energy software companies would find the same six names repeated across every market report, while thirty smaller companies remained hidden behind outdated websites and conference brochures. Even the companies did not always know who their peers were.

There was no shared classification and no searchable view of the market. No reliable way to answer basic questions: who is building, what exactly do they build, who buys it, where has it actually been deployed, and is any of that information still current?

The ecosystem was not empty. It was unindexed. And an unindexed market behaves like a market that does not exist. Companies rebuild tools that already exist, badly, privately and repeatedly. Investors cannot fund a category they cannot describe. Engineers who want to work on energy cannot see the employers, so they go and build another AI meeting assistant, because at least that category is legible.

A visible market compounds. An invisible market keeps introducing itself from the beginning.

So I opened a spreadsheet

Of course I did. The first version of the Africa Energy Software Map was an Excel file. I understand the irony. Please allow me to have this.

It began with five columns. Then it developed tabs. Then tabs explaining the other tabs. At some point, the software map itself began demonstrating why the energy industry needed better software.

Two problems forced it to grow up.

The first was deciding what counts as energy software. Does a payment platform serving solar companies count? A solar developer whose internal monitoring platform is now used by third parties? An “IoT solutions” company, which can mean anything from a sophisticated distributed-energy control platform to a temperature sensor with excellent branding? The answer could not be “I know it when I see it.” That is how taxonomies become vibes with dropdown menus.

The second was more important. An organisation is not a product. A product is not a deployment. And a deployment is not a market. A company may be headquartered in London, founded by Africans, employ a team in Nairobi and have deployments in Nigeria, Ghana and Côte d’Ivoire. Which country does it belong to? The answer depends on the question being asked.

A logo wall cannot represent any of this. Logo walls are attractive. They are also where market intelligence goes to become decorative. Real markets are relationships: organisations build products, products solve problems, deployments happen in markets, financiers back organisations and assets. The interesting information lives between the boxes.

So the spreadsheet gradually became a connected database. That database is now the first version of the Africa Energy Atlas.

The first layer is live

Today, I am launching Phase 1:

The Africa Energy Software Map

The first release contains 2,265 organisations, 528 products, activity across 54 African countries, and 41 software categories: grid and utility software, metering and billing, mini-grid and distributed-energy management, asset monitoring, project development, financial modelling, PAYGo and payment infrastructure, energy data and analytics, carbon measurement, design and engineering, electric mobility, and the infrastructure that refuses to fit neatly inside one box.

You can search by country, market segment, technology, role, product category and maturity. More importantly, the map separates the organisation from what it builds and from where it has been used. A company saying it operates “across Africa” is not the same as evidence of a deployment in twelve countries. A headquarters tells you where the company receives mail, not where the product has proved itself.

Every record should eventually answer three different questions:

Who are they?

What have they built?

Where is there evidence that it is being used?

Those are different questions. The database treats them that way.

What the map is not: a ranking. I do not know which company is best; in several categories I am still learning what “best” should mean. A beautiful interface may hide weak field performance. A visually tragic product may quietly manage ten thousand energy assets without complaint. The Atlas is not an endorsement, not an investment recommendation, and not a paid list of organisations that successfully found my email address. Please perform actual diligence before buying software, investing money or connecting anything to a national electricity system.

It is a living record, not a finished poster: who is building, what they say they have built, where evidence of activity exists, and when that information was last checked.

Why it is open source

There are fifty-four African countries, thousands of energy companies, and products launching, merging, failing and reappearing under new names. I cannot keep all of that accurate by myself. I cannot even keep my Downloads folder accurate by myself.

So the whole project is open:

Africa Energy Atlas on GitHub

You can contribute an organisation, correct a category, challenge a claim, add a deployment, report a company that has closed, improve the data model, or fork the project and build something I have not thought of. Open source is not a decorative badge in the footer. It is the collection method. Claims lead back to sources, corrections have a history, and the people represented on the map can improve the map.

It is also why the Atlas cannot become a commercial directory where companies pay for favourable placement. Visibility can be sponsored. Facts cannot.

I should be direct about one thing. I am building inside this market, and I am not pretending to have no commercial interest in African energy software. A competitor can correct its own listing. A founder does not need to become my friend before their company is included. Neutrality does not mean having no perspective. It means designing a process that does not require everyone to trust the person maintaining it.

The map will be wrong

Let us address this early. A company will be missing. A category will be debatable. A deployment will have changed. Someone will insist that their blockchain-enabled, AI-powered, decentralised energy orchestration protocol does not belong under “Data and Analytics”. They may be right.

This is not a defect in launching. It is the reason for launching openly. A private spreadsheet can remain wrong forever without embarrassment. A public database has the opportunity to be corrected.

It also makes claims falsifiable, starting with mine. One year ago, I wrote that nobody was building the software stack. Today, you can open the map and test that statement. You can show me which layers are crowded and which remain genuinely empty. The point of learning in public is not to preserve your original argument. It is to become easier to correct.

The invitation

Go to map.kaykluz.com. Search for your company. Search for a problem you are trying to solve. Search your country, or the category you think nobody is building.

Then do one of four things: use the map, correct the map, contribute to the map, or build something the map proves is still missing.

This is Phase 1 of a larger Africa Energy Atlas. Software came first because that was the original question, but energy companies do not exist in a software-only universe. Over time the same connected database can grow to cover developers and operators, financiers and advisers, projects and assets, transactions, policies and physical infrastructure. Not seven disconnected directories with seven conflicting copies of the same company: one living model of the ecosystem. A foundation should be small enough to inspect before someone builds another five floors on it. I have worked in infrastructure long enough to respect this principle, usually several months after ignoring it.

One year ago, I asked why nobody was building the software stack for African energy. The answer arrived slowly, one email, meeting and product demonstration at a time. People were building in Lagos, Nairobi, Cape Town, Kigali, Accra, Cairo, Dakar, Lusaka, and in dozens of places I still have not looked carefully enough. They were solving pieces of the system without always being able to see the other pieces.

The builders were here all along. They just could not always find one another.

Now, at least, we have a map. Let us make it accurate together.

Explore the map: map.kaykluz.com

Inspect, contribute or fork the project: github.com/kaykluz/africa-energy-atlas

Read the article that started this: Why Is No One Building the Software Stack for Energy in Africa?

—S
Chief Impostor, provisional cartographer, recovering spreadsheet architect

P.S. Yes, the open-source software map began in Excel. No, the irony was not lost on me. It was, however, saved in a separate tab.

P.P.S. If your company is missing, this is not a strategic statement about your importance. It means I did not find you. Please submit it before writing a seventeen-post thread about gatekeeping.

P.P.P.S. The repository currently has fewer circular references than the spreadsheet it replaced. This is the highest technical standard I am prepared to guarantee.

“Anthropic says it will watermark text generated by its AI models.”

My first reaction was not intellectual. My first reaction was: wait, my memos?

Because, and I say this with the confidence of a man who bills for his judgment, I run half my drafts through Claude. Not to write them. To tighten them. There’s a difference, and I will die on this hill, and apparently the hill is now watermarked.

Somewhere in Brussels, a machine can now detect that my “per my last email” had a co-author.

Subscribe now

So that night I did the thing. You know the thing. One tab becomes nine. Nine becomes an arXiv situation. My mum asked what I was doing and I said “research” in the tone of a man watching football.

Here’s what I learned, roughly in the order I learned it.

First: the watermark isn’t hidden in the text. It IS the text. There’s no invisible ink, no secret character. When a language model writes, it’s rolling dice over which word comes next, and the watermark just shaves the dice. Ever so slightly, using a secret key, the model prefers certain words in certain contexts. You can’t see it. I can’t see it. But someone with the key can run the statistics and go: ah. Loaded dice. Claude was here.

Which means the watermark survives copy-paste, because you’re copying the word choices themselves. You’d have to rewrite the sentences to shake it off. Anthropic’s own support page says it “may persist through some editing,” which is the most lawyer-reviewed sentence I’ve read all year, and I structure energy deals for a living.

Second: this has been running quietly for two years and nobody noticed. Google’s DeepMind shipped SynthID-Text inside Gemini back in 2024. They published it in Nature (actual Nature, the dinosaur journal) and reported testing it across some 20 million live responses. Users didn’t notice any drop in quality. Twenty million people took a blind taste test of loaded dice and shrugged.

I found that genuinely unsettling in a way I couldn’t name yet. Hold that thought.

Third: the reason this is suddenly everywhere is not a moral awakening. It’s Article 50 of the EU AI Act, which kicked in on August 2. AI-generated content must be “marked in a machine-readable format and detectable as artificially generated.” Fines up to €15 million or 3% of global turnover. About 190 companies signed the accompanying code of practice: Anthropic, OpenAI, Google, Meta, Microsoft, the whole class photo.

In deal terms: this isn’t ESG. This is covenant compliance.

Now. Here is where I started enjoying myself, because the attack literature is where academics stop being polite.

There is a paper (an actual peer-reviewed ICML paper) called “Watermarks in the Sand.” Boaz Barak’s group at Harvard, proving that “strong” watermarking is mathematically impossible as long as an attacker can do two things: judge whether text is still good, and keep nudging it. Which is... everyone? That’s just a person with a paraphrasing tool. They didn’t publish a critique. They published an obituary with a poetic title.

It gets better. Researchers at ETH Zurich showed you can steal a watermark: query the model enough to reverse-engineer the dice-shaving, then either scrub it off your text or stamp it onto text the model never wrote. Cost of the attack: under fifty dollars. Success rate: over 80%.

Read that second part again. For the price of a mediocre business lunch, you can frame someone. You can take a paragraph a human wrote and make it test positive for AI. In project finance we’d call a $50 attack surface against a €15M penalty regime an asymmetric risk profile, and we would not be smiling when we said it.

And my favourite: watermarks are radioactive. If someone trains their own model on watermarked output, the watermark contaminates the student model. Detectably. The mark is hereditary. Somewhere out there is a bootleg model that caught a watermark the way you catch an accent.

Meanwhile OpenAI (and I had to double-check this) has reportedly had a ~99.9% accurate detector for its own text sitting in a drawer for about two years. Unreleased. Their stated reasons include that paraphrasing beats it anyway and that detectors tend to false-flag non-native English writers, which is the politest way of saying the thing punishes exactly the wrong people.

So by 1am my summary was: everyone is now required to stamp AI text; the stamp is cheap to apply, cheap to remove, cheap to forge; and the companies enforcing it have published the receipts on why it doesn’t fully work. Great. Wonderful. Regulation as performance art.

I was about to close the laptop.

Then I had the small thought that turned into this essay.

Every single thing I’d read — the dice-shaving, the sand, the radioactivity, the fifty-dollar heists — all of it attaches to one place: the output. The words. The pixels. The mouth of the machine.

And I remembered that the most heavily funded contrarian in AI has spent years insisting that the mouth is the wrong part of the machine entirely.

Yann LeCun. The JEPA guy. He left Meta in December, started AMI Labs in Paris, and raised a $1.03 billion seed round (reportedly Europe’s largest ever) to build “world models” instead of language models.

If you haven’t met JEPA, the short version: a language model predicts the next word. A video generator predicts the next pixel. JEPA predicts the next meaning. It watches the world and forecasts what happens next in an abstract representation space, and Meta’s own description of V-JEPA 2 is that it deliberately discards pixel-level detail as unpredictable noise. It doesn’t paint you a picture of the ball landing. It just... knows where the ball lands.

It is, architecturally, the quiet one in the meeting. Sees everything. Says nothing.

And here’s what stopped me: you cannot watermark that.

Not “it’s hard.” Not “the tech isn’t ready.” There is no attachment point. A watermark is a bias in sampled output: words, pixels, audio. JEPA’s output is an embedding. A vector. A thought, basically. I went looking for papers on watermarking JEPA-class models, fully expecting the usual arXiv pile.

Nothing. Zero. The literature on watermarking language models is a library. The literature on watermarking latent world models is an empty shelf with good lighting.

Even the law agrees, accidentally. Article 50 attaches to synthetic audio, image, video, or text. An embedding is none of those. The EU wrote a mandate for everything that talks, and LeCun’s entire billion-dollar bet is machines that don’t talk.

You can’t be quoted in the minutes if you never speak. Every deal lawyer knows this. Apparently LeCun does too.

And that’s when the actual shape of the thing showed up, five hours after the group chat message:

This was never about detection. It’s a tax on speaking.

Everything we call “generative AI” earns its keep by producing content; content is exactly the thing regulators can mandate marks on, professors can run detectors on, and fifty-dollar attackers can forge. The entire compliance apparatus of 2026, the watermarks and the C2PA stamps and the verify-this-upload tools, all of it polices the mouth.

World models are a bet that the value of AI eventually migrates from the mouth to the mind: from producing plausible content to understanding and acting. A robot that stacks boxes correctly leaves no font. A model that predicts your grid’s load curve outputs a decision, not an essay. There’s nothing to stamp, nothing to spoof, nothing for a professor to feed into a detector at 2am.

So here’s my probably-wrong-but-I-can’t-shake-it observation: if JEPA-style systems end up winning the next decade, the benchmark charts will only be half the story. The other half is that they quietly exited the content economy — right as the content economy got a customs office.

Now let me argue with myself, because the research kept me honest on three points.

One: silent models aren’t invisible. There’s a whole parallel literature (SSLGuard, REEF) on fingerprinting models by their internal representations, robust to fine-tuning, pruning, even merging. You can’t watermark a thought, but you can apparently biopsy the brain that thinks it. If someone forks V-JEPA’s open weights and claims it’s theirs, there’s a lab test for that. Different question (who copied whom, not what did it say), but it exists.

Two: the moment a world model grows a mouth (bolts on a video decoder, a language head, anything that renders for humans), it walks straight back into watermark country. NVIDIA’s Cosmos world models generate actual video, and sure enough, those outputs get watermarked. The exemption isn’t for “world models.” It’s for silence.

Three, and most humbling: JEPA cannot tighten my memo. That’s the whole reason I was in Claude on Tuesday. The talking models are taxed because talking is what we currently pay them for. Nobody watermarks the intern who never speaks, but nobody asks him to draft the IC paper either. Yet.

Still. I keep coming back to the picture of it.

One branch of AI now writes with loaded dice, wears a stamp it can’t fully keep and can’t fully trust, gets forged for lunch money, and files compliance paperwork in three jurisdictions.

The other branch watched a million hours of video, formed some opinions about how the world works, and has never once said anything you could hold against it.

I know which one sounds like the senior partner.

Anyway. I wrote this one myself. You’ll have to take my word for it, which, I learned this week, is the only kind of proof any of us has left.

— Kay

Papers, for the fellow sickos: Watermarks in the Sand (ICML ‘24) · Watermark Stealing (ICML ‘24) · SynthID-Text (Nature, ‘24) · Radioactive watermarks (NeurIPS ‘24) · LLM Idiosyncrasies (ICML ‘25): the one showing models are 97% identifiable by their word habits alone, no watermark needed, which possibly deserves its own essay · V-JEPA 2 · REEF (ICLR ‘25).

Routine work. Check the assumptions, check the returns, write comments, go to bed. I have reviewed dozens of these. I opened the revenue tab and did what I always do first: a lazy sanity check on the back of an envelope, before trusting anyone’s formulas, including mine.

Five hundred connections. Two hundred and fifty naira per kilowatt-hour. Call it one kilowatt-hour per household per day, because it’s late and one is an easy number.

500 × 250 × 30. About ₦3.75 million a month.

I stared at it. That couldn’t be right. That’s the revenue of a decent Lagos restaurant. For an entire power utility.

So I did what any professional does when his own maths embarrasses him: I assumed I was the problem. These people build mini-grids for a living. I build spreadsheets about people who build mini-grids for a living. Clearly I’d dropped a zero somewhere.

I checked the tariff. ₦250/kWh. Correct, and honestly fair; try running a diesel gen for less.

I checked the connections. Five hundred. Correct.

Which left the quiet little number in the middle. The kilowatt-hour. How much electricity does a newly connected rural household actually buy?

I did not know. Twelve years in this industry. I did not know.

So at 9:40pm I was Googling rural household consumption benchmarks like a first-year analyst, and the answer that kept coming back was worse than my lazy guess. Many newly connected households buy half a kilowatt-hour a day. Some less. Lights, phone charging, a fan when the day has been unkind. My envelope had been generous.

I went back to the proposal to find where the real revenue was hiding, because the model showed a perfectly healthy project and my envelope showed a bake sale. And there it was, in Year 3, in a curve that swept upward with total confidence, labelled productive use.

The maize mill. The welder. The cold room. Demand triples, the project sings, everyone claps.

My first reaction, and I wrote this in my notebook, was: “so the whole sector is fiction?”

And then I stopped, because that couldn’t be right either. I have stood in villages where mini-grids work. Actually work. Freezers humming, operator solvent, batteries alive past their warranty. The maths can survive. So what did those sites have that this spreadsheet only had as a curve?

I sat with the cold coffee and made a list. Every working site I could think of had its mill, its cold room, its anchor load, either already there before the panels arrived, or built, financed and nursed into existence by the developer, on purpose, with money and staff and grief.

None of them got their demand from a growth assumption.

And that’s when the envelope finally made sense, at roughly 10:15pm.

The spreadsheet assumes demand exists, waiting to be served, like a queue outside a shop. But in a village that has never had power, there is no queue. There is no shop. Demand isn’t served in these projects. Demand is manufactured. The developer isn’t really selling electricity at all. They are bootstrapping a small economy, and billing for the electrons is the receipt, not the product.

Once you see it that way, everything in the model reads differently.

The grant that paid for the generation equipment? I’d been reading it as a subsidy on power. It isn’t. It’s seed capital for an economy that doesn’t exist yet, and it’s the only reason the tariff can be low enough for that economy to ever start.

The “5% O&M” line? On a working site, half of what that money actually buys is a human being who knows every customer by name and collects payment with a patience no formula captures.

And “productive use” in Year 3? That’s not an assumption. That’s the entire business plan, wearing a revenue line’s clothing. If it appears in the revenue tab but nowhere in the CAPEX, the workplan or the org chart, then nobody is building it, and a curve nobody is building is just hope with a gradient.

So I went back through both proposals one last time, and I searched, genuinely searched, for the mill. For the line item that buys it, the person who runs it, the plan that fills it with maize.

It wasn’t there. In either one.

Which changed what I wrote in my review comments entirely. I’d sat down to answer “is this project financially viable?” and by 10:30pm I’d realised that was the wrong question. The right question, for this project and I suspect for most of the sector, is: who is manufacturing the demand, and did anyone budget for them?

If the answer is in the spreadsheet, you might have a utility.

If the answer is a curve, you have a donation with an invoice printer.

I rebuilt both models from scratch the next day, with the demand line dragged down to what households actually buy and a new cost line for the unglamorous work of building the rest. The projects don’t look dead. They look like what they actually are: patient, subsidised economic development that happens to run on solar, and is entirely honourable so long as nobody calls it something else.

I still don’t know exactly where the line sits between “subsidised but sound” and “insolvency with a ribbon-cutting ceremony”. I’m working on it. Loudly, here, so you can correct me.

The Usual Asterisks

*My maths might be wrong. Please check it.

*The numbers are rounded and anonymised; if they resemble your proposal, that is rather the point.

*If you operate mini-grids and your demand curve came true, I am begging you to tell me how.

—S (Chief Impostor, back-of-envelope enthusiast, cold coffee survivor)

P.S. To the developers actually out there building the mill before drawing the curve: you’re doing the hardest job in this industry, and this piece is a love letter disguised as an audit.

P.P.S. Found an error? Perfect. Drop a comment and I’ll fix it in v47_FINAL_FINAL_actually_final.

P.P.P.S. Yes, I reheated the coffee at 10:45pm. No, it did not help.

Whenever an African startup raises an indecent amount of money, I ignore the funding announcement.

Not completely. I read the number. I am not a monk.

Then I close the article and go looking for the smallest version of the company I can find.

The original pitch.

The first market.

The awkward early product description written before a brand agency taught the logo how to breathe.

Because the latest press release tells you what a company wants to become.

The first hundred customers tell you what it understood before everybody else did.

On 5 August 2026, Moove announced that it had raised $250 million at a $2.1 billion valuation.

I had initially heard $2.5 billion.

I was wrong by $400 million, which is an acceptable rounding error in venture capital and a dismissal offence in project finance.

The announcement came with the usual collection of numbers designed to make the rest of us reconsider what we have done with our lives.

Moove says it now operates approximately 42,000 vehicles across 29 cities in 13 countries, employs 3,300 people and generates $420 million in annual recurring revenue. The company started with 76 vehicles in Lagos.

Seventy-six.

Moove is now building autonomous fleet depots called “Nests”, because apparently once your valuation crosses $2 billion, even the car park needs a product name. It operates autonomous fleets for Waymo and describes itself as the infrastructure layer for mobility at scale.

Somewhere between an Uber driver in Lagos trying to buy a car and a Waymo robotaxi in Phoenix returning to a charging depot with no human inside it, Moove became a very different company.

Or perhaps it did not.

That was the question I could not stop thinking about. Was this an extraordinary pivot? Or had the rest of us misunderstood the business from the beginning?

So I did what any normal person would do.

I opened twenty-seven tabs, ignored several more urgent tasks, read years of funding announcements, investor profiles and founder interviews, and developed an unnecessarily strong opinion about a mobility company I do not work for.

I went looking for the machine.

Because companies do not accidentally become worth two billion dollars.

Somewhere, years earlier, somebody built a machine that kept producing customers, assets, data, revenue and trust. The latest funding round is not the machine.

It is the receipt.

And the more I looked at Moove, the more I realised that its machine has surprisingly little to do with cars.

It is a machine for turning visible demand into financeable infrastructure.

Which means it may be one of the most useful templates we have for thinking about how to build a billion-dollar energy company in Africa.

Let me explain.

I went looking for the 76-car company

Ladi Delano and Jide Odunsi met while studying in London: Jide at the London School of Economics and Ladi at SOAS.

Both were children of Nigerian immigrants. Both had grown up with the strange tension many of us know too well: deep affection for Nigeria alongside the persistent question of why a country with so much human potential could fail to create enough opportunity for its people.

They eventually became business partners and built several companies together, including a pharmacy chain in Lagos. Their stated philosophy was straightforward: identify a real social problem and build a commercially viable solution to it.

Not a charity. Not a panel discussion. A business.

In 2020, they turned their attention to mobility. The problem looked obvious. Millions of people needed work. Ride-hailing platforms needed drivers. Customers needed rides.

But many of the people willing to drive could not afford a vehicle, and banks generally would not lend to them.

The banks saw informal income. No payslip. No acceptable credit history. No conventional collateral.

No thank you.

The driver, meanwhile, faced a particularly African form of absurdity. There was work available. The work could generate enough income to pay for the tool required to do it. But because the driver did not already own the tool, nobody would finance the tool.

This is how poverty reproduces itself with excellent paperwork.

Moove’s first important insight was that a vehicle in this context was not primarily a consumer purchase. It was a productive asset.

In Ladi’s formulation, a car in Africa was not merely a luxury. It could be the foundation of an income-generating business.

That distinction changes everything.

A bank looking at the driver saw a person with insufficient assets.

Moove looked at the same driver and saw an asset that could generate hundreds of small, observable cash flows every week.

The question was no longer: What does this person already own?

It became: What could this person reliably produce if we financed the missing tool?

That is a very different credit question.

And it is much closer to infrastructure finance than consumer lending.

Three parties standing around the same missing car

There were effectively three participants in the original Moove system.

1. The driver had the labour but not the vehicle.

2. Uber had the customer demand but not enough reliable vehicle supply.

3. The financial system had capital but could not confidently underwrite the driver.

Each party possessed something the others needed. Nobody had put the pieces together.

Moove stepped into the middle. It financed the vehicle.

It used data from the mobility platform to assess whether the driver could generate enough income. It monitored the vehicle and the driver’s operating performance. And, crucially, it arranged for repayments to be deducted from the income generated on the platform before the balance reached the driver.

Moove was not simply lending money and hoping the borrower remembered to transfer something at the end of the month.

It was sitting inside the cash-flow system.

The platform could show how frequently a driver worked, how many trips were completed, how stable earnings were, how the vehicle was treated and whether performance was deteriorating. Moove could then use those signals in underwriting and servicing decisions.

That meant it could reduce one of the ugliest risks in lending: the difference between a customer being able to pay and being willing to pay.

The driver’s income came through the platform. Moove took the agreed payment. The driver received the remainder. Over time, the driver could own the vehicle.

Simple.

Except it is not simple at all.

It required Moove to combine financing, data integration, vehicle procurement, collections, insurance, maintenance and fleet operations inside one business.

The elegance was visible to the customer.

The complexity remained with Moove.

That is usually where the value is.

The car was a small factory

I think this is the conceptual leap most people miss. Moove did not finance people who already had jobs. It financed the thing that allowed them to create the job.

The car was a tiny factory.

It converted labour and time into trips. Trips became revenue. Revenue repaid the car. The car eventually became the driver’s asset.

The same object was simultaneously:

• the productive equipment;

• the source of revenue;

• part of the collateral;

• a monitored data point;

• and, through the platform, connected to the collection system.

That combination is unusually powerful.

Traditional lenders often try to predict creditworthiness by examining a person’s financial past.

Moove could examine the productive asset’s present.

• How many trips did it complete?

• How much did it earn?

• How consistently did it operate?

• Was the driver’s activity increasing or declining?

• Was maintenance behaviour creating future risk?

The underwriting file was not merely a payslip, a bank statement and one uncle’s land document from 1987.

It was work happening in real time.

In 2024, Reuters described the model as vehicle finance repaid using a portion of the customer’s weekly revenue. By that point, Moove had expanded beyond ride-hailing into vehicles used for logistics and deliveries as well.

This is the first major lesson for African energy.

LESSON 1 - Do not finance the customer’s poverty. Finance the asset’s productivity.

A business may not have an investment-grade balance sheet. It may still spend money every day on diesel. It may still lose products when the cold room fails. It may still lose production hours during outages.

It may still have customers, revenue, machinery, employees and a commercially useful reason to consume energy.

The absence of a pristine balance sheet does not mean the absence of productive demand.

Sometimes it means the financial system is asking the wrong questions.

Moove asked a better one.

Uber was doing three jobs

People often describe Uber as Moove’s partner.

That is true, but incomplete.

Inside the original model, Uber performed at least three critical functions.

• It was the demand engine: the place where rides were requested.

• It was the data rail: the system that recorded trips, earnings and driver behaviour.

• And it was the payment rail: the place through which the money moved.

This is why the relationship mattered so much.

Moove had an existing connection to Uber from an earlier outdoor-advertising business run by its founders. That helped open the door, but it did not remove the need to prove the model in Nigeria, negotiate deeper integrations and demonstrate that Moove could add reliable supply to the platform. Over time, the relationship became sufficiently important that Uber invested in Moove’s $100 million funding round in 2024.

Moove had not merely found a distribution partner. It had attached itself to an operating marketplace in which demand, data and money already existed.

That allowed Moove to do something most asset-heavy companies get backwards. It found the demand before scaling the assets. The importance of this became clearer through the things that did not work.

Moove experimented with motorcycles, delivery trucks and minibuses. Some of those efforts struggled, not necessarily because Moove could not procure or finance the vehicles, but because the corresponding demand platforms were not mature or dependable enough.

The vehicles existed.

The cash-flow machine did not.

Moove’s leadership later described a strong demand-generation platform as fundamental to the model. The company now plans supply jointly with partners by asking how many vehicles and drivers the market actually needs before deploying more assets.

That may be the most important confession in the entire story.

Moove’s model is not really vehicle-dependent.

It is demand-platform-dependent.

A car without trips is debt with seat belts.

A power plant without contracted demand is the same thing with a fence around it.

African energy has built plenty of both.

The ugly work was the moat

It is tempting to describe Moove as a fintech. That makes the company sound cleaner than it is. Fintech suggests software, dashboards, a tasteful office and people discussing customer journeys while eating catered lunch.

• Moove had to buy cars.

• Register them.

• Insure them.

• Track them.

• Repair them.

• Recover them when things went wrong.

• Manage drivers.

• Integrate with platforms.

• Raise debt.

• Manage currency exposure.

• Report to lenders.

• Navigate transport regulators.

• Build charging facilities.

• And eventually operate fleets around the clock.

This is less “fintech” and more “several difficult businesses in a trench coat”.

Ladi and Jide were repeat founders with experience across finance, consulting and operations. Tingting Peng joined in early 2021 as Moove’s first executive hire and became central to its transformation and capital strategy. The team had to develop capabilities across credit, technology, procurement, fleet operations, capital markets and country execution almost simultaneously.

There was no credible version of the business in which Moove built an app and outsourced all the unpleasant parts.

The unpleasant parts were the product.

• A driver did not merely need a loan approval. The driver needed a functioning car, insurance, maintenance support and enough uptime to continue earning.

• Uber did not merely need another finance provider. It needed vehicles and drivers that remained available.

• Lenders did not merely need a compelling impact story. They needed evidence that the assets existed, the revenue was real, the collections worked and somebody knew what was happening in each market.

Moove had to become the connective tissue between all of them.

This is one reason asset-heavy African businesses become vertically integrated. Not always because the founders woke up wanting to own the universe.

Sometimes because the interfaces between specialist providers do not work.

The bank does not understand the customer. The operator does not control the maintenance. The platform does not own the vehicle. The data does not reach the lender. The insurer does not price the actual risk.

The customer ends up carrying the failure between all five.

Moove absorbed those interfaces.

That is expensive.

It is operationally brutal.

It is also hard to copy.

Even the hiring mistakes were informative

Moove did not scale perfectly.

No company does, although the retrospective case studies usually tidy this part away.

Its leadership has acknowledged that some early hiring decisions placed too much emphasis on prestigious corporate backgrounds. Impressive logos were mistaken for stage-appropriate operating ability.

The result was what Tingting Peng described as a stage mismatch.

A person can be excellent at running one department inside a global company and still be poorly suited to building the department from a laptop, three spreadsheets and an urgent WhatsApp message from somebody at the port.

These are different sports.

As Moove became more complex, it eventually divided the company into four operating businesses—Moove Manned, Kovi, Moove AV and Moove Japan—each with its own leadership and profit-and-loss responsibility. Capital allocation and group strategy remained centralised.

That organisational structure is more important than it looks.

A capital-intensive multinational cannot run every operational decision from headquarters.

But it also cannot allow each market to invent its own capital strategy, underwriting standards and risk appetite.

Local execution.

Central capital discipline.

The same principle will matter for any African energy platform trying to operate across ten countries without discovering ten entirely different definitions of “approved budget”.

Capital was not funding. It was part of the product.

Every asset-heavy startup begins with a mildly offensive paradox.

• It needs assets to generate revenue.

• It needs revenue to attract debt.

• It needs debt to buy the assets.

Welcome to infrastructure.

Moove initially had to use equity to put vehicles on the road.

That was expensive capital doing the work of asset finance, but there was no established portfolio performance to show lenders yet.

The first vehicles produced the first repayments. The repayments produced data. The data made the next pool of vehicles easier to finance.

Gradually, the capital stack could evolve.

This is the sequence:

• Equity pays for uncertainty.

• Debt pays for repetition.

Moove learned this in real time.

Some of its early asset facilities were denominated in US dollars while drivers earned in local currencies. That created the familiar emerging-market magic trick where the customer performs exactly as expected and the company still loses money because the currency has developed other plans.

Moove shifted towards local lenders and local-currency facilities.

It also learned that the structure of the debt mattered almost as much as the interest rate. Amortising term loans consumed operating cash as principal was repaid. Revolving facilities made it easier to keep expanding the asset base while servicing the facility.

Meanwhile, the company invested in lender-grade reporting and internal controls so that institutional capital could understand what it was financing.

This is not glamorous work.

Nobody has ever gone viral on LinkedIn for reconciling an asset register.

But a lender will give more money, at a better price, to the company that can tell it where every asset is, what every asset earned, how every customer paid, what failed last month and what was done about it.

By March 2024, Reuters reported that Moove had raised a cumulative $250 million in equity and $210 million in debt. Its funding sources over time included conventional equity, bank facilities, venture debt and a sukuk. Then came the new $250 million Series C in August 2026.

The capital stack was not sitting backstage supporting the business.

It was the business.

A vehicle-finance company with the wrong financing structure is just a very organised way to go bankrupt.

Moove was not only learning how to underwrite drivers.

It was learning how to make its own balance sheet underwritable.

That distinction is the difference between a small leasing company and a global infrastructure platform.

Then the company started buying the operating system

In January 2025, Moove acquired Kovi, a Brazilian mobility company.

The transaction increased its reported fleet to 36,000 vehicles and its annual recurring revenue to more than $275 million. More importantly, Kovi brought proprietary vehicle telemetry and driver-behaviour technology that could improve maintenance decisions and risk assessment.

This is what a sensible acquisition looks like. Not merely more customers. More learning per customer.

Moove had already built a model around embedded mobility data. Kovi increased both the size of the fleet and the quality of the nervous system attached to it.

The acquisition helped Moove move further away from being a company that merely financed cars.

It was becoming a company that understood how high-utilisation mobility assets behaved. When they failed. How drivers treated them. What maintenance patterns predicted future cost. How utilisation affected economics. Which operating practices extended useful life.

That knowledge becomes valuable far beyond the original loan.

It improves underwriting. It reduces downtime. It informs procurement. It strengthens residual-value assumptions. It makes lenders more comfortable.

And every improvement makes the next asset slightly cheaper to finance.

The loop tightens.

The Waymo “pivot” that was not really a pivot

Then came the part that initially confused me.

In December 2024, Moove announced a partnership with Waymo.

Moove would manage and dispatch Waymo’s autonomous vehicles. It would be responsible for fleet operations, facilities and charging infrastructure. Waymo would continue to operate the passenger platform and remain responsible for the autonomous-driving technology itself. The relationship began in Phoenix and expanded to Miami, with London subsequently announced.

At first glance, this looks like a radical departure.

The company began by helping African drivers own vehicles.

Now it is building infrastructure for vehicles that do not need drivers.

There is an obvious social tension there, and it should not be hand-waved away. A mission built around economic opportunity for mobility entrepreneurs does not map neatly onto a future in which the entrepreneur has been removed from the driver’s seat.

But strategically, the move is much more continuous than it appears.

Look at the original system again.

• Uber had demand and software.

• Drivers had labour.

• The missing layer was financed, reliable vehicle supply.

• Moove built that layer.

Now look at autonomous mobility.

Waymo has the autonomous-driving system and the passenger platform.

The missing layer is still physical.

Someone must finance or own the vehicles. Someone must charge them. Clean them. Maintain them. Inspect them. Store them. Dispatch them. Recover them when a sensor decides the traffic cone has become an existential threat. Run the depot at three in the morning. Coordinate thousands of physical assets across a city.

The driver disappeared. The operating problem did not.

In some ways, it became more demanding.

A human driver can notice a flat tyre, find a mechanic, clean the back seat and complain loudly when the charger is not working.

An autonomous fleet still needs all those things. It simply cannot do any of them itself.

Moove followed the bottleneck.

First, the bottleneck was access to vehicles.

Then it was fleet uptime, maintenance and charging.

Now it is the physical operating infrastructure required to deploy autonomous vehicles at scale.

The company did not abandon its underlying model.

It extended it.

The current Moove website describes the business as a combination of capital, physical infrastructure and city-level operations across both human-driven and autonomous mobility. The new funding will support fleet ownership and “robotics-first” depots where autonomous vehicles can be charged, serviced and continuously orchestrated.

This is the second major lesson.

LESSON 2 - Do not remain loyal to the first product. Remain loyal to the system-level problem.

Moove’s first product was vehicle finance.

Its enduring problem is enabling mobility platforms to deploy productive physical assets at scale.

The car was never the whole business.

The operating layer was.

The five actual Mooves

I tried to turn this into a perfect MOOVE acronym. The second O refused to cooperate. I respect its boundaries.

So instead, here are the five actual moves hiding beneath the company’s growth.

Move one: find the demand before financing the asset

Moove did not begin by buying ten thousand cars and then opening PowerPoint to investigate where ride-hailing might be popular.

It connected vehicle deployment to an existing demand platform.

This reduced utilisation risk. It provided operating data. It created a path to collection.

And it gave the company a partner with a commercial reason to help the fleet succeed.

The experiments that struggled were often the ones without a sufficiently strong demand engine.

Asset-first businesses ask: What can we build?

Demand-led infrastructure businesses ask: What productive activity is already constrained by a missing asset?

That is a much better place to start.

Move two: finance productivity, not existing wealth

The target customer did not have enough conventional collateral. But the financed asset could create measurable income.

Moove underwrote that income.

This does not remove risk.

Drivers can work less. Demand can fall. Vehicles can break. Regulations can change. Platforms can alter commissions.

But these are observable operating risks, not vague judgments about whether an informal worker “looks bankable”.

The data moves the conversation from prejudice to performance.

Move three: control the cash-flow loop

Demand alone is not enough.

Plenty of businesses have customers and still die because the money travels through six accounts, two subsidiaries and one person called Uncle Frank who is “helping with collections”.

Moove embedded repayment inside the platform’s payment flow. That substantially reduced collection friction.

The lesson is not that every company must literally deduct money every week.

The lesson is that the financing structure and the revenue structure cannot live in separate universes.

The closer repayment sits to the productive cash flow, the more financeable the asset becomes.

Move four: turn operations into data and data into cheaper capital

Every vehicle generated more than revenue.

It generated evidence. Utilisation. Maintenance behaviour. Collections. Default patterns. Residual values. City-level economics.

The evidence improved underwriting and lender reporting. Better underwriting should reduce losses. Better reporting should reduce uncertainty. Lower uncertainty should attract larger and cheaper pools of capital. Cheaper capital allows more assets to be deployed. More assets generate more evidence.

This is the flywheel:

Demand → productive asset → embedded collection → operating data → cheaper capital → more assets.

The cars move.

The information compounds.

Move five: follow the bottleneck until you become infrastructure

Moove did not expand randomly into every financial product that could fit on a strategy slide.

It tried adjacent ideas, including other vehicle categories and financial services, and dropped some when the market structure was not strong enough.

But it kept moving deeper into the bottlenecks surrounding its core demand platforms.

• Vehicle access.

• Fleet availability.

• Maintenance.

• Charging.

• Depots.

• City operations.

• Autonomous fleet infrastructure.

That is how a product company becomes an infrastructure company.

Not by announcing a platform.

By repeatedly taking responsibility for the next failure point.

Then I had an unpleasant thought

At some point during this research, I realised Moove sounded more like the energy company I have been trying to describe in this series than most actual energy companies.

That was mildly irritating.

In the first essay, the CITADEL Thesis, I argued that the great African energy company would build a proprietary map of risk, own selected physical assets and use better information to price what everyone else treats as generic “Africa risk”.

In the second, the NVIDIA Thesis, I argued that it would help manufacture its own market: building the enabling platform, securing anchor demand, financing the ecosystem and eventually using its balance sheet to turn uncertain demand into bankable demand.

Moove supplies the missing operating instruction.

Citadel tells you where the information advantage sits.

NVIDIA tells you how to make the ecosystem grow.

Moove tells you how the money reaches your account on Tuesday.

That is not a small contribution.

Because the average African distributed-energy company still scales one bespoke transaction at a time.

Find a factory. Request twelve months of electricity bills. Discover they sent photographs instead of Excel. Request them again. Visit the site. Find out the “available land” belongs to somebody’s aunt. Build a model. Negotiate a power-purchase agreement. Spend eight months explaining the exchange-rate clause. Raise project finance. Build the plant.

Repeat the entire process from zero.

Each project may be good. The company does not necessarily compound. It accumulates.

There is a difference.

Moove did not accumulate unrelated vehicle loans.

It built a repeatable loop around a demand network.

That is what African energy needs.

The closest thing energy has to Uber

The obvious question is: Who is Uber in the energy version?

It is probably not one company across the entire continent. Energy demand is too fragmented. Regulation is too local. The assets vary too much.

A textile factory, telecom tower, cold room and taxi-charging depot do not behave like four branches of the same fast-food chain merely because all of them dislike diesel.

But within individual sectors, there are organisations that perform some of the same functions Uber performed for Moove.

• A telecom tower company controls hundreds or thousands of sites.

• An industrial-park operator aggregates multiple businesses behind one location or grid connection.

• A retail or franchise network has standardised sites and central procurement.

• A logistics platform sees vehicle activity and can forecast charging demand.

• An agricultural off-taker knows which processors and cold stores are active and commercially productive.

• A payment company sees merchant revenues and settles funds into their accounts.

• A bank, insurer or consumer-goods distributor may have hundreds of branches, warehouses or retail partners whose operations depend on reliable electricity.

The perfect anchor is not simply an organisation with many sites. It is an organisation that already possesses three things:

1. Demand visibility.

2. Performance data.

3. Some influence over the payment rail.

That is energy’s Uber.

The more of those three functions the partner controls, the more powerful the model becomes.

Imagine this

Suppose a payment platform serves several thousand merchants. It already knows which merchants transact every day. It knows their revenue patterns. It knows which businesses are growing. It settles money to them.

Many of those merchants run small generators because unreliable electricity threatens refrigeration, lighting, security, connectivity or basic trading hours.

A normal solar company approaches the merchants individually. Five thousand sales conversations. Five thousand site visits. Five thousand credit assessments. Five thousand collections problems.

And, eventually, five thousand reasons the sales team needs another quarter.

The Moove-style company starts elsewhere.

• It partners with the platform.

• It uses transaction history and energy-metering data to identify a first cohort of commercially active businesses.

• It installs standardised solar-and-storage packages.

• The merchant pays a service fee that is designed to sit below its previous power cost.

• The payment is collected through an agreed settlement mechanism.

• The provider owns and maintains the equipment during the contract.

• The platform benefits because its merchants remain open, connected and able to transact.

• The merchant gets more reliable and potentially cheaper power without finding the upfront capital.

• The energy company gets a visible demand pool, a data partner, a distribution channel and a collection mechanism.

• The first 100 systems generate performance and repayment data.

• That data supports a warehouse facility for the next 1,000.

• The larger portfolio can eventually be refinanced or securitised.

This is not merely rooftop solar. It is embedded productive-infrastructure finance.

The same architecture could be built around telecom towers, cold-chain networks, petrol stations, healthcare facilities, bank branches, logistics fleets or industrial clusters.

Different equipment.

Same machine.

Your first 76 systems

Moove’s first fleet was not 42,000 cars. It was 76.

That matters.

The first version of an African energy Moove should not begin with a press release announcing a 3 GW pan-African pipeline.

We already have enough 3 GW pipelines. Some of them are old enough to vote.

The first version should be narrow enough to learn from.

One sector. One anchor partner. One country, or perhaps one operating corridor. A small number of standardised systems. A payment mechanism that actually works. An operating team close enough to reach a failed asset without booking an international flight.

Here is what I would build.

1. Choose a productive-demand network

Do not begin with “African SMEs”. That is not a customer segment. It is a census category.

Choose one network whose operations are sufficiently similar: pharmacies, cold stores, telecom sites, fuel stations, small manufacturers, delivery fleets, clinics or retail branches.

The network should have recurring economic activity and a clear cost associated with unreliable power.

2. Meter before financing

Install monitoring before committing serious capital.

Learn the load. Grid availability. Generator runtime. Fuel consumption. Operating hours. Seasonality. Outage impact. Payment behaviour.

The first product may be measurement. That is fine.

Moove could underwrite because it could see the work.

The energy company must learn to see the electrons.

3. Standardise the assets

Do not custom-design a miniature engineering masterpiece for every customer. Create a limited number of system architectures that cover most of the network.

Standard components. Standard contracts. Standard monitoring. Standard maintenance procedures. Standard spare parts.

The goal is not to eliminate engineering judgment.

It is to stop spending senior engineering time rediscovering that a 50 kW inverter remains a 50 kW inverter in Ibadan.

4. Put payment beside performance

A long-term energy contract is not bankable merely because Microsoft Word allowed everyone to sign it. The collection mechanism must connect to the customer’s actual cash flow.

That may mean a central network invoice. A payment sweep. An escrow account. A parent-company undertaking. A deduction through a marketplace settlement. A prepaid energy wallet. Or a carefully designed revenue-share mechanism.

The exact structure will vary.

The principle will not.

Collections must be designed, not hoped for.

5. Use expensive capital only for the uncertainty

The first systems may require equity, grant-backed first-loss capital or an anchor partner’s guarantee.

That is acceptable. Their job is to prove the model.

Once utilisation, performance, collections and loss rates become predictable, move the assets into debt structures suited to their cash flows.

Prefer local-currency financing where revenues are local currency.

Use revolving or warehouse facilities where repeated deployment requires capital to be recycled.

Do not raise permanent venture equity to fund every panel for the next fifteen years.

Equity should build the machine.

Debt should fill it with assets.

6. Own the operational nervous system

Whether the physical equipment sits in a separate asset company or special-purpose vehicle, the platform must retain the data and operating intelligence.

Every alarm. Every payment. Every maintenance visit. Every battery replacement. Every hour of downtime. Every customer complaint. Every EPC failure. Every forecast error.

That is the underwriting engine for the next portfolio.

Outsourcing installation is fine.

Outsourcing organisational learning is not.

7. Follow the next real bottleneck

Once the company reliably supplies electricity, the next bottleneck may be cooling. Charging. Steam. Power quality. Energy procurement. Equipment financing. Grid balancing.

But the expansion should occur because the existing customer network is pulling the company into a nearby problem—not because somebody found a larger total-addressable-market number during an offsite.

Moove did not become autonomous-mobility infrastructure by randomly adding products.

It followed the operational constraint.

The energy company should do the same.

This is not “solar finance”

There are already many companies offering solar loans, leases and power-purchase agreements.

That is not, by itself, the Moove thesis. The Moove thesis is the closed loop.

1. The demand partner helps identify productive need.

2. The data improves underwriting.

3. The financed asset creates or protects cash flow.

4. The payment mechanism captures part of that cash flow.

5. The operator keeps the asset productive.

6. The resulting performance data attracts cheaper capital.

7. Then the same system is repeated across the network.

Remove any one of those pieces and you may still have a perfectly good solar company.

But you probably do not have the machine.

This distinction matters because panels are not scarce. Batteries are becoming less scarce. EPC contractors are definitely not scarce. I know this because approximately fourteen new ones appear in my inbox every Tuesday.

The scarce capability is coordinating demand, credit, capital, technology, collections and operations at portfolio scale.

That is what Moove built in mobility.

That is where the billion-dollar energy company will be built too.

Where the analogy breaks

Now for the annoying part.

Cars and energy systems are not the same.

I know.

Twelve years in energy has at least taught me that a Toyota Corolla and a 1 MW rooftop have different technical characteristics.

The Moove analogy is useful precisely until it becomes too convenient.

There are several important differences.

1 - Energy assets are harder to recover

A vehicle can be repossessed and redeployed.

A rooftop solar system technically can be removed, but by the time you have taken apart the mounting structure, repaired the roof, found a new customer with the same load and explained the situation to three sets of lawyers, you may develop a renewed appreciation for unsecured lending.

A solar asset is often physically and contractually tied to one site.

That makes customer selection, access rights, termination compensation and step-in provisions more important.

2 - Avoided cost is not automatically cash flow

A driver completes a trip and receives revenue.

A solar system may merely reduce what the customer would otherwise have spent.

That “saving” depends on the accuracy of the baseline. What would grid supply have been? How much diesel would actually have been consumed? Did production fall? Did the factory change shifts? Did the customer exaggerate fuel use because somebody mentioned free solar?

Savings can be economically real and still be difficult to collect.

The energy company must meter the baseline, define measurement and verification clearly, and avoid treating theoretical savings as if they are already sitting in an escrow account.

3 - The contracts are longer

A vehicle-finance arrangement may run for a few years.

An energy-as-a-service agreement can run for ten, fifteen or twenty.

That creates much more exposure to currency movements, regulation, technological change, customer deterioration and people who sign contracts enthusiastically before moving to another company six months later.

The underwriting cannot stop at today’s energy bill.

It must ask whether the customer, contract and asset remain defensible through several business cycles.

4 - There may be no dominant demand platform

Uber provides a concentrated marketplace. Most energy sectors are fragmented.

The energy company may have to assemble its own network through multiple anchor partners.

That makes the go-to-market slower.

It also makes the eventual platform more defensible, because the aggregation capability itself becomes valuable.

5 - Control creates responsibility

Moove’s embedded collection system reduces risk.

It also creates power.

And power can be abused.

In September 2025, some Lagos drivers protested a reported doubling of weekly remittances on certain vehicles. Drivers argued that the increase was unaffordable; reports said Moove attributed the adjustment to inflation and higher operating costs.

I do not know enough about those individual contracts to adjudicate the dispute.

But the lesson travels. A closed payment loop can be a moat. It can also become a chokehold.

An energy company deducting payments from a merchant’s settlement account must make pricing transparent, keep service standards enforceable and provide sensible restructuring mechanisms when reality departs from the model.

Otherwise “embedded finance” becomes a polite phrase for ensuring the customer experiences the downside first.

The company must be bankable.

The customer must also survive.

These are not competing objectives, no matter what the model says in Cell F97.

The autonomous bet is still a bet

It is also worth saying plainly that Moove’s $2.1 billion valuation is not proof that every part of this strategy will work.

A private valuation is a negotiated price in a funding round. It is not a public-market referendum delivered by God.

Autonomous mobility could scale more slowly than expected. Depot infrastructure could be built ahead of utilisation. Technology partners could internalise more operations. Regulations could change. Fleet economics could disappoint.

The capital intensity could become a strength or a spectacularly expensive problem.

Moove is making a large strategic bet that autonomous-driving companies will prefer to focus on software and passenger networks while specialist infrastructure operators own and manage much of the physical layer.

That thesis is credible.

It is not inevitable.

The relevant lesson is not “copy everything Moove does”.

The lesson is to understand the sequence that earned Moove the right to make the bet.

It did not begin by constructing robotaxi depots and waiting for autonomy to arrive.

1. It began with 76 productive assets attached to existing demand.

2. It built operating capability.

3. It accumulated data.

4. It raised increasingly sophisticated capital.

5. It expanded across markets.

6. It bought complementary technology.

7. It became useful to larger platforms.

8. Then it moved into the next infrastructure layer.

9. The order matters.

Build the staircase before announcing the penthouse.

What the billion-dollar energy company actually looks like

It probably does not describe itself as a solar developer.

It may begin there.

But eventually it becomes the operating and financing layer beneath productive energy demand.

It works with networks rather than isolated customers. It sees the demand before funding the asset. It measures actual performance. It standardises deployment. It owns or controls the payment mechanism. It operates the equipment rather than disappearing after commissioning. It separates the capital used to build the platform from the capital used to own mature assets. It refinances proven portfolios into cheaper, preferably local-currency, institutional debt. And it uses every operating asset to improve the underwriting of the next one.

Its flywheel looks like this:

Productive demand
→ financed energy asset
→ reliable operations
→ controlled payment
→ verified performance data
→ cheaper capital
→ more productive assets.

Eventually, the company becomes indispensable to the network.

The retailer does not call it to buy panels. The retailer expects its sites to have power.

The logistics platform does not negotiate separately for chargers, batteries, grid connections and fleet-energy software. The infrastructure layer handles it.

The industrial park does not assemble five vendors every time a new tenant arrives. The platform delivers power, thermal energy, cooling or storage under a standard service architecture.

Customers buy uptime.

Capital providers buy predictable cash flows.

The platform owns the knowledge between the two.

That is the company. Not the panel. Not the battery. Not even the electrons.

The machine around them.

Start with the work

Moove’s original genius was not recognising that people needed cars. Everyone could see that. It recognised that the work already existed. The passengers were already requesting trips. The drivers were willing to drive. The cars were the missing productive infrastructure.

So Moove financed the missing thing, connected it to the existing work, captured repayment from the resulting cash flow, and operated enough of the system to keep the loop functioning.

That is the playbook.

• Find the work.

• Finance the tool.

• Control the cash flow.

• Keep the tool productive.

• Learn from every cycle.

• Use the learning to reduce the cost of the next tool.

• Then follow the bottleneck.

The next billion-dollar African energy company may not begin with a billion-dollar pipeline.

I hope it does not.

It may begin with 76 systems inside one productive network. Seventy-six sites that are properly metered. Seventy-six customers whose payments can be observed. Seventy-six assets using standard equipment. Seventy-six maintenance histories. Seventy-six pieces of evidence that reliable energy improves productive activity and that somebody will pay for it.

Then 760.

Then a local-currency warehouse facility.

Then 7,600.

Then the next infrastructure bottleneck.

The headlines will eventually call it an energy unicorn.

They will write about the valuation. The investors. The geographic footprint. The founders smiling in company hoodies.

And somewhere in the announcement there will be a photograph of a very large facility with a name invented by a branding consultant.

But the company will have been built years earlier. Inside the first small, boring, functioning loop.

Moove’s first productive asset happened to have four wheels. The next one may have panels on the roof and a battery in the yard.

The car was never the company.

The loop was.

And the panel will not be the company either.

-S

P.S. Yes, the correct valuation is $2.1 billion, not $2.5 billion. I have corrected it before anyone from Moove’s finance team sends me a reconciliation schedule.

P.P.S. I remain disappointed that I could not construct a respectable acronym from MOOVE. “Monetise, Originate, Operate, Verticalise, Engineer” sounds like something consultants would charge $600,000 to place inside a pentagon. We have suffered enough.

P.P.P.S. Moove’s latest numbers are company-reported. Private-company revenue, valuation and operating metrics should always be read with the normal amount of adult caution. The strategic pattern is the subject here, not an investment recommendation.

P.P.P.P.S. Found an error? Excellent. The comment section remains smarter than the author. Please correct me before this becomes MOOVE_Thesis_v19_FINAL_FINAL_UseThisOne.docx.

Not financial, legal or investment advice. Also not mobility advice. I cannot drive in Lagos without becoming a worse person. Read this as research, strategy and one energy professional thinking too loudly on the internet.

I sit in meeting rooms across Africa and tell business executives three things: that they can decarbonise almost everything they do, that it will cost them nothing upfront because someone else will finance it and carry the risk, and that in most cases it will be cheaper than whatever they are using today.

Then I watch them not believe me.

Fair enough. If a stranger told me the same thing, I would also assume he was selling timeshares.

So consider this the receipts.

One more thing before we start. This is not theoretical analysis. It is a menu, and every item on it is something I arrange for companies in Africa today, either delivered by my own team or through the platforms I work with daily. Keep that in mind; we will come back to it.

The Numbers Work

Most decarbonisation presentations still begin with a familiar assumption: the green option is environmentally responsible but financially painful.

That assumption is increasingly out of date.

Renewables did become cheaper. IRENA says more than 90% of utility-scale renewable projects commissioned globally in 2025 produced electricity below the cost of the cheapest new fossil-fuel plant available in their respective markets. Solar averaged $44 per megawatt-hour and onshore wind $33.^[8]

The other half of the story is that the incumbent became more expensive.

The green option did not suddenly become cheap. The incumbent became expensive, twice over: once at the fuel pump and once at the bank.

At the pump: in April 2024, Nigeria’s electricity regulator raised the Band A grid tariff from ₦66/kWh to ₦225/kWh, a 3.4x increase for the roughly 15% of customers who consume about 40% of the country’s power^[1], which is to say: factories. Diesel went from around ₦260 per litre in 2021 to ₦1,600 to ₦1,950 in 2026^[2]. Domestic gas is priced in dollars ($2.18/MMBtu base from April 2026)^[3], so every leg of naira depreciation, over 60% against the dollar between 2022 and 2025^[4], lands directly on the boiler bill. Your energy cost is not just an operating expense, It is equally an FX position you did not choose to take.

At the bank: Nigeria’s monetary policy rate sat at a record 27.5% for most of 2025 before a token cut to 27.0% in September^[5], with commercial banks lending to ordinary corporates at 30% and beyond^[6]. Ghanaian bank lending has historically averaged near 30%^[7]. So if you decide to fund your own solar plant, boiler or fleet, you are building it with some of the most expensive money on Earth.

Which is why the sentence that actually matters in this article is this one: you no longer have to pay the capex.

A whole industry now exists to finance, build, own and operate the asset with cheap, patient, mostly DFI-backed dollars, sell you the output at a fixed tariff, and take the performance risk. Your only job is to sign an offtake agreement and pay for what you consume.

Let me walk the stack.

Layer 1: Electricity. Solar, batteries, and now the grid itself

Commercial and industrial solar in West Africa stopped being a pilot programme years ago.

The basic arrangement is now familiar. A developer finances, designs, installs, owns and maintains the solar plant. The customer signs a long-term power-purchase agreement and pays for the electricity produced. There is no upfront customer expenditure on the plant and no solar maintenance team to manage.^[23]

CrossBoundary Energy was already running a $188 million portfolio for 30 corporate customers across 14 African countries by 2022 (Unilever, Diageo, Heineken, AB InBev, Rio Tinto), on 10 to 25 year agreements where the client pays only for the power produced^[9]. By mid-2025 its awarded portfolio had reached roughly $707 million across 560MW of generation and 695MWh of storage in 20 countries^[10].

Daystar Power, bought by Shell in 2022^[11], sells the same promise in four words: no upfront capital expenditure^[12]. Empower New Energy is rolling out a portfolio of commercial-scale solar-plus-storage plants for Justrite Superstores across Nigeria^[13] under the same structure, a programme I worked on personally in a previous role.

And the bench keeps getting deeper: ManoCap Energy replaces diesel generators with solar-and-battery hybrids for C&I clients across Ghana and Sierra Leone, now expanding into Nigeria, Côte d’Ivoire, Guinea, Liberia and Togo with $20 million of Mirova debt behind it^[14], while Germany’s GRIPS Energy runs solar-as-a-service into markets as unfashionable as Senegal’s farming belt^[15].

Onsite solar under a PPA is now the boring, settled, default first move, and my own team writes these contracts too. A two-plant 2.0MW solar array now supplies about 17% of James Finlay’s tea operations in Kericho and saves roughly 2,000 tonnes of CO2 a year, sitting alongside the estate’s own micro-hydro, biogas and wood-fired steam in one of the quietest full-stack energy systems in East Africa.^[25][26][27] Del Monte’s pineapple operation in Thika commissioned an 807kWp plant this February, designed, financed, built and operated by us, next to its new frozen-pineapple line.^[24] Even the shoppers at Nairobi’s Junction Mall are browsing under our rooftop panels.^[28][29] I should know the layer well: I have done more than fifty C&I clean energy projects across the region in the last five years, and this is where most of them started.

The newer story is batteries. BloombergNEF’s December 2025 survey put average lithium-ion pack prices at a record $108/kWh, and, more interestingly, stationary storage packs at $70/kWh, down 45% in a single year^[16], making stationary storage the cheapest battery segment for the first time. That collapse is what lets us size storage not just to smooth solar but to push the backup generator towards retirement. Wood Mackenzie counted 17 African countries with more diesel generator capacity than grid capacity, Nigeria alone sitting on roughly 28GW of gensets^[17]. Every megawatt-hour a battery shifts is a megawatt-hour a genset does not burn at 30-plus US cents^[17]. The flagship proof at scale is CrossBoundary’s Kamoa Copper project in the DRC: 222MWp of solar with a 526MWh battery, contracted to deliver 30MW of round-the-clock baseload to a mine^[18]. If a battery can baseload a copper mine, it can carry your factory through the evening shift.

And then there is the grid itself. Konexa holds one of Nigeria’s first private energy trading licences and wheels renewable power across the national network to Nigerian Breweries: a roughly 40MWp solar plant with a battery in Niger State delivering about 70GWh a year to breweries in Lagos and Enugu, plus 20.5GWh a year from the 30MW Gurara hydro plant, with a MIGA guarantee wrapped around it^[19]. South Africa shows where this goes: licensed traders like Etana Energy are now wheeling at real scale, including a February 2026 deal to supply Sibanye-Stillwater with 600GWh a year of wheeled wind and solar^[20] at a stated 20 to 30% saving against utility rates^[21]. West Africa is years behind that curve, but the model has arrived, and wheeled renewable supply is now part of my own working portfolio here too, not just something I admire from a distance.

At many facilities, we can now replace most routine generator operation with onsite solar and batteries. Where the grid is reasonably available, it supplies the remaining energy. Where it is not, a smaller generator, a biomass CHP plant or remotely supplied renewable power can provide firmness.

That last option matters because not every factory has enough roof or land.

Upfront cost to the corporate in every case above: nothing.

Layer 2: Thermal heat. Steam, and now the whole powerhouse

Electricity is maybe a third of a food factory’s energy problem. The rest is steam^[30], and steam is the hardest line in my order book. Not the only line, as we will get to, but the hardest, which is why it gets its own section. So read this knowing I am talking my book, and then read the receipts anyway.

Steam is invisible, extremely hot and financially consequential.

The proper thermal sequence is usually to eliminate waste first, recover usable heat second, electrify appropriate loads where the power economics permit it, and then replace the remaining fossil fuel with the most suitable renewable heat source.^[30]

For many African food and beverage plants, sustainable biomass is commercially attractive because agricultural residues exist locally while the displaced gas, diesel, LPG or heavy fuel oil is expensive, volatile or imported.^[40]

Thermal is also the part of this stack where I am not merely studying other people’s case studies.

It is my day job.

The proposition: my team finances, builds, owns and operates a biomass boiler at your site, burning palm kernel shells, cashew shells, cocoa husks or rice husk briquettes, and sells you steam at a fixed tariff under a steam purchase agreement with an availability guarantee. Fuel sourcing risk, the thing that actually kills these projects, sits with us, not you, managed through multi-feedstock, multi-supplier contracts with moisture and calorific value specs written in, and every kilogram sourced sustainably under ISCC or equivalent certification^[31].

That last clause matters more than it looks: the three objections every corporate raises about biomass (deforestation, competition with food, supply security) are exactly what a certified, provider-managed, audited supply chain answers. The reputational risk dies in due diligence instead of in your annual report. You buy steam by the tonne, the way you buy electricity by the kilowatt-hour, and your capex contribution is zero.

This is not theoretical, and I no longer need other people’s case studies to prove it. At Berkeley Energy Corporate Solutions, where I lead the business for West and Central Africa, we have signed a steam purchase agreement with Kasapreko PLC^[32], the Ghanaian beverage manufacturer behind Alomo Bitters, to build and operate a biomass-to-steam plant at its Tema facility with a capacity of 15.3 tonnes of steam per hour^[32]: the first project of its kind in the region, delivering clean industrial steam at lower cost and lower emissions than the fossil alternative.

Several further West African projects are well advanced; they will appear here when they are signed and public, because I dislike counting deals before they close.

What is already public: over 150 tonnes per hour of bioenergy projects in development across the continent^[40], and a partnership with John Thompson, the ACTOM boiler house, to roll the same funded steam-as-a-service model through South Africa, Zambia, Mozambique etc. The client, as John Thompson’s chief executive put it, does not need to become an energy expert or an infrastructure owner.^[41] That is rather the point.

A small aside about Kasapreko, because it deserves it. This is also the company behind Ghana’s first PPA-financed commercial solar plant, a CrossBoundary installation since expanded to 1.3MWp covering roughly 19% of the plant’s electricity^[22]. One Ghanaian beverage company, quietly stacking zero-capex decarbonisation layers while larger multinationals are still holding workshops about it.

A correction to a lazy shorthand, including my own: none of this makes us a boiler company. My team sells energy as a service across the whole utility room: electricity, steam, heat, cooling and compressed air, delivered through onsite solar, battery storage, biomass boilers, electric boilers, solar thermal, heat pumps, chillers and absorption chillers, waste heat recovery, anaerobic digestion and compressed air systems^[39], all under one long-term energy purchase agreement in which we fund the full project cost, you pay only for performance, and you can buy the plant outright when the contract ends.^[39] In a typical bioenergy project we demonstrate an 85 to 95% cut in greenhouse gas emissions, and the sourcing is exactly as described above: certified, audited, ours to manage.^[40] The boiler is the flagship. The fleet behind it covers the whole energy bill.

The model has years of operating history behind it, because our sister platform in Asia was doing this before it was fashionable. Heineken’s brewery outside Phnom Penh runs on Cambodia’s largest biomass plant, which eats 45 tonnes of rice husks a day and saves around 17,000 tonnes of CO2 a year^[42][43]; another Heineken plant in Indonesia has taken 100% of its steam from agricultural waste since 2018.^[44] Danone’s Prambanan factory cut its site footprint by up to a third burning rice husk, with the ash going back to local farmers as fertiliser.^[45]

Nestlé’s Cabuyao plant in the Philippines inaugurated an 18 tonne-per-hour biomass boiler this April to run the spray dryers that make its milk powder.^{[49][46][47][48]} Thirteen bioenergy projects across six Asian countries, boilers from 6 to 25 tonnes per hour, and typical steam costs of $22 to $45 per tonne presented at Boiler World SEA 2025, falling below $20 where the feedstock sits on site.^[33] Behind them, a solar book past 300MWp, and lenders like Clifford Capital, Pentagreen, BII and FMO writing eight-figure cheques against the portfolio.^{[79][80][81][82][83]} At some point ‘unproven’ stops being an available objection.

And steam is no longer the ceiling. The same structure now extends to biomass combined heat and power: a plant at your site producing both process steam and electricity, sold under a combined steam and power purchase agreement, with the explicit aim of replacing your generators completely rather than merely supplementing them. This is old technology wearing a new contract. Mauritius has been running sugar-mill cogeneration since 1957, and its sugar-industry power producers grew to supply a large share of the island’s electricity from bagasse^[34].

In Kenya, Tropical Power’s Gorge Farm plant turns about 50,000 tonnes a year of farm residue into 2 MW-plus of engine power and heat, selling electricity at $0.10/kWh against the $0.38/kWh it costs to make the same kilowatt-hour on diesel, displacing roughly five million litres of diesel a year^[35]. A CHP that eats your own husks and shells, powers your plant and steams your process, financed by someone else, is about as close to a free lunch as this industry offers. I now arrange these too.

The template holds beyond my shop, in case you would rather trust the competition: Barry Callebaut’s cocoa processing plant in Itabuna, Brazil runs on a cocoa-husk boiler built and operated by EDP under a service contract^[36]. Nestlé’s Kabirwala plant in Pakistan runs a 20 tonne-per-hour biomass boiler covering 85% of its thermal needs^[37]. Diageo’s East African breweries burn macadamia shells and bagasse^[38]. The fossil side, meanwhile, is a dollar-priced gas bill in a depreciating currency. That is not a baseline worth defending.

Layer 3: Electric Vehicles. Forklifts to 220-tonne mining trucks

Africa’s electric-mobility story is too often reduced to motorcycles.

Motorcycles matter. But inside and around an industrial site, the more immediate commercial opportunity may be machines with jobs, and the as-a-service wrapper now covers essentially all of them: forklifts, reach trucks, tow tractors, distribution trucks, buses, payloaders, excavators, even mining haul trucks, on long-term contracts that bundle the machine, batteries, charging, maintenance, telematics and often the trained operator into one monthly or per-kilometre fee.

You never own the asset. You never carry the battery risk. I can put any of these on a term sheet today.

Start inside the factory gate, because that is where the arithmetic is silliest. Electric forklifts cost 20 to 40% more to buy and roughly half or less to run^[50]; a peer-reviewed field study of a 3.5-tonne electric unit against a Euro V diesel found a nearly tenfold reduction in energy cost per cycle and payback within a year under intensive use^[50].

Which is why the most quietly impressive EV company in West Africa is not a car company at all: Bisedge, founded in Lagos in 2020 by the same people who built Daystar Power before selling it to Shell^[51], runs a zero-capex machine-as-a-service fleet as Linde Material Handling’s partner across Nigeria, South Africa, Kenya and Tanzania, with 250-plus electric forklifts deployed^[52] for clients like Coca-Cola and Heineken^[53], and a fresh $20 million from Metier Private Equity (May 2026) to scale^[54]. Each forklift converted saves around 55,000 litres of diesel and roughly 124 tonnes of CO2 a year^[55]. Nobody puts forklifts in a sustainability report. They should; the payback is boring and fast, and someone else will finance the whole fleet.

On the road, Kenya’s BasiGo sells electric buses on a pay-as-you-drive model^[56] (near-diesel upfront price, then a per-kilometre fee covering battery, charging and maintenance, with up to 90% bank financing and a 90% uptime guarantee^[57]), with 114 buses in service by November 2025 and operators reporting up to 70% operating cost savings^[58], helped by Kenya Power’s e-mobility tariff dropping to KES 8/kWh off peak^[59]. Ampersand and Spiro run tens of thousands of motorcycles on battery-swap subscriptions^[60].

Globally, the IEA counted electric heavy-truck sales nearly tripling from about 84,000 in 2024 to 230,000 in 2025^[61], and truck-as-a-service operators like WattEV in California already price battery-electric haulage per mile at parity with diesel^[62], while Germany’s Hylane leases DHL its electric Mercedes trucks strictly pay-per-kilometre, breakdown risk included^[63].

Heavy iron is arriving in Africa on the same terms. Exxaro signed with XCMG in February 2026 for seven 220-tonne electric mining trucks at its Grootegeluk mine^[64]; Zambia hosts a copper mine running what its contractor calls the first large-scale electric mining truck deployment on the continent^[65]; DHL, Unilever and Volvo launched Africa’s first electric Superlink truck in South Africa in August 2025^[66]; DSV followed with electric Volvos in December^[67]. Volvo’s electric excavator cut hourly operating costs 74% in a 90-day pilot with Skanska^[68].

The honest caveat: electric machines win on high utilisation and lose on idle assets and absent charging, so the multi-shift forklift fleet and the fixed-route distribution run pencil today, while the long-haul intercity truck mostly does not, yet.

Layer 4: Cooling. Buy the Cold, not the Chiller

Here is a number that surprises people: for a brewery, a dairy, a pharma plant or a data hall, cooling and refrigeration can be the largest single electricity load on site, and badly maintained systems quietly drift to using 30 to 40% more power than they should^[30].

Cooling-as-a-service fixes this by removing the client from the equation entirely: the provider designs, finances, builds, owns and operates the whole cold stack (chillers, cooling towers, air handling, cold rooms, process refrigeration) and bills per unit of cooling delivered, the way Kaer in Singapore has done since 2013, charging per refrigerant-ton-hour and famously cutting one campus’s cooling energy 70% within six months^[69].

Africa’s proof point is Energy Partners in South Africa. In February 2025 it committed R300 million to build, own and operate a centralised ammonia cooling plant at Aspen Pharmacare’s Gqeberha site over 20 years, a 17MWR system expected to cut energy consumption by more than 40%, the largest ammonia cooling plant in the country^[70]. Its book includes a R360 million servitisation deal with dairy group Clover projecting R792 million of savings over the contract life^[71]. The refrigerant transition mandated by the Kigali Amendment^[72] (out with high-GWP HFC gases, in with natural refrigerants like ammonia) comes bundled in, which converts a looming compliance headache into somebody else’s engineering problem.

Nor is any of this exotic. In Asia, BECIS designs, installs and operates entire HVAC systems for a fixed rate per ton of refrigeration delivered, bundling cooling into the same energy-as-a-service book as its steam and solar^[73][79], and August Energy, a Singapore platform built by veterans of that same scene, launched a $30 million joint venture with PIDG in December 2025 to roll out solar, storage and cooling-as-a-service across the Philippines, Vietnam and Thailand^[74], with modelled energy cost savings of around 34% on combined cooling and power.^[74] And the grandfather of the whole idea, Veolia, has been selling outsourced industrial utilities, cooling included, to corporates for decades.^[75]

The concept is not the risk. The only question is who operates it well in your geography, and mine sells it directly: chillers, absorption chillers and full cold-stack contracts under the same energy purchase agreement as the steam and the solar, billed per unit of cooling delivered the way our Asian sister platform already prices it per ton of refrigeration-hour, fully outsourced and fully financed, now available in West Africa.^[39][73]

And while we are in the utility room, the same wrapper covers the most forgotten machine in the factory: compressed air as a service, metered and guaranteed^[77], because compressed air is famously the most expensive utility per unit of useful work in most plants^[76], and almost nobody owns the problem.

The entire factory can become a service contract

Put the layers together and the shape of the opportunity becomes clearer.

The factory’s structure can use lower-carbon materials.^[78]

1. Its daytime electricity can come from onsite solar.

2. Additional renewable electricity can be wheeled from a remote project.

3. Batteries can manage power quality, peaks and outages while reducing generator runtime.

4. Sustainable biomass can provide process steam.

5. Biomass CHP can produce both heat and electricity where the load supports it.

6. Cooling can be procured as a guaranteed temperature service.

7. Forklifts, trucks and heavy equipment can be electrified under long-term fleet-service contracts.

8. Waste heat can be recovered. Organic waste and wastewater can become biogas. The biomass supply chain can be managed and certified by the service provider rather than pushed back onto the factory.

The company no longer has to appoint nine contractors and then discover that nobody is responsible for the total result.

It could instead have one decarbonisation programme, one technical baseline and one commercial architecture.

The units would differ—kilowatt-hours, tonnes of steam, refrigeration-ton-hours, kilometres, machine-hours or tonnes moved—but the underlying contract would be the same:

We finance the infrastructure.
We deliver the output.
We guarantee the performance.
You pay for what your factory uses.

This is what “everything on someone else’s balance sheet” means in practice: provider-funded infrastructure and transferred operating responsibility, rather than a corporate cheque for every asset.

One accounting footnote, because accountants can smell an overstatement from another continent: a provider-funded service is not automatically off-balance-sheet under IFRS. Control rights and contract terms matter, and your auditor gets the final word.^[84]

The claim here is about upfront cash, ownership burden and risk allocation (not an unsolicited accounting opinion, I am definitely not an accountant).

Why someone else may finance your factory more cheaply than you can

There is no magic pool of philanthropic money paying for all this. The service provider still needs a return. It can nevertheless produce a lower customer tariff for four reasons.

First, the replacement technology may consume cheaper energy. Solar displaces diesel. An electric forklift replaces fuel and engine maintenance. Waste heat replaces purchased steam. Agricultural residues may replace imported boiler fuel.

Second, the new system may use energy more efficiently. Centralised cooling, professional plant operation, modern controls and properly sized equipment reduce the energy required for each unit of output.

Third, the provider absorbs and manages risks that the customer currently pays for badly: performance, maintenance, spare parts, fuel aggregation, battery degradation and residual value.

Fourth, a specialist platform can finance assets differently. It can combine projects, obtain long-tenor institutional capital and use portfolio structures that are unavailable to an individual factory.

The $30 million BECIS bioenergy facility is one example.^[80] MIGA’s guarantees for Konexa are another.^[19] They demonstrate how contracted corporate demand can be turned into financeable infrastructure cash flows.

The multinational’s creditworthiness effectively becomes infrastructure.

But cheaper must be a test, not a slogan.

The relevant baseline is not merely the price of diesel or the grid tariff. It is the customer’s all-in cost: fuel, maintenance, labour, consumables, downtime, inventory, capital, FX exposure, future replacement and the value of reliability.

The proposed service price must beat that number on an agreed set of assumptions.

Currency matters too. Financing a project in dollars and then charging a naira-earning customer an aggressively dollar-indexed tariff can simply move the FX risk from the fuel invoice into the PPA.

A good contract allocates currency and indexation according to the actual cost base, financing structure and customer revenues. A bad one takes a cheaper technology and finances it into an expensive problem.

My commercial rule is simple: I will not ask the sustainability department to justify a project that the operating economics cannot support.

If the financed tariff does not beat the properly calculated incumbent cost (or offer a reliability improvement valuable enough to justify the difference) the project is not ready.

The honest ledger

Now the part a marketing department would usually remove.

I cannot point to a publicly documented West African factory that has placed every layer in this article under one master contract.

There are facilities that have combined several of them. Kasapreko has both financed solar and a signed biomass-steam project.^[22][32] Clover has outsourced power, steam and refrigeration.^[71] Nigerian Breweries is using site-side systems and wheeled renewable power.^[19] Bisedge is electrifying industrial material handling.^[52] CBI Ghana is lowering the carbon content of construction materials.^[78]

The pieces exist.

The integrated category is still forming.

Nor does “end to end” mean a multinational has reached literal net zero across every Scope 1, 2 and 3 emission.

A food company may decarbonise its factory while emissions remain in fertiliser, agriculture, packaging and third-party logistics.

A data centre can address electricity, cooling, backup power and construction materials while server racks, semiconductors and networking equipment remain deeply carbon-intensive.

A mine can electrify light vehicles, loaders and some processing loads while its heaviest haul routes remain difficult.

A cement plant can reduce clinker, substitute fuels and improve efficiency, but conventional cement chemistry still releases process emissions.

Batteries are excellent at seconds, minutes and hours. They are not automatically the cheapest answer to several consecutive days without grid power or sunlight.

Biomass is useful only where the feedstock is genuinely sustainable, traceable, locally available and contractually secure.

Electric heavy trucks work best on the routes that suit them. Low-utilisation, long-distance vehicles with no charging infrastructure remain a weak starting point.

And not every “green” product is cheaper. Some lower-carbon materials still carry a premium, particularly where supply is immature.

So the defensible claim is not:

“Every African industry can become absolutely zero-carbon tomorrow at a lower cost.”

It is:

A very large share of the controllable emissions from African industrial facilities can already be removed using commercial technologies, long-term finance and service contracts—and many of the first reductions can lower the customer’s total operating cost.

That is still a very big claim.

It is also one I am prepared to put into a model.

The actual pitch

Here is the disclosure I have delayed less than most people would.

I sell this for a living.

Thermal decarbonisation is the most publicly visible part of my current work. The Kasapreko agreement carries my name and quote because I helped originate and progress it. The Asian operating portfolio gives us evidence that the model works beyond a presentation.

But my work in Africa today is wider than boilers.

I can structure onsite solar, battery storage, wheeled electricity, Cooling-as-a-Service, electric commercial fleets and heavy equipment, biomass steam, biomass CHP and the sustainable feedstock supply chain required to support them.

The assets can be financed over long periods. The operations, maintenance, technology and performance obligations can be bundled into the service. The customer can buy electricity, steam, cooling, kilometres or machine availability instead of becoming the owner of a collection of unfamiliar infrastructure.

I am not claiming that I have already closed every layer under one agreement.

I am saying that I can sit with an industrial customer, examine the entire facility from foundation to loading bay, determine which parts can be decarbonised commercially today, assemble the appropriate technology and operating partners, and put financed offers behind the options that work.

A technical baseline. A risk allocation. A tariff. A performance guarantee. And a counterparty prepared to write the cheque.

Bring twelve months of electricity and fuel bills, interval load data, steam and cooling profiles, fleet routes, equipment duty cycles, site drawings and planned capital projects.

I will bring the engineering options and the term sheets. Your signature on an offtake agreement is the entire ask.

This is what I do for a living, some of it delivered by my own team, the rest arranged through the platforms I work with every day, across Africa, for exactly the kind of companies reading this.

If any layer of this stack is sitting on your to-do list behind “get through this quarter,” reply to this email or find me on LinkedIn.

Bring your last twelve months of energy bills, your fleet list, or your next build’s bill of quantities. I will bring the term sheets.

Worst case, you get a free benchmark of what you are overpaying. Best case, you get cheaper, cleaner everything, paid for by someone else’s balance sheet.

P.S. Full disclosure, because this newsletter is called what it is called: I sell several of these services for a living and arrange the rest, and one of the press releases cited below carries my own quote in it. I am talking my book from the first paragraph to this one. That is also precisely why I know these term sheets exist rather than merely believing in them, so weigh the bias against the access and make your own call.

A confession: I still cannot find an independently audited, third-party cost-per-tonne-of-steam benchmark for biomass versus HFO in West Africa. Every number I have is from a vendor, a conference presentation (yes, including my employer’s) or my own models. If you have real operating data from a real plant, I will trade you a very good dinner for it.

P.P.S. If you think I am wrong about any of this, especially the “cheaper” claims, say so in the comments or by reply. Being corrected in public is the cheapest consulting I ever receive.

P.P.P.S. If this was useful, subscribe, and forward it to the one person at your company who owns the energy budget and the one who owns the sustainability slide. They are rarely the same person, which is roughly 40% of the problem.

Next up: I am building an open, live map of every software company in African energy, and the launch piece explains why the boring plumbing matters more than the apps. Stay tuned.

References

Repeated citation numbers point to the same source.

Energy economics

1. NERC — April 2024 Supplementary MYTO Order (Band A)

2. Nigeria National Bureau of Statistics — Automotive Gas Oil (Diesel) Price Watch

3. NMDPRA pricing analysis — 2026 domestic base price for natural gas

4. IMF — Nigeria: 2026 Article IV Consultation

5. Central Bank of Nigeria — 2025 Monetary Policy Committee decisions

6. Nairametrics — Nigerian commercial lending rates

7. Bank of Ghana — Interest-rate data

8. IRENA — Renewable power generation costs in 2025

Electricity, solar and storage

9. CrossBoundary Energy — $188m C&I renewable-energy portfolio

10. MIGA — $495m framework supporting CrossBoundary Energy projects

11. Shell — Acquisition of Daystar Power

12. Daystar Power — No-upfront-capex service model

13. Empower New Energy and AFRIGREEN — Justrite solar-plus-storage financing

14. Mirova — $20m ManoCap Energy debt financing

15. GRIPS Energy — SCL solar project in Senegal

16. BloombergNEF — 2025 Lithium-Ion Battery Price Survey

17. Wood Mackenzie — Africa’s distributed diesel-generation landscape

18. Kamoa Copper and CrossBoundary Energy — Solar-plus-storage baseload project

19. MIGA — Konexa renewable supply for Nigerian Breweries

20. Etana Energy and Sibanye-Stillwater — Wheeling PPA

21. Sibanye-Stillwater — 2025 results and energy-cost savings

22. CrossBoundary Energy — Kasapreko solar project

23. Daystar Power — Corporate service model

24. BECS — Del Monte Kenya 807kWp solar commissioning

25. BECS — James Finlay Kenya 2.0MW solar commissioning

26. The Star — James Finlay solar-project coverage

27. Standard Media — James Finlay solar-project coverage

28. LinkedIn — Junction Mall Nairobi rooftop-solar site visit

29. LinkedIn — Junction Mall Nairobi rooftop-solar project post

Industrial heat and bioenergy

30. Review — Energy use and decarbonisation in food and beverage manufacturing

31. ISCC — Sustainability certification system

32. BECS — Kasapreko biomass steam-purchase agreement

33. Boiler World Update — BECIS biomass Steam-as-a-Service

34. Study — Bagasse cogeneration in Mauritius

35. Tropical Power — Gorge Farm Energy Park

36. EDP and Barry Callebaut — Itabuna cocoa-husk biomass project

37. Nestlé Pakistan — Kabirwala biomass boiler

38. Diageo — Biomass energy at East African Breweries

39. BECS — Energy-as-a-Service solutions

40. BECS — Sustainable biomass, 150 t/h pipeline and 85 to 95% GHG reduction

41. Engineering News — John Thompson and BECS biomass-steam partnership

42. BECIS — Heineken Cambodia biomass plant

43. Khmer Times — Heineken opens Cambodia’s largest biomass plant

44. BECIS — Heineken Indonesia bioenergy case study

45. BECIS — Danone Prambanan bioenergy and circularity case study

46. Nestlé Philippines — Cabuyao biomass-boiler inauguration

47. Philstar — Nestlé Cabuyao biomass facility

48. BusinessWorld — Nestlé Philippines Cabuyao biomass facility

49. Enertech — Nestlé Cabuyao 18 TPH biomass-boiler project

Electric mobility and machinery

50. Research Square — Comparative field study of electric and diesel forklifts

51. Endeavor Nigeria — The founders behind Bisedge and Daystar Power

52. GCPF and responsAbility — Investment in Bisedge’s electric logistics fleet

53. Bisedge — Electric logistics clients and service model

54. Metier Private Equity — $20m investment in Bisedge

55. Energise Africa — Bisedge forklift fuel and emissions savings

56. BasiGo — Pay-As-You-Drive model

57. BasiGo — Financing and uptime support

58. Africa Signal — BasiGo fleet deployment and operating savings

59. Kenya Power — E-mobility off-peak tariff

60. Associated Press — African electric motorcycles and battery swapping

61. IEA — Global EV Outlook 2026: heavy-truck sales

62. WattEV — Battery-electric trucking economics

63. Hylane and DHL — Pay-per-kilometre electric trucks

64. International Mining — Exxaro and XCMG mining trucks

65. Africanews — Electric mining trucks deployed in Zambia

66. DHL, Unilever and Volvo — Africa’s first electric Superlink

67. DSV — Electric Volvo truck deployment in South Africa

68. Volvo CE — EC230 Electric pilot results

Cooling and industrial utilities

69. Ellen MacArthur Foundation — Kaer Cooling-as-a-Service case study

70. Energy Partners — Aspen Pharmacare cooling-as-a-service project

71. Business Report — Energy Partners and Clover servitisation agreement

72. UNEP Ozone Secretariat — Kigali Amendment overview

73. BECIS — Cooling-as-a-Service

74. PIDG and August Energy — $30m Energy-as-a-Service platform

75. Veolia — Integrated industrial utilities management

76. U.S. Department of Energy — Cost of compressed air in industrial plants

77. Kaeser — Compressed air supplied as a metered utility

Cement and buildings

78. World Cement — CBI Ghana’s calcined-clay plant

Finance, standards and integrity checks

79. BECIS — 300MWp-plus solar portfolio presentation

80. BECIS, Clifford Capital and Pentagreen — $30m green-loan facility

81. BECIS, BII and FMO — $50m financing facility

82. BECIS — New equity financing

83. BECIS — $45m in new equity

84. IFRS Foundation — IFRS 16 lease definition and control rights

Let me show my working, which regular readers will recognise as the part where I usually get into trouble.

The 32.5 GW figure comes from BloombergNEF’s Africa Power Transition Factbook 2024, which worked out what the African Union’s 300 GW renewables-by-2030 target actually requires: quadrupling the 7.9 GW added in 2023, every year, for the rest of the decade. The 4.2 GW comes from IRENA’s Renewable Capacity Statistics 2025, and it was not a fluke of a bad year. It was 0.7% of the 585 GW the world added in 2024, a record year globally in which Africa, home to 60% of the planet’s best solar resources by the IEA’s own assessment, went backwards in relative terms.

I structure energy deals in West Africa for a living, which means I am partly responsible for these numbers. Keep that in mind as I explain why the next four years decide the next thirty.

The money numbers come from three different reports, and the difference matters

If you have read anything about African energy investment, you have seen a version of this sentence: investment must more than double, from around USD 100 billion a year to somewhere between USD 190 billion and USD 240 billion.

The sentence is true. The actual numbers to believe depend on what reference reports is used.

The IEA’s Africa Energy Outlook 2022 built the Sustainable Africa Scenario and set the original bar: over USD 190 billion each year from 2026 to 2030, with two-thirds going to clean energy.

It also produced my favourite underused statistic: in that scenario, energy investment rises to 6.1% of Africa’s GDP over 2026 to 2030. Pretty much a macroeconomic reorganisation.

The IEA and African Development Bank’s Financing Clean Energy in Africa (2023) restated the requirement as over USD 200 billion annually by 2030, against actual 2022 investment of under USD 90 billion, of which clean energy was roughly USD 25 billion. Do the arithmetic: two-thirds of USD 200 to 240 billion is about USD 130 to 160 billion of clean spending. From a USD 25 billion base, that is a five-to-sixfold increase. When you see “clean energy investment must rise sixfold” in a conference deck, this is where it comes from, usually without the working shown.

The IEA’s Clean Energy Investment for Development in Africa (2024) supplied the granular targets: total investment reaching almost USD 240 billion annually by 2030; electricity sector investment growing from just under USD 30 billion in 2022 to more than USD 120 billion in 2030; private capital growing 2.5 times, from around USD 75 billion to USD 190 billion.

And what actually happened?

The IEA’s World Energy Investment 2024 put actual African energy investment at roughly USD 110 billion, about USD 70 billion of it fossil and about USD 40 billion clean. So against a clean energy requirement of USD 130 to 160 billion a year by 2030, the continent is running at roughly a quarter of the destination, with four years left and the ramp not visibly ramping.

One genuinely encouraging line hides in the World Energy Investment 2025: private clean energy investment in Africa tripled from about USD 17 billion in 2019 to almost USD 40 billion in 2024. Private money is showing up. The problem is where the rest went. Public and development finance institution funding for African energy fell by roughly a third over the decade, to USD 20 billion in 2024, driven largely by a reduction of more than 85% in spending by Chinese DFIs. The capital that is growing chases the most bankable markets. The capital that reaches the hardest places is the capital in retreat.

Africa, with about 20% of the world’s population, receives about 2% of global clean energy investment and about 3% of total energy investment. Two different denominators, routinely blurred in op-eds, kept separate here because I would like to still be quotable in five years.

Projects are slow, and the calendar knows it

The mechanism that turns “2030 target” into “2026 decision”, is a Gantt chart.

A utility-scale solar project in sub-Saharan Africa routinely takes well over a year from tender to financial close, and that is under the World Bank’s streamlined Scaling Solar programme, which exists specifically to make this faster. Ordinary projects take longer. I have watched projects take longer. I have, on at least two occasions, been the reason projects took longer, which I mention so you know I am critiquing the mirror as well as the market.

After financial close comes construction and commissioning, another one to two years for solar if everything goes well, which it will not, because something involving a customs official and a shipping container never goes well.

Transmission is worse. A 400 kV line takes up to seven years in a best case, and South Africa, the continent’s most sophisticated power market, is estimated to be roughly two decades behind on its transmission build. Kenya’s 310 MW Lake Turkana wind farm remains the canonical cautionary tale: the turbines stood finished while the 428 km Loiyangalani to Suswa line ran years late, and Kenyans paid for electricity that physically could not reach them.

Chain it together. A project that has not reached financial close by 2027, or generously 2028, will not be generating by December 2030. Which means every megawatt that counts toward the 2030 targets is being originated, negotiated, financed or abandoned right now, by identifiable people, in identifiable meeting rooms, some of which I am sitting in while writing this instead of paying attention.

That is why 2026 to 2030 is not just another five-year planning cycle. The 2030 scoreboard is being written in the present tense.

Let us concede 2030 now and get it over with

Having spent two sections building the window, I am now going to say the quiet part: the 2030 targets are almost certainly gone.

Tracking SDG7: The Energy Progress Report 2025 counted 666 million people without electricity in 2023, 85% of them in sub-Saharan Africa, and projects that on current trajectory 645 million will still lack power in 2030, with 1.8 billion people worldwide still without clean cooking. BloombergNEF’s own 2030 forecast for African wind and solar additions falls 43% short of the required build, and even the upper range of countries’ own targets misses by 35%. When the optimistic case misses by a third, the target is not in danger. The target is in the past tense.

So why write 3,000 words about a window onto a landscape that has already burned down?

Because the window was never really about the date. Missing 2030 while adding 4 GW a year and missing 2030 while adding 20 GW a year are entirely different civilisational outcomes. The first means entering the 2030s stalled: same access deficit, same institutional muscle memory, same pitch decks with the year find-and-replaced. The second means entering the 2030s compounding: auction programmes that work, offtakers that pay, transmission that exists, local banks that have seen a solar loan perform. Institutions and pipelines compound the way interest does, and 2026 to 2030 sets the base of the exponent.

I honestly think that this window does not decide whether Africa hits 2030. It decides the slope of everything after it.

The real constraint is the price of money

The strongest objection to this entire essay goes: the calendar is not the binding constraint, the cost of capital is, and no amount of window rhetoric changes a discount rate. I know this objection intimately because it is what I mutter in financial models at 1 a.m.

The numbers behind it are brutal. The Clean Air Task Force’s October 2024 analysis puts Africa’s average power sector weighted average cost of capital at 15.6%, more than three times the 2 to 5% typical of Western Europe and the United States (their breakdown: 2.4% in Japan, 4.2% in Western Europe, 5.1% in the US, 6.6% in China). The IEA has separately estimated African power WACC above 18% in 2023. And the IEA’s World Energy Investment 2025 contains the single most quietly devastating sentence in the whole literature: in Africa, overall debt servicing costs are equivalent to over 85% of total energy investment in 2025.

Read that again. For every dollar the continent invests in energy, it pays roughly 85 cents servicing existing debt. The panels are cheap. The money is not.

Cost of capital is not weather. It is the output of decisions: sovereign guarantees, currency frameworks, concessional blending, offtaker reform, DFI risk appetite. Financing Clean Energy in Africa (2023) estimated that about USD 28 billion a year of concessional capital could mobilise USD 90 billion of private clean energy investment by 2030, a more than tenfold increase on current concessional flows. Whether that concessional capital materialises is being decided in exactly this window: in Mission 300 country compacts, in the direction of EU Global Gateway, in whatever replaces the American development finance apparatus, of which more shortly.

The calendar and the cost of capital are not competing explanations. The window is when the cost of capital gets decided.

Waiting will not get cheaper

The second objection is the patient one: solar keeps getting cheaper, so a project built in 2032 beats one built in 2027, and the window is just impatience with a deadline attached.

This argument had a good decade. It just ended.

Solar modules bottomed at USD 0.07 to 0.09 per watt in 2024 and early 2025, an absurd, historic low. Then China withdrew its solar export VAT rebate and Wood Mackenzie tracked module prices rising roughly 9% in late 2025. SolarPower Europe’s Global Solar Market Outlook 2026 to 2030 forecasts global installations declining 8% in 2026 to 612 GW, after a record 664 GW in 2025: the first contraction in more than twenty years, led by a 24% drop in China.

The great deflation is over, or at least on pause. Meanwhile the thing that actually determines African project economics, the 15.6% cost of capital, does not fall while you wait. It compounds against you. A country that waits five years does not collect cheaper panels. It collects five more years of diesel imports, five more years of 85 cents on the dollar to creditors, and the same financing penalty at the end of it.

Waiting was never free. Now it is not even cheap.

The leapfrog has a weight limit

Every essay in this genre is legally required to mention mobile money, so let us do it properly and then put it down.

Yes, Africa skipped fixed-line telephony and built M-Pesa. Yes, there is a real energy version of this: Afrobarometer’s 34-country survey finds 23% of households using electricity from a source other than the national grid, 16% relying on off-grid sources alone, mostly solar. At the household level, the leapfrog is not a metaphor. It is happening in several hundred million living rooms.

But a 2024 paper in Science on technology leapfrogging in Africa states the limit plainly: off-grid solar can leapfrog the grid for lighting and phone charging, but it cannot leapfrog the massive generation and grid investment needed to power a growing economy. A phone call needs milliwatts. A cement kiln does not care how inspiring your fintech comparison is.

So the greenfield advantage is real but bounded. Africa can absolutely deploy AI-native grid management, smart metering and distributed architecture without ripping out legacy systems, because in much of the continent there is mercifully little legacy to rip out. What it cannot do is app-store its way past transmission lines and baseload. The leapfrog carries households. Industry still needs the boring, slow, seven-year infrastructure, which is precisely the infrastructure whose 2030s existence is being decided in this window.

The window is a decision interval

If the window were merely a countdown, this essay would be a eulogy. It is not, because the defining feature of 2026 to 2030 is how much of it is being actively, visibly decided by named actors, in both directions at once.

On the ledger’s good side: Mission 300, launched by the World Bank and AfDB, targets 300 million new connections by 2030. In January 2025, thirty African heads of state endorsed the Dar es Salaam Energy Declaration, with 48 countries eventually signing on and partners pledging more than USD 50 billion. By mid-2026 the AfDB reported over 50 million people connected across 40 countries. The World Bank has committed roughly USD 30 billion for African energy through 2030, nearly double its historic pace. These are the largest coordinated electrification commitments in the continent’s history, and they were made inside this window.

On the ledger’s other side: USAID, and with it Power Africa, was effectively terminated in early 2025. Power Africa had brought roughly 12,000 MW to financial close and mobilised about USD 22 billion since 2013. Germany, France and the UK have cut development budgets. Chinese DFI energy lending is down more than 85%. The concessional layer that the whole USD 28 billion mobilisation arithmetic depends on is being dismantled and rebuilt simultaneously, live, by people with calendars and elections.

That is what makes this a decision interval rather than a countdown clock. Countdown clocks run whether you watch them or not. This window is made entirely of choices: a regulator gazetting a tariff methodology, a DFI board approving a guarantee facility, a finance ministry choosing which compact to actually implement, a developer choosing to reach financial close in 2027 instead of writing a strategy document about 2035.

The next four years will not decide whether Africa hits its 2030 targets. That question has quietly answered itself.

They will decide whether the 2030s begin with momentum or with a rebrand of the same deficit. And unlike the targets, that outcome is still entirely for sale.

P.S. Honest disclosures, as always. The “quarter of the destination” and trajectory arithmetic in this piece are my own calculations from IEA figures, not IEA findings; the underlying numbers span report vintages from 2022 to 2026, and the IEA has an endearing habit of restating the same requirement as USD 190 billion, USD 200 billion and USD 240 billion depending on the framing, so I have cited each figure to its specific report and you should too. My maths might be wrong. It has been before. Check it.

P.P.S. A confession. I started researching this essay to prove the window thesis was airtight, and the evidence kept politely informing me that the thesis was right but the deadline was dead. I have chosen to publish the version the data supports rather than the version I pitched myself, which my MBA professors would call intellectual honesty and my inner impostor calls losing an argument to a spreadsheet.

P.P.P.S. If you think I am wrong, and in particular if you think 2035 is the honest deadline and this whole window framing is consultant theatre, I genuinely want to hear it. The comments are open and I respond to everything, usually at hours that concern my family.

P.P.P.P.S. If this was useful, subscribe, and forward it to the one person you know who still has “300 GW by 2030” in a live slide deck. They need this more than either of us.

P.P.P.P.P.S. Next up: the Africa Energy Software Map goes live, and possibly a new short-form section for those of you who have complained, with some justification, that my essays are too long. Still not your lawyer.

Every C&I solar developer in Africa is, whether they admit it or not, building for one of two exits: a trade sale or a refinancing. The trade sales are no longer hypothetical. Shell bought Daystar Power, the Lagos and Accra C&I platform, in a deal that closed in December 2022 and was described at the time as the oil major’s first power acquisition in Africa. Helios and AIIM merged Starsight Energy with SolarAfrica to create a pan-African C&I platform that completed with 520 MW of installed and contracted solar and 60 MWh of storage across seven countries. At utility scale, Infinity Power’s purchase of Lekela’s 1 GW wind portfolio was billed as Africa’s biggest renewables deal.

The consolidation phase of African clean energy has started, and it will accelerate: AFSIA’s 2026 outlook identifies 23.4 GW of operational solar on the continent, and possibly as much as 63.9 GW once Chinese module export data is factored in, growing faster than any other region on earth. By project count, more than 80% of the operating projects AFSIA tracks are C&I. Thousands of small contracted assets, in other words, most of them built by developers who will eventually want to sell.

When a buyer does arrive, they do not buy your panels. Panels are a commodity that loses value the moment it clears the port. What they buy is a bundle of paper: power purchase agreements, leases, land documents, permits, EPC and O&M contracts, loan agreements, shareholder agreements, tax filings. The panels convert sunlight into electricity; the paper converts electricity into a legally enforceable cash flow; and the discounted value of that cash flow, adjusted for everything that could go wrong with the paper, is the price. A megawatt with bad paper is not a discounted megawatt; past a certain point it is scrap on a roof you may not even have the right to access.

I have spent twelve years on the sell side, buy side, and occasionally the confused-bystander side of West African energy transactions, and the pattern repeats. Deals rarely die because the solar resource was overstated or the inverters were the wrong brand. They die, or bleed out through price chips, because of clauses written years earlier by people who were optimising for signature speed, not exit.

This guide is an attempt to be exhaustive about those clauses. It is written with a C&I bias, because that is where I live and where the portfolio problem is most acute, but utility scale and mini-grids get their own sections, because they fail differently.

How bad paper shows up in a deal

Before the catalogue, it helps to understand the mechanics. A buyer’s diligence team (lawyers, technical advisers, financial advisers, increasingly an ESG adviser too) produces a red flag report. Every defect in that report is triaged into one of five buckets, and knowing the buckets tells you how expensive each defect is:

1. The price chip: quantifiable risk gets priced. A portfolio where 30% of revenue sits with offtakers who can walk away in 90 days is not worthless; it is worth what a 90-day revenue book is worth, which is a lot less than what your model says a 15-year revenue book is worth.

2. The condition precedent: fixable defects become CPs to closing (obtain this consent, stamp that lease, register this charge). Each CP adds months, and time kills deals, because currencies move, boards change and rival sellers appear.

3. The indemnity, escrow or holdback: unquantifiable but bounded risk (a possible tax reassessment, a threatened dispute) gets covered by holding back part of your price, sometimes for years. Warranty and indemnity insurance, which softens this in developed markets, remains thin on the ground for African C&I portfolios, so the escrow comes out of your pocket.

4. The perimeter exclusion: individually rotten assets get carved out of the transaction. You keep the problem children, which is exactly as fun as it sounds.

5. The walk: risk that cannot be quantified, bounded or excluded ends the process. Walks rarely trace back to a single catastrophic clause. They happen when the buyer realises, somewhere around week six, that the data room is a museum of improvisation and nothing in it can be relied upon.

Legal diligence practitioners describe the same dynamic from the other side: sparse or inconsistent data rooms do not merely slow a transaction, they broaden the red flag classification across every workstream, because a buyer who cannot verify one thing stops believing everything.

Standard renewable M&A checklists (Enverus publishes a representative one, and FTI describes the platform-level version) all converge on the same short list of value drivers: offtake quality, site control, permits, construction and equipment paper, and financing terms. So that is the order I will take them in, with the African complications attached.

The offtake contract: the crown jewel and the crime scene

The PPA (or solar lease, or energy service agreement; the label matters less than the economics) is 70 to 90% of the value of a C&I asset. It is also where the worst drafting lives, because early-stage developers negotiate PPAs under commercial pressure against procurement departments whose bonus depends on extracting concessions. Every concession you made in year two gets repriced against you in year eight.

1. Termination for convenience without a real make-whole

The single most value-destructive clause in African C&I solar, and the most common. The customer negotiates a right to exit the contract early, “for convenience”, on 60 or 90 days’ notice, against payment of a termination sum. The termination sum is where the crime happens. I have seen buyout schedules set at depreciated book value of the equipment, at “outstanding equipment cost” with no definition, at a flat percentage of remaining invoices, and, memorably, at a number “to be agreed in good faith at the time”. None of these covers the net present value of the remaining contracted cash flows plus debt breakage costs, which is the only definition a buyer or lender respects.

The consequence is mechanical: a rational buyer values your contract not at the tariff line, but at the lower of the tariff line and the buyout schedule, because the buyout schedule is what the offtaker will pay if it is ever cheaper to leave. If your buyout table dips below the outstanding project debt at any point in the term, you have drafted a contract with embedded negative equity. Lenders’ lawyers find this in about eleven minutes.

What good looks like: a buyout table appended as a schedule, calculated as the greater of (a) NPV of remaining payments at a stated discount rate and (b) outstanding senior debt plus breakage plus a stated equity return, updated for any capex additions, with no “good faith” language anywhere near it.

2. Assignment and change of control clauses that hand your customer a veto over your exit

This is the clause nobody reads at signing and everybody reads at exit. Two separate mechanics matter. First, assignment: whether the SPV can transfer the contract itself (relevant in asset deals and refinancings). Second, change of control: whether a transfer of shares in the SPV, or in the holding company two levels up, requires the offtaker’s consent. Badly drafted C&I PPAs require written consent for any direct or indirect change in shareholding, with no deemed-consent mechanism, no reasonableness standard, and no carve-outs.

Now multiply. Starsight alone had over 656 sites across Nigeria, Kenya and Ghana at the time of its merger announcement. A platform sale over a portfolio like that, on bad assignment language, requires hundreds of individual consents from procurement departments who have zero incentive to hurry and every incentive to reopen pricing in exchange for a signature. Buyers call this hold-up risk, and they price it savagely, or they restructure the deal to close over the missing consents with indemnities you will be paying against for years.

What good looks like: consent not to be unreasonably withheld or delayed; deemed consent after 15 to 30 days’ silence; express carve-outs for transfers to affiliates, to lenders enforcing security, and to any transferee meeting objective criteria (a stated net worth, a stated MW of operating experience); and change of control defined at the SPV level only, not at the ultimate-parent level, so that a platform-level transaction does not trip 200 site-level triggers.

3. Offtaker credit that exists only in the logo

The African C&I model was built on a seductive syllogism: multinationals are creditworthy, our offtakers are multinationals, therefore our revenue is creditworthy. Except you did not contract with the multinational. You contracted with a locally incorporated subsidiary whose balance sheet you never saw, with no parent company guarantee, no letter of comfort, and a payment security package consisting of one month’s deposit that was quietly consumed by an invoice dispute in 2021. Buyers unwind the syllogism in the first week of diligence. They sort your revenue book into: investment-grade parent paper (guaranteed), local blue chip, SME, and “aspirational”. Each tier gets its own default assumption. The absence of any security (deposits, standby letters of credit, escrowed collections) moves an offtaker down a tier; a clean 36-month collections history moves it up. Which is why your billing and collections ledger, reconciled and exportable, is worth real money at exit. A portfolio with a documented 98% collection rate is an asset class. A portfolio where collections live in a WhatsApp thread with the finance manager is an anecdote.

4. Currency clauses that were stress-tested against the wrong decade

Most West African C&I contracts price in local currency because offtakers earn in local currency. The question is what happens when the currency does what West African currencies do. Nigeria answered it in June 2023, when the new administration collapsed the managed exchange rate windows and let the naira float, dropping it 36% in a single day, from the mid-400s per dollar to 750, on its way to considerably worse. Ghana’s cedi had its own 2022 annus horribilis. For a dollar-funded portfolio with naira tariffs, the FX clause stopped being boilerplate and became the income statement.

The defect spectrum runs from fatal to merely painful. Fatal: pure local-currency tariffs with CPI-only escalation, funded with dollar debt (the equity is simply gone; buyers will value the asset on a restructured-tariff basis, if at all). Painful but survivable: USD-indexed tariffs with a capped adjustment, or indexation to an official reference rate that lags or diverges from the rate at which anyone can actually source dollars. The structurally sound version is in Nigeria’s bankable utility-scale documentation, where payments are calculated in US dollars but payable in naira, indexed to the exchange rate on the payment date, so the FX risk sits with the party best able to pass it through.

A perfectly drafted USD-indexed tariff, though, is only as good as your offtaker’s ability to absorb a 250% naira-terms price increase without defaulting, disputing or demanding renegotiation. Post-float Nigeria has been a live experiment in this. Industry analysis after the float estimated that the currency move alone could raise C&I project costs by around 40% and the resulting cost of electricity by 27%, sustainable only because diesel, the alternative, is itself dollar-linked. Buyers therefore diligence two things: the indexation clause as written, and the collections data showing whether it was actually applied and actually paid. A contract you never dared enforce is, for valuation purposes, the contract you actually have, not the one you signed.

5. Volume risk dressed up as partnership

C&I tariff structures in the region come in three broad flavours, in descending order of acquirability: fixed tariff per kWh with a minimum offtake or capacity charge; fixed tariff with no minimum (you carry consumption risk); and “discount to grid or diesel” pricing, where your tariff floats at, say, 15% below whatever the customer would otherwise pay. That last structure sold beautifully in 2019 and is a valuation horror in 2026, because your revenue is now short the grid tariff, short diesel, and short the regulator, with no floor. When a market reform or a fuel-price collapse moves the reference price, your cash flow moves with it, and no buyer will underwrite an uncapped short position they cannot hedge.

Adjacent defects in the same family: no deemed-energy clause when the offtaker’s site is unavailable (factory maintenance shutdowns become your production losses); no protection on relocation or closure of the site (the customer’s business risk becomes your stranded-asset risk); ambiguity on excess energy, metering standards and meter-failure fallbacks (every ambiguity eventually becomes an invoice dispute, and invoice disputes become the collections data that poisons your valuation); and contract volumes that ignore panel degradation, so that year-12 you is contractually committed to year-1 output.

6. The paper trail itself

The final offtake defect is incompleteness itself. Unsigned annexes. Tariff schedules marked “to be agreed” that never were. Amendments evidenced by email chains, or worse, by conduct (”we agreed the new rate on a call and they have been paying it”). Contracts that expired in 2023 and rolled over into an undocumented holdover. Side letters granting discounts that live in one sales director’s inbox. In Nigeria this shades from sloppy into legally dangerous, because an unstamped agreement is inadmissible in evidence until stamped, with penalties for late stamping under the Stamp Duties Act. In practice the buyer’s counsel cannot even rely on your contract as evidence of itself. Every one of these gaps costs more to fix during a transaction, under time pressure and adversarial scrutiny, than it would have cost to fix on a quiet Tuesday in the year it happened.

Site control

C&I solar has a structural quirk that developed-market templates never contemplated: very often, the entity buying your power does not own the building your panels sit on. The factory is leased. The landlord never signed anything. Your entire site right is a clause in the PPA saying the offtaker “grants access to the roof”, which is a right the offtaker may not have had to give. Now run the failure cases a buyer’s counsel will run.

The offtaker’s lease ends in year 6 of your 15-year PPA. The landlord’s bank enforces its mortgage over the building and the receiver regards your array as a fixture of the property. The offtaker relocates and you discover your removal rights, reinstatement obligations and continued access rights after termination were never drafted.

A proper site package for a rooftop asset is boring and short: a direct licence or lease from the building owner (not just the tenant), coterminous with or longer than the PPA, with access rights that survive PPA termination, a waiver or subordination from the landlord’s financiers where they exist, and clarity that the system remains your movable property and not a fixture.

Ground-mounted C&I and captive projects import all of the region’s land-law complexity. In Nigeria, the Land Use Act vests land in state governors, and Section 22 makes it unlawful to assign, sublease or mortgage a statutory right of occupancy without the Governor’s consent first obtained; transactions done without it are voidable, and instruments that are unstamped or unregistered cannot be registered and are inadmissible as evidence. Consent processes take months and cost percentage points of assessed value, which is why so many portfolios are sitting on unperfected leases that “everyone was going to sort out later”. Ghana adds stool and family land, with the associated risk that the signatory who took your money did not bind the family; Kenya adds county-level approvals; francophone West Africa adds its own titling regimes. None of this is exotic to a buyer who works the region, but all of it must be paper-perfect or priced.

The standard buyer responses are a CP to perfect title (adding three to six months to your deal) or an indemnity backed by escrow. The worst version is the ground lease whose term is shorter than the PPA it supports, which I keep seeing, and which converts your terminal value into a question mark.

Permits

Regulatory diligence in African C&I is a moving target, because the frameworks themselves keep moving; the discipline is to be current, documented and conservative at every threshold. Nigeria is the worked example. The Electricity Act 2023 replaced the 2005 framework and allows licence-free generation up to 1 MW per site and distribution up to 100 kW, while the constitutional amendment let states build their own electricity markets: where a state has established its own regulator, NERC’s role recedes to interstate matters, and by mid-2026 15 of Nigeria’s 36 states had created their own power regulators.

For a portfolio owner this is a full compliance matrix: an asset compliant under federal rules in 2022 may now sit in a state market with its own licensing, tariff and reporting regime, and the transition paperwork is precisely the kind of thing that never got done.

The recurring C&I defects: the 1.2 MW plant that was described as “below threshold” because someone measured the inverter rating creatively; the phased expansion that crossed a licensing line in phase 3 and nobody noticed; captive generation permits never obtained because “we are behind the meter”; missing environmental permits and building approvals; and wheeling or third-party supply arrangements entered into without the trading or supply licence the structure actually required.

The buyer’s problem with permit gaps is that the downside is not a fine you can price; it is the theoretical unenforceability of the revenue contract sitting on top of the unpermitted asset. Unquantifiable downside is bucket five. That is the walk.

Construction and equipment paper

Operating C&I portfolios in the region were mostly self-built: the platform was its own EPC, procurement arm and O&M contractor. Efficient, and fatal to the warranty stack. There is no arm’s-length EPC wrap, no defects liability period enforceable against anyone who is not you, no performance ratio test conducted at commissioning (so no contractual baseline for degradation claims), and no liquidated damages regime that ever bound anyone. Where there was a third-party EPC, it was frequently a related party at a related-party price, which invites the buyer to restate your capex, and therefore your asset base, and therefore your price.

Equipment paper has its own traps. Manufacturer warranties on modules, inverters and batteries are typically personal to the original purchaser and require consent to assign; a share sale may preserve them, an asset reorganisation may void them. Battery warranties are conditional on operating envelopes (cycling, depth of discharge, temperature) that you must be able to prove you observed, which loops back to monitoring data; a battery warranty without the telemetry to demonstrate compliance is a decorative PDF. Serial-number traceability matters twice over: once for warranty claims, and once because institutional buyers, especially those with DFI capital in their chain, now run supply-chain provenance checks on modules as standard, and “we bought whatever the Lagos distributor had that quarter” is not an answer that survives an investment committee.

Data and operations

A megawatt without data is a rumour. Buyers underwrite yield from actuals: monitored generation history against irradiance, availability records, curtailment logs, O&M tickets, and a billing ledger that reconciles metered energy to invoices to cash. Where that exists, they will finance your P50 with confidence and banks in the more mature corners of the market are now comfortable with 18-year tenors against private PPAs. Where it does not exist, they underwrite a theoretical yield with punitive haircuts, or decline the asset. Missing SCADA history, estimated billing, uncalibrated meters and undocumented downtime do not make your portfolio look scrappy and entrepreneurial. They make it unauditable, and unauditable is a valuation, and the valuation is low.

The debt stack

Project and platform debt is paper too, and it is paper with tripwires. The classic: your facility agreement’s mandatory prepayment on change of control, which converts your exit into your lender’s exit, on your lender’s terms, including make-whole and breakage costs nobody modelled. DFI-sourced debt (and in African C&I, most institutional debt has a DFI somewhere in it) adds environmental and social covenants, action plans and reporting obligations that travel with the loan and that a buyer inherits; undisclosed E&S non-compliance is a modern deal-killer because the buyer’s own capital providers will not waive it. Then the quieter defects: shareholder loans advanced over the years with no loan agreements (creating tax exposure on imputed interest, withholding tax questions, and subordination ambiguity in any enforcement); security never perfected, charges never registered within statutory windows at the companies registry, leaving your lender unsecured in fact and furious in diligence; negative pledges in one facility that conflict with security granted under another; and cross-default clauses knitting every SPV in the group into a single point of failure. A buyer can live with debt. A buyer cannot live with a debt stack whose true terms can only be established by forensic archaeology.

The corporate layer

Everything above can be immaculate at asset level and the platform can still be unsellable, because platform buyers acquire shares, and shares carry the entity’s whole history. The recurring defects, from the merely embarrassing to the fatal: contracts signed by the parent company while the SPV operates the asset, with no novation, so the revenue and the asset live in different legal persons; board and shareholder approvals absent for the very contracts being sold; the share register and the companies-registry filings telling different ownership stories; equity promised to early employees by email and never papered, surfacing mid-transaction as a claim; shareholder agreements with pre-emption rights but no drag-along, handing a 6% holder a veto over your exit; and unresolved founder separations that everyone stopped mentioning but nobody documented.

Then tax. Payroll and pension arrears; withholding tax on cross-border service and interest payments never remitted; VAT treatment of the PPA (is it a supply of goods, services or a lease? the characterisation drives the invoice, and the invoice history is discoverable); transfer pricing on all those intra-group EPC and O&M charges; and capital gains exposure on the transaction itself. One structural mercy in Nigeria is that share transfer instruments attract only nominal stamp duty, one reason regional deals are overwhelmingly share deals rather than asset deals; the corollary is that share deals inherit everything, so the corporate hygiene you skipped is now the buyer’s problem, and the buyer will make it your price’s problem.

Governing law and disputes

International buyers discount contracts that resolve disputes exclusively in local courts. The reason is arithmetic, not disrespect: enforcement timelines measured in years change the present value of every remedy in the document. The bankable default remains arbitration with a neutral seat under recognised rules, in a jurisdiction whose awards travel well under the New York Convention, paired with governing law the parties’ counsel can actually opine on. The C&I-specific defects are mundane: dispute clauses copied across templates with the wrong seat or contradictory law-and-forum pairings; multi-tier clauses (negotiation, then mediation, then arbitration) with broken or impossible timelines; and asymmetric clauses that let the offtaker sue anywhere while confining you to one forum. Nobody reads these until the first serious invoice dispute, at which point they are the whole game.

The utility-scale detour

Utility-scale PPAs in Africa fail less through sloppy drafting and more through counterparty and sovereign dynamics, and the last seven years have been one long demonstration. The core structural problem is the single buyer: a state utility with a weak balance sheet, which is why bankable projects historically stacked sovereign guarantees, partial risk guarantees and MIGA-style political risk cover on top of the PPA before lenders would engage. When that support architecture is missing or ambiguous (a “letter of comfort” doing the work of a guarantee, termination compensation drafted vaguely, no deemed-energy protection against grid curtailment), the project may still get built in a good year, but it becomes very hard to sell, because the secondary buyer underwrites the documents, not the ribbon-cutting.

And even signed sovereign paper gets repriced when the fiscal arithmetic breaks. Ghana contracted so much take-or-pay thermal capacity that it was paying over $500 million a year for electricity it did not use, nearly $1 billion across two years by the government’s own account, prompting the Energy Sector Recovery Programme, a declared shift to take-and-pay contracting, and direct renegotiation with a dozen IPPs and their lenders; the Energy for Growth Hub’s case study of Ghana’s 32 PPAs is the best public autopsy of how uncoordinated contracting became a fiscal crisis. Kenya ran the same movie with different subtitles: a Presidential Taskforce on PPAs in 2021, a moratorium on unconcluded agreements, renegotiation pressure justified by findings that IPPs supplied 37% of energy but took 59% of power purchase costs, and a freeze that was only formally lifted by the National Assembly in November 2025, with a pivot toward competitive auctions, after years in which law firms openly warned the process was damaging Kenya’s standing with private capital. Nigeria’s cautionary tale is quieter: the 975 MW of utility-scale solar PPAs signed with NBET in 2016 that then spent years unable to reach financial close as tariff, guarantee and FX questions went unresolved.

For an acquirer, four lessons. First, the offtaker’s credit and the sovereign support package are the asset; diligence them as hard as the tariff. Second, renegotiation risk is real even for signed contracts, so contracts drafted with clear compensation mechanics, deemed energy, and robust arbitration survive renegotiation waves with their value intact, while vague ones get rewritten. Third, curtailment without deemed-energy protection converts grid weakness into your equity loss. Fourth, transferability again: government consents and support instruments (guarantees, put-and-call option agreements, political risk policies) must be assignable to a new owner, or your exit needs a minister’s signature, and ministers change.

The mini-grid detour

Mini-grids are the inverse of C&I: instead of one contract with one creditworthy offtaker, you have a regulated tariff over thousands of retail customers, plus a lattice of permits, community agreements and grant contracts. The acquirability problems follow. Tariffs are set or approved by regulators (in Nigeria via cost-based methodologies under NERC’s framework), so tariff-review risk replaces offtaker-credit risk. The historic existential fear, the main grid arriving and stranding your asset, has been progressively addressed in Nigeria: the framework has moved from the 2016 rules through the 2023 update to the Mini-Grid Regulations 2026, which raised capacity thresholds to 5 MW for isolated and 10 MW for interconnected mini-grids with a unified licence and registration-only treatment up to 100 kW, require 12 months’ written notice from a DisCo before grid extension into a mini-grid area, and define compensation on asset transfer as the higher of indexed historical cost net of depreciation and net depreciated replacement cost, plus unrecovered development costs and 12 months’ revenue where grid arrival occurs in the first decade of operation. That is a genuinely bankable protection, and a buyer will check whether each site in your portfolio actually holds the permit class that unlocks it, because unregistered or informally operated sites do not.

The rest of the mini-grid diligence pain is portfolio heterogeneity. Fifty sites developed over eight years under three regulatory regimes, four grant programmes and two management teams means fifty slightly different permits, community land arrangements of varying formality, plus grant and results-based-financing agreements whose fine print contains clawbacks and change-of-control consent rights held by the funder. Selling a subsidised portfolio without the grant provider’s consent can convert grant income into a repayable liability at exactly the wrong moment. Sector coordination is improving (Nigeria’s REA and the Africa Minigrid Developers Association have formalised cooperation on this kind of standardisation), but for now the seller’s job is the unglamorous one: a site-by-site matrix of permit class, tariff basis, land instrument, grant conditions and customer-level revenue data, maintained as if a buyer will read it, because one will.

A brief word on South Africa

South Africa’s contract stack grew up under different rules. Liberalisation from 2021 unlocked wheeling, and the market promptly grew a layer of licensed traders and aggregators between generators and offtakers, with Eskom’s virtual wheeling platform extending the model to multi-site, lower-voltage customers and rapid growth in trading licences issued. The acquirable asset there is increasingly a stack of contracts (generator PPA, trading agreement, wheeling and use-of-system agreements, municipal frameworks), as in the Mpact deal, where a paper producer buys solar via a licensed trader wheeling from a SolarAfrica plant, or NOA’s aggregation platform signing a 20-year PPA with a healthcare group across six Eskom-connected sites. The diligence question mutates accordingly: not just “is this PPA bankable” but “does every link in this chain (trader licence, wheeling agreement, municipal framework, settlement mechanics) survive a change of control simultaneously”. Back-to-back risk is the new termination risk. The SALGA municipal wheeling status report is the reality check on how uneven the municipal layer still is.

What good looks like

Everything above reduces to one habit: draft and file, from the first contract, as though a buyer’s counsel were already reading. The checklist I wish someone had handed me in year one:

• Run one PPA template and version it ruthlessly, with every deviation logged in a register alongside the commercial reason. A buyer who can diligence one template plus a deviations schedule prices a portfolio in weeks; a buyer facing forty bespoke documents prices it in quarters, against you.

• Engineer assignability in at signing: deemed consent, permitted-transferee carve-outs, lender step-in, change of control defined at SPV level. The clause costs nothing on day one and everything at exit.

• Attach a buyout schedule to every contract that is a genuine make-whole, never below debt plus breakage plus a stated equity return, with no good-faith placeholders.

• Perfect the boring paper in real time: stamp, register, obtain consents, file charges, novate contracts into the right entity in the year it happens, not the quarter you sell.

• Keep a contract register and consent matrix as living documents: every agreement, counterparty, term, consent trigger and security interest in one place, reconciled to what is physically in the data room.

• Reconcile metering, billing and collections monthly, and keep the output exportable. Your collections history is your credit story; treat the ledger as an asset under management.

• Once a year, read your own portfolio the way a hostile buyer’s counsel would, and fix what you find. This is cheaper in every year except the year you sell, when it is priceless.

None of this is intellectually difficult. All of it loses to the urgent, every week, for years, which is why clean paper commands a premium: it is scarce because discipline is scarce. The developers who got bought (Daystar by a supermajor, Starsight and SolarAfrica into a pan-African platform) were not the ones with the shiniest panels. They were the ones whose paper could survive being read by strangers with an incentive to find problems.

The best time to draft for your exit was the day you signed your first PPA. The second-best time is before a buyer’s counsel opens your data room with a highlighter, a checklist, and a professional obligation to assume the worst.

- S

The Impostor is not your lawyer, and given what you have just read about his early contracts, you should be relieved.

P.S - Nothing above is legal advice. It is scar tissue, arranged by section heading. Proper lawyers exist, they charge by the hour, and the hours are cheapest years before a transaction and ruinous in the middle of one. Book them in the boring years.

P.P.S - The views are mine alone. Not my employer’s, not any lender’s credit committee’s, and not those of the counterparties who may recognise a clause or two above. Any resemblance to contracts living or dead is, regrettably, not coincidental.

P.P.P.S - Three of the defects above are autobiographical. I will not be indicating which three, but the counterparties know, and one of them still sends a Christmas card.

P.P.P.P.S - No confidential deal information appears above. Every example has been sanded down to the point where four different former colleagues will each be privately certain it is about them. It is about all four.

P.P.P.P.P.S - A reader will ask why I do not simply publish a model PPA. Because a template without the discipline is just a nicer museum. Also because my lawyer, who is real and long-suffering, asked me not to.

P.P.P.P.P.P.S - The consent matrix and the deviation register are spreadsheets, not software. You could build both this week. You will not, but you could.

Sources and further reading

The heavyweight documents behind this piece, for the sufficiently masochistic: the Energy for Growth Hub case study of Ghana’s PPAs; the Kenya Presidential Taskforce report on PPAs and Bowmans’ running commentary on the moratorium and its lifting; CrossBoundary on PPA bankability beyond sovereign guarantees; the GCF funding proposal for Nigeria’s solar IPP programme, which doubles as a field guide to Nigerian power-sector risk allocation; commentaries on Nigeria’s Electricity Act 2023 and the Mini-Grid Regulations 2026; AFSIA’s Africa Solar Outlook; and the SALGA report on municipal wheeling in South Africa.

I keep my promises. Because Template 2 is a chip company that looked at its customers, decided they were too poor to buy its chips, and responded by becoming their investor, their anchor customer, and eventually something that looks suspiciously like their bank.

If that sounds like financial engineering with extra steps, hold that thought. We will get to the extra steps.

But first, a joke.

Revenue is a Mindset 😂

Revenue, Kuveke explains, is a mindset.

• OpenAI commits roughly $300 billion of compute spending to Oracle for cloud capacity (WSJ, Reuters).

• Oracle buys mountains of Nvidia GPUs to serve it for tens of billions.

• Nvidia announces a letter of intent to invest up to $100 billion in OpenAI (Reuters, Bloomberg).

Money out the front door, orders in through the back, and valuations rising on every corner of the triangle. Michael Burry compared it to Enron and Lucent. Jim Chanos, the man who actually shorted Enron, warned about "arcane financial structures" stacked on top of money-losing companies.

The money goes around the circle, everyone’s stock goes up.

Bloomberg ran a whole feature on the circularity of these deals and whether the trillion-dollar AI boom is partly financing itself (Bloomberg). Serious analysts are genuinely worried, and I want to be careful here: circular financing in AI is an open analyst concern, not an adjudicated fraud. Nobody has been marched anywhere in handcuffs. The deals may all prove out. But the structure of the worry gives us a diagnostic, which is the most useful sentence in this essay:

When the money stops cycling, what remains?

If the answer is “nothing,” you have Kuveke’s mindset revenue. If the answer is “a developer ecosystem, a manufacturing base, standards everyone builds on, and customers who keep paying real money for real things,” you have something else entirely. You have bootstrapped an ecosystem into existence using your own balance sheet as scaffolding.

Meet NVIDIA

Some orientation, for anyone who last checked in on Nvidia around 2016. Nvidia designs the chips that train and run AI models, and it currently trades at around $5.2 trillion. Africa’s entire annual output is roughly three trillion. One company, priced above the yearly production of 54 countries and one and a half billion people.

Sit with that for a moment. I had to.

Nvidia, whatever you think of its stock price, has been running the second play for nearly twenty years. And I think it is the most complete template that exists for building the energy company Africa actually needs.

This is not a copy-Nvidia essay. You cannot copy Nvidia. You do not have $4 trillion of market cap, and if you did you would not be reading a Substack called The Impostor’s Guide to Clean Energy.

This is an essay about extracting the operating logic underneath Nvidia and asking what it looks like when the chips are transformers and meters, the developers are gensets and mini-grids, and the infinite money glitch has to survive contact with a Ghanaian utility’s collections department.

The disclaimer from last time carries over. Nobody on this continent needs to cosplay as the Jensen Huang of Ogbomosho. We are here to study the architecture, not to buy leather jackets.

The five things Nvidia actually did

Nvidia’s playbook is five moves, executed in sequence, over two decades.

1. Move one: build the platform before the demand exists.

   In 2006 Nvidia unveiled CUDA, a software layer that let programmers use graphics chips for general-purpose computing. Almost nobody cared. Wall Street treated it as an expensive hobby for the better part of a decade. Nvidia kept paying for it anyway: libraries, compilers, university curricula, developer tools, whole conferences.

   Then deep learning arrived, discovered that graphics chips are accidentally perfect for training neural networks. Today CUDA counts roughly six million developers, every major AI framework treats it as the default dialect, and Nvidia books something like 86 percent of all data-centre GPU revenue. Rivals ship silicon with comparable specs on paper and lose anyway, because two decades of accumulated software means the Nvidia chip performs better in the field, and every engineer already speaks its language.

   Readers of the Citadel essay will recognise this move. It is the map. Nvidia drew the map of accelerated computing before there was any traffic to put on it, and now all the traffic consults the map.

   CUDA looked like an expensive hobby for a decade and a half before it paid. As an Arsenal supporter, I am professionally trained to respect a long wait for silverware.
   

2. Move two: sell the stack, never the component. Nvidia does not really sell you a chip anymore. It sells you a rack, the networking, the software, the orchestration, the whole integrated system, which is how a hardware company sustains gross margins above 70% in an industry where components get commoditized for sport. Value lives in integration. Component sellers compete on price; platform sellers compete on switching costs.
   

3. Move three: manufacture your own demand side. By early 2026 Nvidia had deployed over $40 billion in equity into its own ecosystem, including a stake in OpenAI that ended up around $30 billion, plus positions in xAI, CoreWeave, a data-centre developer or two, and, my personal favourite, a glass company. Corning. They invested in glass!!!. (I guess when you are underwriting an entire supply chain, you eventually reach the sand), and a small nation’s worth of other companies whose common feature is that they buy or enable the purchase of Nvidia products. Your customers cannot afford your product? Fund the customers. It sounds circular because it is circular. The question, as always, is what remains when the circle stops.
   

4. Move four, the strange one: become the credit enhancer for your own ecosystem. This is the move that made me sit up. The neoclouds, companies like CoreWeave that buy GPUs and rent them out, had a problem: lenders had no idea what a GPU would be worth in four years, so they priced the debt like it was radioactive, north of 10% unsecured.

   Nvidia’s response was to backstop them. In September 2025 it signed an agreement to buy up to $6.3 billion of CoreWeave’s unsold cloud capacity through April 2032 (Reuters, CNBC). A take-or-pay floor under utilization. With Nvidia guaranteeing the downside, the same borrowers could suddenly raise secured debt around SOFR plus 225 basis points, call it 5.9%. Sit with that number.

   The backstop compressed the cost of capital by four to five percentage points. For an infrastructure business, that is the difference between a viable project and a PowerPoint. And here is why Nvidia could rationally do it when no bank could: Nvidia is the only party on Earth that actually knows what GPUs are worth over time. It sees utilization, resale values, failure rates, the roadmap of the next chip that determines the residual value of the current one. Asymmetric information usually shows up in finance as a reason deals die. Nvidia weaponized it in the other direction. It sold its superior knowledge of its own ecosystem as a guarantee, and got paid for it in ecosystem growth.
   

5. Move five: know when the scaffolding comes down. In February 2026, the famous $100 billion OpenAI letter of intent quietly became a $30 billion investment (The Guardian), with Jensen Huang clarifying that the original figure was never a binding commitment (Yahoo Finance). You can read that as retreat. I read it as discipline. Scaffolding is supposed to come down. The companies that died doing this play died because they forgot that part, and we will meet the most famous corpse shortly.

Now I am supposed to shout AI bubble

The most useful book I studied on this was Carlota Perez’s Technological Revolutions and Financial Capital, published in 2002. Perez studied five technological revolutions across two centuries and found the same rhythm every time: an installation period, when speculative finance floods in and overbuilds the new infrastructure amid a bubble; a crash; then a deployment period, when the surviving infrastructure gets put to work by everyone and the actual golden age happens. The railway manias bankrupted investors and left railways. The dot-com crash vaporised trillions and left the fibre that later carried YouTube, and eventually this newsletter to your inbox. Bubbles inside a real technological revolution are wasteful, unfair, and strangely productive. Tulip manias leave nothing. Fibre manias leave fibre.

So my answer on AI is that parts of it will end in tears, and the tears will water something. When the music stops, the test is simple and brutal: look at what remains. Round-tripped revenue leaves a bigger number and a court date. Overbuilt infrastructure leaves infrastructure.

Keep that test in your pocket.

We are about to use it on my side of the world, where the collateral does not rot in three years, and where I will now argue this entire playbook works better than it does in California.

How do all these analogies actually apply to African Energy

In Silicon Valley, revenue is a mindset. In African energy, revenue is a collections problem.

When OpenAI commits $300 billion to Oracle, nobody asks whether OpenAI will physically refuse to pay the invoice while continuing to consume the compute. In African power markets, that scenario has a name. Several names, actually, and case files.

Ghana signed a stack of take-or-pay power contracts in the mid-2010s, then demand undershot and the state found itself paying for capacity nobody used. The distribution utility ECG recovers only around 62% of what it bills, the sector accumulated a shortfall of roughly $12.5 billion between 2019 and 2023, and at one point Karpowership, the Turkish company that literally parks a ship full of generators off your coast, was owed over $370 million. A country was in arrears to a boat.

Nigeria’s version: the bulk trader NBET spent years paying generation companies around 35% of their invoices. The accumulated debt to gencos ran into trillions of naira, and in January 2026 the government moved to settle part of it with a ₦501 billion bond issue, which is what it looks like when a sector’s receivables get so bad they have to be converted into sovereign paper to be believed.

So no, you cannot run Kuveke’s play here. Two Lagos startups buying each other’s services for a million dollars a month would not produce $12 million of ARR. It would produce two unpaid invoices and a very awkward wedding, because the founders are probably cousins.

But look at what those horror stories actually did to the market. They taught every international lender that African power revenue is unknowable, and lenders price the unknowable brutally. That pricing shows up in one number, and that number is the entire game.

The cost-of-capital spread

A 2025 study in Applied Energy by Dato, Dioha, Hessou and colleagues, working with the Clean Air Task Force across 48 African countries, put the average weighted cost of capital for African power projects at 15.6%, against 5.1% in the United States, 6.6% in China, and 2.4% in Japan (Applied Energy, CATF).

Ghana reaches 20.5 to 22.9%. The sun in Ghana is magnificent. The discount rate is a war crime.

This is why the continent with 20% of the world’s population attracts under 3% of global energy investment, roughly $110 billion of a $3.4 trillion total (IEA World Energy Investment, Africa), while around 600 million people remain without electricity and the World Bank and AfDB scramble to connect 300 million of them under Mission 300.

Solar panels cost broadly the same everywhere. Money does not. The thing we do not have is cheap money

What you actually build

Here is the five-move sequence, translated.

• Stage one, the platform before the demand. Build CUDA for electrons. Before the serious capital arrives, someone has to build the instrumentation layer nobody wants to pay for. Meters on factories, cold rooms, telecom sites and mini-grids. Payment histories. Load profiles. Diesel burn. Outage logs. Degradation curves by panel brand, by battery chemistry, by city. The last essay called this the map and spent four thousand words on it, so today I will only add the Nvidia corollary: the map must be built before the market arrives, it will look like an expensive hobby for years, and that is precisely the point. Moats grow in the dark.

• Stage two: sell the stack. Do not sell panels. Nobody here buys components either. The African industrial customer does not want panels. He wants his factory to run through the afternoon and his diesel bill to die quietly. The winners already sell exactly that: power as a service, financed, installed, operated and metered under one contract. Daystar Power was doing it across West Africa when Shell bought the whole company in 2022, and its installations now include a 4.2 megawatt system at Nigerian Breweries and a string of Nestlé plants across three countries. CrossBoundary Energy runs the model continent-wide with an awarded portfolio around $700 million, including my favourite deal in African energy: 222 megawatts of solar and a 526 megawatt-hour battery built to feed the Kamoa-Kakula copper mine in the DRC thirty megawatts of clean baseload, around the clock, in one of the hardest operating environments on earth.

  
  Recurring kilowatt-hour revenue against contracts that run ten to twenty years, instead of one-off construction margins. Teece's law again: the profit pools around the complementary assets, the financing, the operations, the relationship, and never around the hardware.

  
  Sell electricity, uptime, cold storage that works, milling that happens, an outcome with an SLA. Integration margins survive commoditization. Component margins do not.
  

• Stage three: manufacture the demand side. Nvidia’s equity stakes translate here as anchor offtake. Nvidia converts speculative chip demand into bankable demand by writing cheques into its own customer base. The African energy version is less exotic and frankly more elegant: the anchor offtake agreement. A fifteen-year contract with a brewer, a cement plant, a mine or a tower company, some entity with a hard-currency parent and a treasury that answers emails, does for a power project precisely what Nvidia's $30 billion does for OpenAI. It converts a speculative asset into a financeable one before construction starts. And the newest anchor class arriving on the continent is, pleasingly, the data centre itself, with African capacity headed from around one and a half gigawatts toward two and a half by 2030. Anchor first, then aggregate the smaller loads around the anchor. Never generation first. Generation without offtake is a monument.

• Stage four, the crown jewel: become the credit enhancer. Everything above is prologue. My actual claim is that the billion-dollar African energy company is the one that runs Nvidia’s fourth move. It takes the ignorance premium, those five to seven points of WACC priced off unanswerable questions, answers the questions at portfolio scale, and stands behind the answers with a guarantee.
  
  The astonishing part is that every component of this machine already exists on this continent. They are lying in separate rooms.
  
  In Lagos there is InfraCredit, set up in 2017 by GuarantCo and Nigeria’s sovereign wealth authority, which wraps naira infrastructure bonds in an irrevocable, locally AAA-rated guarantee so that Nigerian pension funds can buy them. It works well enough that it listed its own shares last year and is being cloned in Pakistan and Kenya. That is the wrapper: proof that credit enhancement turns African infrastructure into paper that institutions will hold. (InfraCredit, World Bank PPP).
  
  In Nairobi, in July 2025, Sun King closed $156 million of securitised debt backed by nothing more exotic than the repayment behaviour of about 1.4 million Kenyan solar customers, with Absa, Citi, Co-operative Bank, KCB and Stanbic taking the senior notes and development banks holding the mezzanine. It was the largest securitisation ever completed in sub-Saharan Africa outside South Africa. M-KOPA has raised over $200 million against similar collateral. That is the converter: proof that granular African payment data, gathered meter by meter, transforms directly into cheap senior capital.
  
  And in Washington, that same month, MIGA, the World Bank’s political-risk insurer, executed a $495 million guarantee framework covering up to one hundred CrossBoundary distributed-energy projects at once. (MIGA). Portfolio-level cover, replacing the old one-project-one-negotiation grind. That is the rented balance sheet: proof that the multilaterals will now wrap African energy platforms wholesale.
  
  A wrapper. A converter. A rentable balance sheet. Nvidia’s backstop, disassembled, its parts scattered across three cities, every part already functioning, waiting for the company that bolts them together.
  
  Picture that company. It operates distributed energy assets across half a dozen countries, which is move two. It has metered everything obsessively for a decade, which is move one, so it holds the continent’s deepest verified dataset on the only two questions the ignorance premium is made of: do the offtakers pay, and do the assets perform. Then comes the Nvidia move. It starts underwriting other people’s projects. It issues floor guarantees on portfolios of commercial and industrial solar the way Nvidia guarantees GPU utilisation, wrapped in DFI counter-guarantees the way InfraCredit borrows sovereign trust, and it takes a spread for standing behind numbers it alone can verify.
  
  Nvidia can write its backstop because it authors the roadmap that sets its collateral’s residual value. The African platform can write its backstop because it holds ten years of metered truth about offtakers and assets that no bank possesses. The same asymmetry, on different collateral. The bank prices fear. The platform prices knowledge. The spread between fear and knowledge is the product.
  
  And the arithmetic transfers with indecent precision. Compress a portfolio’s cost of capital from 15 percent toward 9 or 10, territory the best guaranteed African deals already reach, and you have roughly doubled the equity return on identical hardware, the same violence the backstop performed on CoreWeave’s margins. Then the flywheel closes. Every guaranteed project reports its operating data back to the platform. The data sharpens the underwriting. The underwriting strengthens the next guarantee. The guarantee attracts cheaper capital. The capital builds more projects. The projects produce more data. Around and around, widening with every turn.
  
  Now take out the test from the bubble section, because this is where Kuveke’s joke finally gets its answer. Stop the music on the fake flywheel and nothing remains except a bigger number and litigation. Stop the music here and what remains is a twenty-year power asset, metered and contracted, feeding a factory that needed it, plus a dataset that appreciates with age. Nvidia’s collateral is landfill in four years. Ours hums for two decades. The African version of this playbook rests on better collateral than the original.

  

• Stage five: sunset discipline. Write the scaffolding-removal triggers before you write the first guarantee. My proposal: if a backstopped project cannot refinance into ordinary market debt within 24 to 36 months, the demand was never real and the guarantee must shrink, not grow. Nvidia walking its $100 billion LOI back to $30 billion looked embarrassing and was actually the healthiest thing in the whole saga. The alternative is the Lucent ending, and the Lucent ending deserves its own section, because it is the ghost that should haunt every founder who reads this far and feels excited.

The ways this kills you

Lucent Technologies, the pride of Bell Labs, spent the late 1990s vendor-financing its own customers: lending money to upstart telecom carriers so they could buy Lucent equipment, with commitments that reached about $8.1 billion. Revenue looked spectacular right up until roughly 47 of those carriers went bankrupt between 2000 and 2003, at which point Lucent discovered it had been booking its own loans as demand. Revenue collapsed from around $38 billion to $8 billion and the company only survived by merging with Alcatel (American Affairs, MIT Technology Review); Tom Tunguz has a sobering (comparison of Nvidia’s exposure to the Lucent and Nortel playbooks). Same play as Nvidia. Same mechanics. Opposite ending. What separated them was discipline and information: Lucent financed anyone with a pulse and a purchase order, and its only informational edge over its customers’ lenders was superior desperation.

So, the failure modes, stated plainly, because this series has a policy of telling you how the hero dies:

1. You backstop vibes instead of data. If your guarantees are priced off optimism rather than years of verified payment behavior, you are Lucent with better weather. The whole edge is informational. No data layer, no business.

2. Your balance sheet is a rounding error at the wrong moment. Nvidia backstops CoreWeave from atop a fortress. You will be making guarantees from something closer to a bouncy castle. This is the most speculative leg of the entire thesis, and I will say it in bold so nobody accuses me of burying it: **operator data alone does not fully substitute for an AA balance sheet.** The honest version of stage four is done in partnership, your data plus InfraCredit’s or GuarantCo’s or MIGA’s capital, with you taking a sliver of first-loss and a slice of the economics. Solo heroics here are how you become a cautionary LinkedIn carousel.

3. The take-or-pay trap. Ghana’s crisis was caused by take-or-pay commitments detached from real demand. Your stage-three anchor offtakes are take-or-pay commitments. The instrument that saves you is the instrument that killed Accra’s balance sheet. Respect it accordingly, and size every floor against demand you can verify rather than demand you can forecast after two Guinnesses.

4. The commodity layer commoditizes underneath you. When DeepSeek released open-weight models at a fraction of the incumbents’ costs, the panic was instructive: if the expensive layer gets cheap, who keeps the margin? Solar-plus-storage is on the same curve; panels and cells get relentlessly cheaper. If your business is selling the commodity, the cost curve is coming for you. If your business is distribution, verified data, integration, and cheaper capital, the cost curve is your best salesperson. Cheap hardware plus expensive money is still an unbankable project. You sell the money part.

5. Political weather. A single regulation on guarantee structures, a currency crisis, a new minister with opinions about foreign-linked credit enhancers, and stage four freezes for a year. Local-currency structures like InfraCredit’s exist precisely because of this. Copy the homework.

Why this is hard to copy, and the endgame

Recall the CITADEL loop from the last post: the map tells you where to build, the assets generate returns and better data, the merchant layer monetizes the flexibility, and each turn of the loop makes the next turn cheaper. The Nvidia thesis is what happens when that loop grows a financial arm. The map and the meters produce the data. The data prices the risk. The priced risk becomes guarantees. The guarantees cut the cost of capital. The cheaper capital builds more assets, which produce more data. Around and around, and every rotation widens the moat, because a competitor arriving in year five is up against your five years of payment histories and the basis points those histories are worth, and there is no procurement process for that. Panels can be bought. Longitudinal collections data on ten thousand African offtakers cannot, at any price, on any timeline shorter than the one you already traveled.

The endgame, then. The billion-dollar African energy company will not look like a company. It will look like an ecosystem of mutually reinforcing platforms, a data layer, an integrated power operator, a portfolio of anchor demand, and a credit-enhancement arm, each one de-risking the others’ capital expenditure, the whole thing worth more than the sum because the whole thing is the reason each part can borrow. From the outside it will be confusing. Analysts will draw circular diagrams of it and mutter about related-party transactions, and the diligence question will be the same one we started with: when the money stops cycling, what remains? The answer had better be: meters that read, plants that run, mills that pay, and a loan book priced off truth.

Twenty percent of humanity, three percent of the investment, six hundred million people in the dark, and a five-to-seven-point ignorance premium sitting there like free money for whoever does the unglamorous decade of homework first. Jensen Huang built a $4 trillion company by knowing what his ecosystem’s assets were worth when nobody else did. The African version starts smaller, with a meter, a mobile-money receipt, and the radical proposition that a paying customer in Kisumu, properly measured, is a better credit than a hyperscaler’s letter of intent.

Next in the series, we go from strategy to anatomy. If the CITADEL post was the map and this one is the money, the next one is the machine: what the first vertical of this ecosystem actually looks like at ground level, and why it starts in the least glamorous corner of the market on purpose. Wear boots.

- S

P.S. - Nothing here is investment advice. If you lose money backstopping a maize mill because a Substack told you to, that is between you and your risk committee.

P.P.S. - The views are mine alone. Not my employer's, not any board's, and certainly not the investment committee that reads my actual models, who are, for the record, wonderful people with excellent judgment of whom I am not afraid at all.

P.P.P.S. - Long-time readers will suspect, correctly, that this template describes a company that looks a great deal like the one I have been building on nights and weekends. The data layer comes first. It always comes first. Ask the CUDA engineers who spent a decade being asked when they would get a real job.

P.P.P.P.S.- To be scrupulously fair to Nvidia one more time: the circular-financing worry is an analyst concern, not a finding. The commitments are mostly non-binding letters of intent, the totals are not additive, and the company may well be vindicated. Lucent, on the other hand, was adjudicated by reality. Learn from the second one while watching the first.

P.P.P.P.P.S.- Again, none of this is investment advice. I model energy projects for a living, and some of the most beautiful models I have seen ended up face down in the mud because someone forgot that customers, currencies, regulators, and batteries have free will.

Read this as research. Read it as strategy. Read it as vibes. Read it as me, once again, thinking too loudly on the internet.

Leave a comment

Sources and further reading

The meme, and the diagrams underneath it

• Robin Wigglesworth and colleagues at Bloomberg mapped the circular deals first: “OpenAI’s Nvidia and AMD deals boost $1 trillion AI boom with circular deals,” Bloomberg, October 2025. https://www.bloomberg.com/news/features/2025-10-07/openai-s-nvidia-amd-deals-boost-1-trillion-ai-boom-with-circular-deals

• Jack Kuveke’s LinkedIn oeuvre, filed under Jabroni Capital. Consume responsibly.

The Nvidia playbook

• CNBC, “CoreWeave stock jumps on disclosure of $6.3 billion order from Nvidia,” September 15, 2025 (the original backstop, obligated through April 2032). https://www.cnbc.com/2025/09/15/coreweave-stock-jumps-on-disclosure-of-6point3-billion-order-from-nvidia.html

• NVIDIA corporate blog, “The AI Compute Partnership,” July 2026 (the formalised backstop programme, co-signed by CFO Colette Kress; Sharon AI and Firmus as first adopters).

• SemiAnalysis, on AI debt markets, July 7, 2026 (backstopped pricing near SOFR plus 225 basis points against roughly 10 percent unsecured; the projection of more than $7 trillion of AI debt outstanding by 2029).

• CNBC, May 9, 2026, on Nvidia’s more than $40 billion of AI equity investments in the first four months of the year, led by the $30 billion OpenAI position.

• Wall Street Journal, January 2026, on the Nvidia-OpenAI letter of intent cooling; Bloomberg’s coverage of Jensen Huang at the Morgan Stanley TMT Conference, March 4, 2026 (”might be the last time”).

• NVIDIA’s own CUDA twentieth-anniversary materials (roughly six million developers); Mizuho estimates of data-centre GPU revenue share, via Forbes.

• Epoch AI, on hyperscaler capital expenditure crossing operating cash flow around Q3 2026; Financial Times and Goldman Sachs compilations of the $725 billion 2026 capex guidance.

The ghosts

• Lucent Technologies, Form 10-K for fiscal year 2000 (customer financing commitments of up to $8.1 billion), together with the standard histories of the 2000 to 2003 telecom carrier bankruptcies. The single best cautionary tale in vendor financing.

The theory shelf

• Carlota Perez, Technological Revolutions and Financial Capital: The Dynamics of Bubbles and Golden Ages, Edward Elgar, 2002. The book that explains why some bubbles leave railways behind and others leave nothing.

• David J. Teece, “Profiting from Technological Innovation: Implications for Integration, Collaboration, Licensing and Public Policy,” Research Policy, Vol. 15, No. 6, 1986, pp. 285-305. https://www.sciencedirect.com/science/article/abs/pii/0048733386900272

• Michael G. Jacobides, Carmelo Cennamo and Annabelle Gawer, “Towards a Theory of Ecosystems,” Strategic Management Journal, Vol. 39, No. 8, 2018 (the co-specialised complements argument, for those who want the DFI point in academic dress).

The African cost of capital

• Clean Air Task Force, “Evaluating the Weighted Average Cost of Capital (WACC) in the Power Sector for African Countries” (the 15.6 percent average across 48 countries, with some above 25 percent). https://www.catf.us/resource/evaluating-weighted-average-cost-capital-wacc-power-sector-african-countries/

• The peer-reviewed version: Dato et al., “Computation of weighted average cost of capital (WACC) in the power sector for African countries and the implications for country-specific electricity technology cost,” Applied Energy. https://www.sciencedirect.com/science/article/abs/pii/S0306261925010633

• IEA, Cost of Capital Observatory (the two-to-three times premium, and the decomposition showing the country base rate at 60 to 90 percent of WACC in developing markets). https://www.iea.org/reports/cost-of-capital-observatory

• IEA, World Energy Investment 2026, Africa chapter (the roughly $110 billion, the 3 percent share of global energy investment against 20 percent of world population, and the comparison with the more than $105 billion invested globally in data-centre energy infrastructure in 2025).

The disassembled machine

• Citi press release, “$156M Sun King Securitization to Deliver Solar for Over a Million Kenyans,” July 2025 (the largest securitisation completed in sub-Saharan Africa outside South Africa; senior notes held by Absa, Citi, Co-operative Bank, KCB and Stanbic). https://www.citigroup.com/global/news/press-release/2025/citi-sun-king-securitization-deliver-solar-million-kenyans

• MIGA, “CrossBoundary Energy C&I Africa Portfolio,” the $495 million portfolio guarantee framework covering up to one hundred projects. https://www.miga.org/project/crossboundary-energy-ci-africa-portfolio

• CrossBoundary Energy, “$200M senior debt to further renewable energy portfolio in Africa,” November 2025 (the Standard Bank-led facility, and the Kamoa Copper solar-plus-storage baseload deal). https://crossboundaryenergy.com/crossboundary-energy-secures-us200m-senior-debt-to-further-renewable-energy-portfolio-in-africa/

• PV Tech, “Shell acquires African C&I solar provider Daystar Power,” December 2022. https://www.pv-tech.org/shell-acquires-african-ci-solar-provider-daystar-power/

• InfraCredit Nigeria, established 2017 by GuarantCo and the Nigeria Sovereign Investment Authority; listed on NASD in April 2025, with sister vehicles InfraZamin in Pakistan and Dhamana in Kenya.

The cautionary tales

• Ghana’s Energy Sector Recovery Programme documentation and associated World Bank reporting (the $12.5 billion 2019 to 2023 shortfall estimate and the PPA moratorium), plus Ghanaian press reporting on Karpowership’s outstanding receivables, February 2025.

• Nigerian reporting on the ₦501 billion NBET bond issuance of January 27, 2026, and Association of Power Generation Companies statements on invoice settlement rates.

And the one that started the series

• The Impostor’s Guide to Clean Energy, “How to Build a Billion-Dollar Energy Company in Africa - The CITADEL Thesis,” June 2026. Read it first if you have not. Today’s essay stands on its shoulders

There is a small town in northern Louisiana called Haynesville. It is known for two things: an annual butterfly festival, and the fact that it sits on top of one of the largest natural gas fields on earth. There are dozens of drilling rigs scattered across the region. And according to the Financial Times, the single biggest owner and operator of those rigs is not an oil company.

It is a hedge fund.

Ken Griffin’s Citadel, the firm most people associate with bond trades and quants in Miami, is now one of the largest physical energy players in the United States. The Financial Times recently traced how it got there, and if you flick a light switch in California, there is a decent chance the electrons passed through something Citadel touched.

I have read that article more times than is healthy, because I keep getting asked a version of the same question.

What would the African renewables version of that look like?

Now, let me get something out of the way, because I can already hear it.

The last thing this continent needs is someone reading one FT article and deciding they are now the Ken Griffin of Kwara. We have enough of those. The man with the 1 GW pipeline that has been “at final stages” since the last AFCON does not need a hedge-fund cosplay phase on top of everything else.

So this is not a “copy Citadel” essay. Copying Citadel would actually be a fast way to lose money in Lagos.

This is a “study the architecture beneath Citadel” essay. Because underneath the mythology, Citadel is not really a story about trading. It is a story about how information advantage, patient capital, talent, and physical ownership compound into something that becomes very hard to dislodge. And that architecture, translated honestly and not copied blindly, is the most useful mental model I have found for what a billion-dollar African energy company could actually be.

Let me show you what I mean. Then let me show you the one place the whole analogy breaks, because that broken place is the most important part.

What Citadel actually did

Strip away the drama and Citadel’s energy empire rests on a handful of moves. None of them are “be good at trading.”

The first move was distress. When Enron filed for the biggest bankruptcy in American history in 2001, Griffin did not wait for the dust to settle. He chartered a jet, sent his people to Houston within days, and they interviewed what insiders describe as nearly everyone in Enron’s trading operation, at every level. Other firms set up a few polite meetings. Citadel reverse-engineered the entire business. They walked away with a research team and, more valuably, a working map of how US energy markets actually function. A few years later, when the energy fund Amaranth blew up on bad gas bets, Citadel teamed up with JPMorgan, bought the entire book, and made a fortune cleaning it up.

The pattern is the thing to notice. Citadel did not fight incumbents at their strongest. It waited for capital structures to break, then acquired the assets and the people at a discount, and built the capabilities the previous owners never realised.

The second move was data, long before data was fashionable. By the mid-2000s Citadel was bragging about a “meteorology room.” Over the following decade it hired atmospheric scientists, built high-performance computing, and turned weather, shipping, satellite, and production data into something it could actually trade on. The claim the firm makes today is that it is the most sophisticated user of this kind of alternative data in the business. Whether or not that is literally true, the posture is the point. They treated information as the edge and everything else as plumbing.

The third move, and this is the one that matters most for us, was the slow walk from financial to physical. For years Citadel was a financial actor: buying and selling contracts on gas and power. Then in 2014 it set up an energy marketing business that physically moves gas from the people who produce it to the people who need it. Last year that business moved volumes equal to roughly a tenth of all US natural gas consumption. It now owns gas fields and rigs. It bought a German renewable-power trading firm and a Japanese power company. It went from betting on the flow to owning the flow.

That sequence (financial exposure first, physical control later) is the spine of the Citadel story. Understand the market on paper, then go own the tanks and pipes and wires that the paper represents.

And here is the deeper logic, the part that turns a clever trade into a moat. Once you own the physical layer, every asset becomes a sensor. The gas field tells you about production. The power assets tell you about flow. That proprietary information makes the next decision sharper, which funds the next asset, which generates more information. Information becomes assets becomes better information. The loop is the empire.

So far, so transferable. Distress acquisition, an obsession with data, ownership of the physical layer, and the patience to let the loop compound. You could write that on a whiteboard in Nairobi and it would still be true.

Except for one thing.

The lesson that breaks in the heat

Citadel got rich in deep, liquid, electronified markets. The US gas and power markets it trades are some of the most transparent and tradable on the planet. You can see prices by the second. You can hedge almost anything. Griffin’s whole worldview, by his own description, is that everything can eventually be automated and electronified.

Now look at the African commercial and industrial power market.

It is the opposite of electronified. Origination happens over relationships and site visits and warm bottles of water. Tariffs are negotiated, often with one side knowing far more than the other. Payment behaviour is idiosyncratic. Currency risk hangs over every contract. Land, permitting, and FX intrude at every step. There is no second-by-second price for “reliable industrial power in Aba.” There is barely a price at all.

Which means the part of the Citadel playbook that made Citadel its billions, the trading, the market-making, the merchant layer, is exactly the part that does not yet exist here.

This is the single most important thing I can tell you, so let me say it plainly.

You cannot copy Citadel’s order of operations. Citadel went financial first, then physical. In Africa you have to go physical first, because the financial and trading markets that reward a pure trader are years away from being deep enough to pay you. The African Citadel is built backwards from the original.

And once you accept that, something interesting happens. The opacity stops being a bug.

In a transparent market, information is a commodity. Everyone subscribes to the same data feed and competes on cost of capital. In an opaque market, the firm that quietly builds the only map wins everything. It originates better, buys better, and prices risk that nobody else can see. The very thing that makes African power impossible to automate, the fog, is what makes a proprietary information layer so valuable. You are not trying to be Citadel in a liquid market. You are trying to be the firm that builds the missing market’s nervous system, and then owns the best assets inside it.

So what actually transfers? Distress acquisition transfers, and in Africa distress is not a once-a-decade event, it is the weather. The data obsession transfers, except you cannot buy the data, you have to manufacture it. The reliability premium transfers, and harder than anywhere: in a market defined by counterparty fragility, “we will still be standing in twenty years, and our power will still flow” is itself the product. And the patient-capital discipline transfers, with a twist I will come to.

What does not transfer is the dream of a quant trading desk arbitraging a liquid market. Not yet. Pretending otherwise is how you end up with a beautiful model and an empty bank account.

The real scarcity is not panels

Here is the part where I am supposed to overwhelm you with market-size statistics. I will keep it to the ones that actually matter, because the numbers are not the argument. The numbers just tell you the argument is large.

Start with the wall of diesel. Wood Mackenzie counts roughly 100 GW of generators across the continent, and calls its own number conservative. Nigeria alone runs something like 28 GW, Ghana around 10, Kenya around 8. Seventeen African countries have more backup generator capacity than actual grid capacity. The Central Bank of Nigeria has noted that fuel for power can eat up to 40% of a manufacturer’s operating costs. Across sub-Saharan Africa, most firms experience regular outages, and more than half of manufacturers own or share a generator. This is not a market that needs to be created. It exists, it is enormous, and it currently runs on the most expensive, dirtiest power available.

Then the arbitrage that pays for everything. All-in diesel power lands around forty cents a kilowatt-hour, and higher once you honestly price the downtime, the theft, and the maintenance. Solar paired with storage now delivers somewhere between ten and fourteen cents in the same markets, with paybacks for a diesel-dependent factory that can compress toward a couple of years. That is not a marginal efficiency. It is a collapse in the cost of one of an industrial firm’s three biggest line items, available today, with technology that works.

Now the part I find genuinely funny, in a dark way. Nobody agrees how much solar Africa installed last year. The Global Solar Council says it was a record, around 4.5 GW, up more than fifty percent. The Africa Solar Industry Association, using its own bottom-up project tracking, counts roughly 2.4 GW. And then AFSIA looks at how many Chinese solar panels actually landed on African ports, and concludes that the real installed base might be 63.9 GW against the 23.4 GW it can formally see. A gap of something like 40 GW.

Sit with that. The official numbers might be wrong by nearly a factor of three, and most of the missing capacity is behind-the-meter: factories, mines, and commercial parks quietly going solar where no statistician can find them.

A 40 GW blind spot in the data is a tragedy for analysts. For a builder, it is a 40 GW map that nobody else has drawn. The disagreement is the opportunity.

And underneath all of it sits the real constraint, the one that decides who wins. It is not solar resource. We have absurd amounts of sun. It is not technology. Anyone can import a container of modules. The constraint is capital, and specifically the price of it. The IEA estimates the cost of capital for African energy projects runs at least two to three times higher than in advanced economies. Meanwhile private clean-energy investment on the continent has tripled, from around 17 billion dollars in 2019 to almost 40 billion in 2024, which sounds great until you remember it is a fraction of what is needed, held back precisely because lenders price the risk so high.

So here is the whole thesis in one line. In African energy, the scarce resource is not panels or even capital. It is risk judgment. Whoever can price African energy risk better than the banks, the DFIs, the utilities, and the developers will win, because better risk judgment means lower losses, lower losses mean a lower cost of capital, and a lower cost of capital wins the projects. Everything else is choreography.

That is the citadel. Not the megawatts. The judgment.

What you actually build

So you do not build another solar EPC. You do not build another IPP that prays at the altar of announced megawatts. You build the thing that prices risk better than anyone else, and you wrap physical assets around it. In practice that is three layers, and the order they go in is the entire game.

First, you build the map. Before you deploy serious capital, you build the best proprietary dataset in African energy. Not market reports. Real operating data. You meter the factories and the cold rooms and the agro-processors and the telecom sites. You learn their load profiles, their diesel runtime, their power-quality events, their payment history, their currency exposure, their roof and transformer constraints. You watch which feeders actually fail, as opposed to which ones the utility claims are fine. You track which inverter models die early in the heat and the dust, which battery chemistries degrade, which EPCs deliver and which ones vanish after commissioning.

Most developers underwrite a project from a stack of bills and a single site visit and a prayer. What you are building instead is a live underwriting engine, an African energy risk graph that knows which sectors pay, which geographies are real, which contract structures survive an FX shock. That graph is the thing the whole company is organised around, and it is the one asset that cannot be bought. It can only be accumulated, deal by deal, site by site, over years. A competitor with a fatter balance sheet can clone your app in a quarter. They cannot clone four years of verified knowledge of who burns what and who pays.

Second, you own what the map points to. This is where many platform founders flinch, and they are wrong to. The FT quotes a rival who studied Citadel and walked away, and his conclusion is the most useful sentence in the whole piece: unless you are willing to actually own assets, it is hard to build a real edge. Brokering and software alone leave you renting other people’s economics and, crucially, other people’s data.

Owning physical generation does three things nothing else can. It captures the full diesel-to-solar arbitrage instead of a thin fee on it. It turns you into the reliable, long-term counterparty that commands a premium, which is precisely why Shell wrote a cheque for Daystar in 2022 and why CrossBoundary can raise a half-billion-dollar-plus portfolio against blue-chip offtakers like Unilever, Diageo, Heineken, and a copper mine in the DRC, wrapped in a 495-million-dollar World Bank guarantee against the currency risk. And, most importantly for our purposes, every asset you own becomes a permanent sensor feeding the map: real yield, real degradation, real payment behaviour over a fifteen-year contract. The developer with assets but no data layer is wasting the telemetry. The software company with a data layer but no assets has no telemetry to waste. Only the firm that owns both keeps the map alive.

You do not buy randomly, though. You own selectively, where ownership produces information and optionality: C&I solar and storage portfolios, hybrid systems in weak-grid corridors, battery fleets where reliability is scarce, eventually wheeled positions where the rules allow. And you treat distress as the accelerant, because the consolidation wave is already releasing supply below replacement cost: underperforming portfolios with good sites and bad operators, sub-scale installers who cannot raise the next round, developers caught on the wrong side of a currency move, and the occasional fund-level governance break that strands an entire book. The Amaranth move, adapted. When a competitor’s capital structure cracks, the assets and the people are available to whoever did the work and kept capital ready.

Third, and last, you cash the asymmetry. This is the merchant layer, the part that made Citadel its fortune, and I am going to be honest about it in a way that most pitch decks are not: in African power, it barely exists yet. So you rank it by what is real today.

Biomass feedstock trading is real now. It does not wait on any power-market reform. It is physical, relationship-gated, and information-intensive, which is exactly the terrain where a proprietary supply-and-logistics map plus contracted tonnage is a genuine edge. Certificates are real, and growing: I-RECs from African projects still trade for less than a dollar a megawatt-hour, which is small, but if you own the generation you originate them at zero marginal cost, and the impact variant, the Peace REC, has sold to the likes of Microsoft and Google at around forty-five dollars a certificate. When you already own the asset, that is pure margin sitting on top of power you were selling anyway.

And then there is the tell, the single clearest signal that the merchant layer is coming and that the data company is positioned to own it. Look at South Africa. Eskom launched virtual wheeling commercially in 2025, Vodacom became the first to use it at scale, and the wholesale electricity market is being stood up as we speak. But traders and aggregators, the people who would actually make that market liquid, are still locked out, waiting on the rules. And the reason everyone keeps naming for why it cannot scale yet is automation. With many generators and thousands of meters, somebody has to track allocations, match them to consumption, and reconcile settlements at a scale that is impossible by hand. Eskom had to hire a firm to build the software platform to do it.

Read that again with a builder’s eyes. The continent’s most advanced power market is being throttled by precisely the settlement-and-aggregation capability you would be building anyway. That is not a coincidence to admire from a distance. It is the merchant opening, and it arrives the moment the rules catch up to the code.

So the discipline is this. Underwrite the data and the assets on their own economics, because those pay today. Treat the merchant layer beyond biomass and certificates as cheap optionality on a market that is visibly, if slowly, electronifying. Do not put a single naira of required return on a power-trading market that has not been built. But be the one company positioned to walk through the door the instant it opens.

Why this is hard to copy

Each of those three layers is a business on its own. Composed, they become a moat, and the reason is the loop.

The map points you at the best assets. The assets generate proprietary operating and payment data. That data sharpens both the next round of origination and the merchant decisions. Better decisions mean lower losses and higher utilisation, which lowers your cost of capital, which wins you the next assets, which feed the map. Around it goes, and it widens with every turn.

Now watch why a single-layer competitor cannot catch up. The pure software company sees only the deals that cross its own platform, earns a thin fee, owns nothing, generates no operating data, and is one well-funded copycat away from irrelevance. It is also worth remembering that horizontal software in this space already exists and is genuinely good: there are platforms moving billions in financing for distributed energy across the continent. Which is exactly the point. Software alone is contestable. The pure developer owns assets and the telemetry they throw off, which is real, but it only ever sees its own deal flow. It selects from what walks in the door rather than from a map of the whole field, and it has no structural trading edge beyond its own generation. The pure trader has no proprietary supply or demand signal and no assets to generate one, so in a foggy market it is trading half-blind.

Only the integrated entity closes the loop. And it is defended by all the boring, durable things at once: network effects, because the benchmarking gets better with every site you add; switching costs, because offtakers are embedded in multi-year metered contracts; scale economies, because you finance portfolios instead of projects; and a proprietary-data advantage that a balance sheet cannot buy because it can only be lived.

The replicable layers are the ones founders love to talk about: the slick app, the clever lease product, the brand. The defensible layers are the quiet ones: the accumulated data, the calibrated asset telemetry, the earned trust of a guarantor like the World Bank. The strategy is not to build the best software. It is to use the software to accumulate the one asset capital cannot buy, and bolt it to a balance sheet a software company could never raise.

The ways this kills you

I have spent enough years in this market to know that the graveyard is full of beautiful models. So in the spirit of scar tissue, here are the ways this particular ambition gets you killed, and roughly how you survive each one.

It kills you through reputation and politics. The FT puts it perfectly: if half of California loses power and it traces back to you, that is bad. In Africa the charge is doubly live, because electricity is welfare and governments intervene, sometimes retroactively. The moment you are simultaneously the firm that sees the most, owns the assets, and profits from the spread, you attract scrutiny a humble installer never would. Eskom is already in court trying to slow the very traders the market was meant to unlock. You survive this by being additive rather than extractive, by never being the single point of failure for a critical load, and by treating regulators as people you help build investable markets, not people you outsmart.

It kills you through currency. This is the structural killer, the one that has buried more African energy projects than any inverter fault. Your revenue is in naira and cedi and CFA. Your capital and your panels are priced in dollars. A single currency move can add a fifth to your costs overnight, and several have. You survive this the way the serious players do: local-currency debt where it exists, guarantee wraps against inconvertibility, tariffs structured to breathe with the shocks, and a geographic spread across currency regimes so no single devaluation sinks the book.

It kills you through the wrong capital. The FT lands on this quietly: once you start owning assets, you are really a private equity firm wearing a hedge fund’s clothes, and you need locked-up capital. An asset-owning platform run on hot, redeemable money is a contradiction that ends in a fire sale. This is the twist on the patience lesson. Citadel returns capital to stay nimble. You need the opposite: committed, long-duration capital that can sit through an eighteen-month development cycle without flinching. The cleanest answer is to split the house. Let a capital-light data and software company carry the venture money and the high margins and the proprietary datasets. Let a separate, patient asset vehicle carry the infrastructure capital and the long-tenor debt and the guarantees. Wire them together with one nervous system: the data company feeds origination and risk intelligence to the asset vehicle and is paid for it, the asset vehicle feeds operating data back. Investors can back either risk appetite. The value accrues to whoever owns the layer running between them.

And it kills you through the smaller, dumber things, the ones that do not make it into thesis essays but do make it onto gravestones. Selling solar when the customer is buying reliability, and watching them churn the moment a cheaper quote appears. Chasing announced megawatts instead of collected kilowatt-hours, because announced megawatts are vanity and collected kilowatt-hours are sanity. Financing customers you do not actually understand, the ones who serve you cold water, nod gravely through the meeting, show you three generators, and then treat your invoice as a philosophical suggestion. Customising every single project until you are not a scalable business at all, just a very tired consultancy with some panels attached. Ignoring operations and maintenance until dust and heat and a dead inverter quietly destroy an IRR that looked beautiful in the model. And making carbon credits the base case, when carbon should improve a return, never be the thing standing between your project and financial sadness. If a project only works because of carbon, it is not a project yet. It is a prayer request with an emissions factor.

None of these are clever. All of them are fatal. The companies that scale are not the ones with the best slides. They are the ones that internalised the boring discipline early.

The endgame

Here is what I actually believe, underneath all the jokes.

The African renewable-energy unicorn will not be a solar company. It will be an energy intelligence company. A credit-underwriting company. An asset-management company. A reliability company. A financing company. A data company. And yes, somewhere in there, also a renewable-energy company.

The endgame was never panels on rooftops. It is control of the reliability layer of African industrialisation. The next phase of this continent’s growth needs power for factories and cold chains and data centres and ports and irrigation and the processing of the minerals the world is suddenly desperate for. The company that wins is the one that knows, before anyone else, which factories will grow, which feeders will fail, which customers will pay, which states will regulate well, which batteries will die early, which portfolios are mispriced, and which markets are about to open. That is not a solar installer. That is an energy institution.

And it gets there backwards from Citadel, on purpose. Build the map first, because in the fog the map is everything. Own the assets the map points to, because ownership captures the economics and keeps the map alive. Let the merchant layer pay out last, as the markets electronify, fastest in South Africa, where the bottleneck is already the exact capability you are building. Do that, and the thing that makes you hard to dislodge is not any single layer. It is that no competitor can assemble all three in the order the market actually rewards.

A hedge fund became the biggest rig operator in a Louisiana town famous for butterflies. It did not happen because they were the best traders. It happened because they understood the flow before anyone else, and then went and owned it.

The African version is sitting there waiting to be built. Not by the loudest installer. By the best risk-underwriter, data owner, operator, financier, and eventually market-maker in African power.

That is the template. The market is open. The window is not permanent.

Next in the series, we leave the hedge funds and go somewhere stranger.

A few honest disclosures, because this is the internet and someone always asks.

P.S. - None of this is investment advice. I model power projects for a living, and I have watched genuinely beautiful ones die, in spreadsheets and in the field. That has left me allergic to confident predictions, my own included. Read everything above as a way of thinking, not a tip sheet.

P.P.S. - The views here are mine alone. Not my employer’s, not any board’s, not the committee that has to sign off on my actual day job.

P.P.P.S. - And yes, I am aware that I have just spent four thousand words describing, in some detail, a company that looks a great deal like the one I have been building on nights and weekends for years. We will get to that another day. Or we will not, and you can simply wonder.

P.P.P.P.S. - If this was useful, the kind thing is to subscribe. The useful thing is to forward it to the one person you know who is about to raise money to install panels and call it a strategy. And if you think I have this wrong, I would honestly like to hear why. I have been wrong about enough things to find the exercise interesting.

P.P.P.P.P.S. - Template 2 is already taking shape. It is stranger than a hedge fund, which is a high bar.

Read this as research. Read it as strategy. Read it as vibes. Read it as me, once again, thinking too loudly on the internet.

Sources and further reading

The Citadel source

1. Robin Wigglesworth, “Citadel: the hedge fund that became an energy giant,” Financial Times. https://www.ft.com/content/3d0842dd-a8f4-435d-b888-5587b4b9eeda

The scale of the problem: market size and economics

2. “Africa’s hidden solar boom: how C&I platforms are quietly consolidating a 40 GW market,” African Exponent (the ~100 GW of African gensets per Wood Mackenzie; Nigeria ~28 GW, Ghana ~10 GW, Kenya ~8 GW; all-in diesel above $0.40/kWh versus solar-plus-storage at $0.10 to $0.14/kWh). https://www.africanexponent.com/africas-hidden-solar-boom-how-c-i-platforms-are-quietly-consolidating-a-40-gw-market/

3. Adewale A. Adesanya and Joshua M. Pearce, “Economic viability of captive off-grid solar photovoltaic and diesel hybrid energy systems for the Nigerian private sector,” ScienceDirect (diesel captive generation at roughly $0.28 to $0.33/kWh). https://www.sciencedirect.com/science/article/abs/pii/S1364032119305568

4. “Solar PV powering the C&I sector in Africa,” Synergy Consulting (financed captive solar at $0.08 to $0.12/kWh, 60 to 70% below backup diesel at $0.35 to $0.40/kWh). https://www.synergyconsultingifa.com/noteworthy/industry-knowledge/solar-pv-powering-commercial-industrial-sector-in-africa/

5. “Africa Diesel Genset Market,” Market Data Forecast (Central Bank of Nigeria estimate that businesses spend up to 40% of operating costs on fuel for power). https://www.marketdataforecast.com/market-reports/africa-diesel-genset-market

6. Nigeria electricity prices, GlobalPetrolPrices.com (business grid tariff around $0.048/kWh, September 2025). https://www.globalpetrolprices.com/Nigeria/electricity_prices/

7. “Tipping point for African C&I solar,” pv magazine (Nigeria’s C&I solar opportunity put at “at least $15bn per year”). https://www.pv-magazine.com/2022/11/01/tipping-point-for-african-ci-solar/

The data, and the disagreement about it

8. “GSC: Africa adds record 4.5GW of new solar PV capacity in 2025,” PV Tech (Global Solar Council figure, a 54% year-on-year increase). https://www.pv-tech.org/gsc-africa-adds-record-4-5gw-new-solar-pv-capacity-2025/

9. “AFSIA: 2.4GW of new solar deployment in Africa in 2025,” PV Tech (the African Solar Industry Association’s lower, bottom-up tracked figure). https://www.pv-tech.org/chinese-solar-imports-batteries-2-4gw-new-solar-deployment-africa-2025/

10. “Africa’s installed PV capacity estimated above 63 GW,” pv magazine (23.4 GW tracked versus 63.9 GW estimated once Chinese export data is reconciled). https://www.pv-magazine.com/2026/01/22/africas-installed-pv-capacity-estimated-above-63-gw/

11. “Solar PV in Africa: conflicting data confuse the messaging,” Empower New Energy (a direct read on the AFSIA versus GSC disagreement). https://www.empowernewenergy.com/post/solar-pv-in-africa-conflicting-data-confuse-the-messaging

The developers and the deals

12. “CrossBoundary Energy C&I Africa Portfolio,” MIGA / World Bank Group (the $495m guarantee framework against currency inconvertibility and transfer restriction). https://www.miga.org/project/crossboundary-energy-ci-africa-portfolio

13. “CrossBoundary Energy secures US$200M senior debt...,” CrossBoundary Energy, November 2025 (Standard Bank-led facility with Absa, MCB, FEI, DEG, FMO; blue-chip offtakers; the Kamoa Copper baseload deal). https://crossboundaryenergy.com/crossboundary-energy-secures-us200m-senior-debt-to-further-renewable-energy-portfolio-in-africa/

14. “Shell acquires African C&I solar provider Daystar Power,” PV Tech, December 2022 (a supermajor’s first power acquisition in Africa). https://www.pv-tech.org/shell-acquires-african-ci-solar-provider-daystar-power/

15. “Merger activity heats up in African commercial solar market,” pv magazine (the Starsight and SolarAfrica merger, backed by Helios and AIIM). https://www.pv-magazine.com/2022/09/28/merger-activity-heats-up-in-african-commercial-solar-market/

The merchant layer: wheeling, certificates, biomass

16. “Eskom moves to automate virtual wheeling,” Energize (the Enerweb platform; automation as “a prerequisite, not a nice-to-have”). https://www.energize.co.za/article/eskom-moves-to-automate-virtual-wheeling

17. “Virtual Wheeling,” Eskom Distribution (the mechanics of the product). https://www.eskom.co.za/distribution/virtual-wheeling/

18. “South Africa: why Eskom’s virtual wheeling is a breakthrough moment for SMEs,” Zawya (Vodacom as first at scale via a Sola Group PPA, September 2025; the wholesale market arriving in 2026). https://www.zawya.com/en/economy/africa/south-africa-why-eskoms-virtual-wheeling-is-a-breakthrough-moment-for-smes-x1ejwt8q

19. “The trader’s legal landscape,” Cliffe Dekker Hofmeyr (traders and aggregators excluded from virtual wheeling pending NERSA’s trading rules, anticipated around mid-2026). https://www.cliffedekkerhofmeyr.com/en/news/publications/2025/Sectors/Projects-Energy/projects-and-energy-alert-20-august-The-traders-legal-landscape

20. “A guide to renewable energy certificates and global attribute markets,” Innovo Markets (I-RECs, and the Peace REC variant for fragile and conflict-affected regions). https://www.innovomarkets.com/blog/renewable-energy-certificates-and-global-attribute-markets

21. “The market outlook for International Renewable Energy Certificates,” Ecohz (I-REC issuance, redemption, and cross-border trade). https://www.ecohz.com/news/irec-market-outlook

22. Energy Peace Partners, Peace REC FAQs (average prior transactions around $45 per Peace REC; corporate buyers including Microsoft and Google).

23. Odyssey Energy Solutions / Transforming Energy Access (horizontal distributed-energy platform; billions in financing enabled across thousands of company users).

Investment and capital

24. “Africa, World Energy Investment 2025,” IEA (private clean-energy investment tripling from around $17bn in 2019 to almost $40bn in 2024; roughly 600 million Africans still without electricity access; cost of capital for African energy projects at least two to three times higher than in advanced economies). https://www.iea.org/reports/world-energy-investment-2025/africa

It is also, if you read it carefully, a $176 million institutional admission that the existing financing architecture is broken.

Not because the money is unimportant. The money matters enormously.

But because the shape of the money tells you what the market has been missing: patient capital, platform capital, and structures built for distributed assets that do not fit neatly inside traditional project finance.

I need to correct something.

In Part 1 and Part 2 of this series, I made the argument that African renewable energy companies keep dying because the market structure is wrong, and that history, from semiconductors to oil and gas to telecoms, tells us exactly what the structural fix looks like. The argument was right. The scope was wrong.

I analysed one-third of the market and presented it as the whole picture. The entire analysis was about C&I solar: developers chasing 500 kW to 20 MW projects for industrial offtakers, navigating project finance structures, dying in the gap between origination and financial close.

But the market is wider than that.

While I was writing about the missing middle in C&I finance, Nigerian households were adding 803 MW of solar capacity in a single year. A 141 percent year-on-year increase. Africa’s second-largest solar market behind South Africa. Almost entirely self-financed. Almost entirely invisible to the institutional energy conversation.

And the mini-grid sector, the most DFI-discussed segment in all of African energy, had already quietly built the prototype of the structural solution I said was missing.

This piece is about the other two-thirds. Then it is about what would actually fix all of it. Then it is the survival playbook for developers who need to know what Monday morning looks like.

The Residential Boom

Let me start with this incredible number.

Nigeria’s cumulative solar capacity rose from roughly 385 MW in 2024 to nearly 1.19 GW in 2025. Of that, approximately 96 percent is off-grid: solar home systems, private rooftop installations, commercial systems, mini-grids. The utility-scale share is a rounding error.

And here is the data point that should be projected onto the wall of every DFI conference room in Washington and London: formal capital importation into Nigeria’s electricity sector has been declining even as solar panel imports climb significantly. The boom is being financed by Nigerian households and small businesses making individual economic decisions. Not by institutional investors. Not by blended finance facilities.

The market cleared without us.

Two triggers arrived simultaneously in May 2023. Fuel subsidy removal caused petrol prices to rise nearly 500 percent. The national grid collapsed at least ten times in 2024 alone. For Nigerians, a solar panel now pays for itself within six months against diesel costs. Six months is not an investment thesis. It is an emergency decision.

The 6-month payback period is doing more for Nigerian residential solar adoption than twenty years of DFI technical assistance ever did for grid-connected IPPs.

Think about what this means through the Carlota Perez framework from Part 2. The residential solar market has entered the deployment phase without passing through an institutional installation phase. There was no government-sponsored foundry. No IREDA. No SECI auction. No MASEN procurement intermediary. Individual Nigerian households did the installation phase with their own capital, one rooftop at a time, driven by the simple economics of diesel avoidance. This is simultaneously the most impressive and the most fragile form of energy transition imaginable.

Three Markets Pretending to Be One

The residential boom is not one market. Collapsing it into a single category obscures the dynamics that matter.

Solar home systems and PAYG. Sun King has sold over 2 million kits in Nigeria, growing from 3,000 kits per month in 2020 to 75,000 per month today. Nigeria is Sun King’s fastest-growing market and shows its lowest rates of non-payment globally. The company operates through more than 45,000 agents across Africa and had extended $1.2 billion in total loans to customers as of May 2025. In May 2025, Sun King closed an $80 million naira-denominated facility with IFC and Stanbic IBTC Bank, explicitly designed to mitigate foreign exchange risk. The World Bank’s NEP programme has delivered over 1.08 million verified standalone solar connections.

These are small systems, 50 to 200 watts, serving households with no or minimal grid connection. But at 75,000 kits per month, Sun King alone is connecting nearly a million Nigerian households per year. That is more connections per year than many African utilities manage.

Urban rooftop solar. Arnergy scaled its lease customer base rapidly between 2023 and 2024, deploying over 1,800 systems across 35 Nigerian states, totalling 9 MWp of solar and 23 MWh of battery storage. These are larger systems, 3 to 20 kWp with meaningful battery storage, serving grid-connected households experiencing chronic reliability failure. Battery storage is not optional in this segment. Without it, you have daytime-only power.

Estate-embedded solar. This is the sub-market I missed entirely, and it is the most structurally interesting. Property developers are now installing large-scale solar as a primary selling point. Solar-powered estates now house approximately 200,000 people across Nigeria, up from fewer than 50,000 in 2020. These developments embed solar at the estate level, typically 200 kWp to 3 MW, with costs recovered through service charges.

Read that again. Property developers have quietly become de facto electricity utilities. They own generation assets. They distribute power within a private boundary. They bill residents through service charges. They are doing everything a DisCo does, but better, within a gated perimeter, without a licence, and funded through real estate economics rather than energy finance.

This is the Samsung-as-IDM problem from Part 2, but in reverse. The estate developer vertically integrates because the specialist electricity market literally does not serve them. Williamson’s transaction cost economics again: when the external market is too unreliable, firms internalise the function. Every estate developer who installs solar is making the same calculation that every African RE developer makes when they try to be an IDM. The difference is the estate developer has a captive offtaker and a real estate margin to absorb the cost.

The Quality Crisis Underneath the Boom

The Nigerian Energy Support Programme reports that up to 40 percent of solar components in the market fail to meet international standards. REAN estimates that nearly 50 percent of installed systems suffer from design flaws: incorrect panel orientation, undersized inverters, insufficient battery capacity. There are over 350 solar installers listed on ENF in Nigeria alone. The vast majority are tiny, uncertified, with no formal training.

A homeowner who spent ₦500,000 on a solar system that stopped working after six months tells twenty people. Those twenty people do not buy solar. The substandard installer has already moved to the next customer. The reputational damage is absorbed by the entire industry.

This is the residential equivalent of the bankability problem in C&I. In C&I, financiers do not trust the project data. In residential, consumers do not trust the product and the installer. Both are information and trust failures. Both kill the market’s ability to scale. But they manifest differently: in C&I, the trust failure slows down deals. In residential, the trust failure actively destroys demand.

The Cleantech 1.0 parallel from Part 2 is instructive here. The MIT study found that 90% of cleantech companies funded after 2007 failed because they were hardware companies trying to compete in a commoditised market. The Nigerian residential solar market has the same commodity problem. Panels are panels. The differentiation is in installation quality, system design, and after-sales service. But there is no market infrastructure to signal quality. No installer certification that consumers trust. No system passport that a secondary buyer could evaluate. No warranty enforcement mechanism. The quality crisis is the residential version of the missing foundry.

The Mini-Grid Frontier: The Foundry That Already Exists

The World Bank estimates Sub-Saharan Africa needs 160,000 new mini-grids to achieve universal access by 2030. As of 2024, AMDA members had deployed roughly 600. That is 0.4 percent of the stated target with fewer than five years remaining.

That is not a miss. It is a structural failure that has been building for a decade.

CAPEX for mini-grids in Africa has declined about 20 percent since 2020, to a four-year average of $6,824 per kWp. That is more than double the global benchmark of $3,000 per kWp. Average consumption per customer is only 6.1 kWh per month across the continent. At $0.30/kWh, a household is paying roughly $1.80 per month. The only way the maths works is productive use: commercial anchor loads like grain mills, cold storage, water pumping, and small manufacturing that generate 10 to 50 times the revenue of a pure residential connection.

But here is what changed my understanding of the whole thesis.

The mini-grid sector has already built the structural solution I said was missing in C&I. It just built it quietly, without anyone using the semiconductor analogy to describe it.

CrossBoundary Energy Access is Africa’s first project finance facility dedicated to mini-grids. Their model: CBEA finances construction and owns the assets. The developer focuses on origination, community engagement, and operations. The developer retains a minority equity stake and the O&M contract. In January 2026, CBEA completed the first acquisition of an operational mini-grid platform, ANKA in Madagascar, with the developer retained as minority shareholder and operator.

Map it to the semiconductor analogy. CBEA is the foundry. It provides the capital-intensive manufacturing capacity, asset ownership and balance sheet, that individual developers cannot afford. The developer is the fabless designer. They bring the knowledge-intensive, relationship-intensive, local capabilities. CBEA owns the assets. The developer retains the IP equivalent: local knowledge and operational expertise, and a carried interest in the asset they originated.

This is the TSMC model for distributed energy. Beyond conceptual. Beyond just a proposal. Operating.

Husk Power operationalised the other half: the productive use anchor load thesis. Their deal with Olam Agri, a 1.3 MW solar system with 860 kWh BESS under a 10-year PPA for a rice farm in Nasarawa State, is a C&I solar project sitting inside a mini-grid portfolio. The anchor industrial load subsidises the community residential tariff while generating higher-margin C&I revenue.

And the regulatory infrastructure is catching up. Nigeria now has more than 1,000 deployed mini-grids, with more than half financed through the REA. The December 2025 commercial framework for interconnected mini-grids establishes, for the first time, a clear commercial relationship between mini-grid operators and distribution companies. NERC’s 2026 regulations expand allowable interconnected mini-grid capacity to 10 MW. The $750 million DARES programme targets 1,350 solar mini-grids, including 250 interconnected systems. A developer who has been hesitating on interconnected sites because the commercial terms with the local DisCo were undefined now has a regulatory basis for those negotiations. That window is open now.

Three Markets, One Root Cause

Step back and look at all three segments together.

C&I’s problem: financiers do not trust the project data, so deals do not close.

Residential’s problem: consumers do not trust the product and installer, so bad installations proliferate and demand is poisoned.

Mini-grids’ problem: developers cannot predict demand accurately, investors cannot assess risk without historical comparables, so capital stays on the sidelines.

All three are data and trust failures. Different manifestations. Same root cause. The absence of standardised, verified, Africa-specific energy data infrastructure.

Khanna and Palepu’s institutional voids framework from Part 2 explains this precisely. In emerging markets, the absence of credit bureaus, contract enforcement, quality certification, and professional intermediaries forces every firm to fill those voids individually. In the Nigerian energy market, every C&I developer builds their own financial model from scratch. Every residential installer self-certifies. Every mini-grid developer conducts their own demand assessment. Each is independently solving the same information problem at enormous individual cost.

A C&I project needs a verified asset record. A residential installation needs a system passport. A mini-grid needs a community energy profile. All three data products share common underlying infrastructure: grid reliability data, solar irradiance data, equipment performance benchmarks, regulatory frameworks, fuel pricing.

And here is the network effect that makes this genuinely interesting. A platform serving only C&I has maybe 500 data points. One that serves C&I plus residential has 50,000. One that serves all three segments has 100,000 data points spanning urban industrial zones, suburban residential areas, and rural communities. The intelligence gets more accurate. The benchmarks get more meaningful. The carbon portfolio gets commercially viable.

The residential market is where the data volume lives. Mini-grids are where the data diversity lives. C&I is where the data value per record is highest. Together, they create a dataset that no single segment could produce alone.

The data infrastructure IS the bankability infrastructure. They are the same thing.

What Would Actually Fix This

Enough structural analysis. Not more accelerators. Not more pitch competitions. Not more capacity building workshops. Actual structural interventions that change the economics of being a small developer.

1. A development finance facility at the right ticket size. The single biggest killer of small developers is the development phase: $20,000 to $50,000 per project for feasibility, site assessment, legal documentation, and financial modelling. If the project does not close, that money is gone. What would work: a revolving development finance facility, capitalised by DFIs or philanthropic capital, providing $20,000 to $100,000 in recoverable grants. If the project reaches financial close, the facility gets its money back. If it fails, the loss is absorbed. For mini-grids, the window needs to be larger, $50,000 to $150,000. For residential installers, the equivalent is a revolving inventory credit line secured against PAYGO receivables. ETAFA Nigeria’s $50 million local currency fund, where $10 million in GEAPP concessional capital catalyses $40 million in commercial finance, demonstrates the leverage ratio works. The missing piece is ticket size: ETAFA is still structured for medium-scale projects, not the $20,000-$50,000 development phase.
   
   Zafiri is important here because it shows the capital stack is beginning to move in the right direction: patient, blended, permanent capital aimed at distributed renewable energy rather than only large grid-connected infrastructure. But vehicles like Zafiri will still face the same market-structure question. How do they find, compare, diligence, price, and monitor hundreds or thousands of small distributed assets without recreating the same bespoke due-diligence burden deal by deal?
   
   That is where the foundry thesis comes back. Patient capital helps. But patient capital still needs trusted project data, standardised documentation, verified demand, equipment quality signals, and operating benchmarks. Otherwise, it simply becomes patient capital waiting patiently for bankable projects to appear.

2. Standardised documentation that nobody has built. Every small developer is reinventing the wheel on every deal. A lawyer for the PPA: $5,000 to $10,000. A feasibility study: $10,000 to $20,000. A financial model from scratch. Each a fixed cost that a platform company has already amortised across fifty deals. What would work: open-source standardised PPA templates in naira and USD, bankable feasibility study templates, financial model templates with pre-built assumptions. For residential: standardised system sizing methodology, installation checklists, commissioning test procedures. For mini-grids: five to ten reference designs for common community sizes. If you can collapse the development cost of a 500 kWp C&I project from $50,000 to $15,000, you have tripled the number of projects a small developer can pursue with the same balance sheet. This sounds boring. It is boring. It is also the single most powerful cost reduction available. Germany’s Sparkassen did not fund the Energiewende because they were innovative lenders. They funded it because the 20-year FIT made the lending decision boring. The African energy sector needs the same boring standardisation.

3. An equipment purchasing consortium. A small developer buying 500 kW of panels pays 15 to 25 percent more per watt than Starsight or BECS buying 50 MW. That margin difference alone can make a project uncompetitive. A purchasing cooperative where 20 to 30 small developers aggregate their annual equipment needs. Agricultural cooperatives have done this for a century. The REA model of aggregating demand across 1,350 mini-grids under DARES is the same logic at the programme level. The solar industry just has not built it at the developer level yet.

4. State-level registration with fast-track permitting. The Electricity Act 2023 is the single biggest structural opportunity for small developers. Fifteen Nigerian states have transitioned to independent electricity regulation as of early 2026. LASERC inaugurated its board in March 2026. A state-level registration system where developers meeting minimum criteria get pre-approved, with fast-track permitting, 30 days instead of six months, and access to a state-maintained database of potential sites and offtakers. This is directly analogous to how the marginal field programme worked: reduced entry barriers, simplified terms, starter opportunities. For residential, this translates to installer licensing that creates a quality floor protecting consumers and rewarding good installers. Nigeria’s solar manufacturing capacity has already grown from 120 MW to roughly 300 MW, with 3.7 GW in the pipeline and locally produced panels now being exported from Lagos to Accra. The state-level market infrastructure needs to keep pace.

5. A deal exchange connecting originators to capital. Small developers have deals but no capital. Platform companies have capital but struggle with origination. What would work: a platform where a developer uploads a standardised project package and qualified financiers or platform companies review and bid. The developer retains a carried interest or development fee. This is the direct translation of the oil and gas farm-in/farm-out structure from Part 2. A developer who secures an offtake agreement and permit for a 2 MWp project has created $50,000 to $100,000 of development value. Instead of trying to raise the full $2 million, they farm out to a platform company. The developer retains 10 to 20 percent and the O&M contract. For mini-grids, this is exactly what CBEA is already building.

6. Market infrastructure for energy trading. When diesel has a market price and solar electricity does not, banks will always find it easier to lend against diesel. The NBET-to-NENEX transition, modelled on India’s IEX which has over 8,100 participants including more than 2,100 renewable generators, creates the possibility of transparent price signals for generated electricity. Once electricity has a visible spot price in naira, lending against solar output becomes dramatically simpler for Nigerian banks. The analogy to oil is direct: the reason oil companies can borrow against reserves is that there is a transparent, liquid market for their output. Nigeria’s equivalent does not yet exist.

7. Data infrastructure that makes the whole market bankable. This underpins everything else. A standardised data schema for African energy assets, a common format defining exactly what information is needed to describe a project at every stage of its lifecycle, with provenance tracking for every data point. A residential system passport with verified component quality and installer credentials. A mini-grid community energy profile with verified demand data. If a critical mass of projects were described in a common format, the entire financing ecosystem would become more efficient. Lenders could compare projects on a standardised basis. Insurers could build actuarial models. Carbon auditors could verify emission reductions against structured records. Secondary market buyers could evaluate assets with confidence. The M-KOPA model from Part 2 proved this at consumer scale: the embedded GSM chip and daily M-Pesa payments created the information infrastructure that made un-bankable transactions bankable. The C&I and mini-grid sectors need the same thing at project scale.

The Survival Playbook for Developers Who Are Still Standing

While the ecosystem interventions above get built, and they will take years, small developers need to survive today. Here is the honest playbook.

Rule One: No offtake commitment, no spend. Ever. Before spending money on an energy yield assessment, a grid connection study, legal structuring, an ESIA, or a full feasibility report, you need a signed letter of intent from a creditworthy offtaker. Not a warm conversation. Not an expression of interest. A signed document from someone with a balance sheet. The LOI is not just a commercial milestone. It is a financing instrument. Get it first. Never before it.

Rule Two: Structure as a service company that also develops. Pure development companies have no recurring revenue between financial closes. Every month of pre-development is a month of cash burn with nothing coming in. A developer structured as an energy services company, with O&M contracts, energy audits, feasibility consulting, technical advisory work, generates monthly income that funds the development activity. The O&M contracts build a track record. The feasibility studies build client relationships. The service revenue keeps the company alive during the long stretches between closings. This is the Cleantech 1.0 survivor lesson from Part 2: Opower and First Solar survived because they had services layers and recurring revenue, not because they had better technology.

Rule Three: Use the consortium SPV model. Two or three small developers who each have one anchor client relationship but neither has the balance sheet to reach financial close should form a consortium SPV. Pool your pipelines. Approach a European, Asian, or Middle Eastern energy company that wants African exposure but cannot originate locally. The foreign partner provides 60 to 70 percent of the equity. The local developers retain 30 to 40 percent of a transaction much larger than either could have done alone. This is the Seplat model precisely transposed. Platform Petroleum spent years building a track record on a marginal field before the Shell divestment opportunity came. The consortium SPV is the marginal field for small solar developers.

Rule Four: For mini-grids, do not develop without an anchor load. An agro-processor, a cold chain facility, a fuel station, a telecoms tower cluster. Identify the anchor first. Structure the PPA with the anchor load first. Size the system for the anchor. Treat the community residential connections as both a social benefit and a revenue diversification. This is the Husk/Olam Agri model at smaller scale.

Rule Five: Get a Lagos Electricity Market licence now. LASERC inaugurated its board in March 2026. Lagos is the obvious example because LASERC inaugurated its board in March 2026 and is now a live regulatory body. But the opportunity is no longer Lagos-only. Under the Electricity Act 2023, several Nigerian states have transitioned to regulating their own intrastate electricity markets, with state electricity regulators taking over from NERC for local market activity.

Under the Electricity Act 2023, several Nigerian states have transitioned to regulating their own intrastate electricity markets, with state electricity regulators taking over from NERC for local market activity.

So the instruction is not simply “get a Lagos licence.” It is: map your pipeline against the new state electricity markets, understand which regulator now controls intrastate activity in each state, and get properly registered or licensed where required.

The cost of being early is usually small compared to the strategic value of being a recognised participant while market rules are still being written. Developers who are visible during market formation have standing to push for practical licensing categories, simplified permitting, standardised documentation, reserved procurement windows, and fair treatment for small distributed-energy players.

Those who wait until the rules are locked may find themselves subject to a market structure they had no hand in shaping.

Rule Six: Target the prosumer gap. NERC’s final Net Billing Regulations 2026 apply to systems between 50 kWp and 1.5 MWp. Earlier drafts contemplated up to 5 MWp, but the final framework is narrower. Hospital campuses. University clusters. Gated estates. Large religious institutions. This segment requires enough engineering sophistication to differentiate from residential installers but operates at a scale that Daystar and Konexa cannot cost-efficiently serve. It is precisely the terrain where knowledge beats capital. The Samsung-vs-TSMC lesson: the foundry model requires institutional separation. In the prosumer gap, small developers ARE the foundry. They bring the local knowledge, the relationship, the system design expertise. They just need a capital partner for the asset ownership.

Rule Seven: Document everything obsessively. Your feasibility studies, site assessments, client interactions, regulatory filings. This documentation IS your asset. It is what you bring to a farm-out negotiation, a partnership discussion, or eventually your own fundraise. In the semiconductor analogy, this is your design IP. You may not own the fab. But you own the design.

Rule Eight: Do not die. Keep fixed costs brutally low. Do not hire ahead of revenue. Do not rent an office until you have cash flow from at least one operating project. The developers who survive the next three to five years while the market matures will be the ones who were still standing when the ecosystem catches up. The market is coming. The question is whether you have enough runway to be there when it arrives.

The Bottom Line

The foundry problem has three dimensions, not one.

C&I needs a project development foundry that reduces development costs, standardises documentation, and connects developers to capital. Residential needs a quality and distribution foundry that certifies installers, ensures equipment quality, and provides consumer financing infrastructure. Mini-grids need an integrated delivery foundry that standardises designs, aggregates portfolios, and manages the complex interplay of subsidy, tariff, and grid arrival risk.

All three need shared data infrastructure. All three benefit from the same network effects. All three are currently being served by different companies, different programmes, and different investors who mostly do not talk to each other.

The semiconductor industry did not scale because chip designers got access to more capital on better terms. It scaled because the foundry model separated design from fabrication and routed the right capital to the right function. African renewable energy will not scale because small developers get marginally better access to project finance. It will scale when the architecture changes.

The market structure that will make African renewable energy work at scale is being built right now. Some of it in the Lagos Electricity Market rules. Some of it in the bilateral trading framework. Some of it in the December 2025 interconnected mini-grid commercial framework that almost nobody outside the sector is talking about. Some of it in conversations between developers who have not yet decided to collaborate rather than compete on terrain where neither of them can win alone.

The question is not whether this market will eventually work. It will. The question is who shows up to shape it while it is still being designed, and who arrives after the design is finished and wonders why the structure does not seem to have room for them.

I have been in this market long enough to know which side of that question I want to be on.

This is not a technology problem. It is an architecture problem.

The views expressed here are the author’s own and do not represent the views of his employer.

P.S. On the mini-grid section: I was initially sceptical of the CBEA model because I could not see why a mini-grid developer would voluntarily give up majority ownership. Then I talked to three developers who had spent two years trying to reach financial close on projects they had originated. Two of them had burned through their savings. One had taken a consulting job to pay rent. The question is not “why would you give up equity?” The question is “what is 30% of something versus 100% of nothing?” The Seplat founders did not insist on owning Shell’s entire onshore portfolio. They took what the structure offered and built from there.

P.P.S. The seven interventions section took me longer to write than Parts 1 and 2 combined. Structural analysis is easy. Specific, actionable interventions that do not read like a consultant’s slide deck are genuinely hard. If any of them sound too neat, I assure you the spreadsheet behind each one was messier than the prose. If they sound too obvious, ask yourself why none of them exist yet. Obvious and unbuilt is a more damning diagnosis than complicated and aspirational.

P.P.P.S. Rule Eight of the survival playbook, “do not die,” was the original working title for this entire series. My editor, pointed out that “The Foundry Problem” would attract readers who were interested in structural analysis, whereas “Do Not Die” would attract readers who were already in trouble. She was right. She is always right. This is why she edits and I write. If you are reading this and you are in trouble, go back and read Rule Two. Then call me.

P.P.P.P.S. Yes, I am aware that writing a blog series with four P.S. sections per instalment is arguably a structural problem of its own. Consider it a design challenge. Or consider it proof that even the author of “The Foundry Problem” has not fully solved his own architecture. A full reference list with all verified source links is available below. If you have made it through all three parts and all fifteen P.S. sections, you have spent more time reading about African energy market structure than most DFI programme officers spend before approving a $50 million facility. I am not sure what that says about you, but I am grateful.

Complete Reference List

1. Residential Boom

1. BusinessDay / Global Solar Council (February 2026). Nigeria Ranks as Africa’s Second-Largest Solar Market in 2025. businessday.ng

2. Global Solar Council. Africa Market Outlook Solar PV 2026-2029. LinkedIn / BusinessDay

3. Nairametrics / AFSIA (January 2025). Nigeria’s Installed Solar Capacity Rises to 385.7 MWp in 2024. nairametrics.com

4. The Electricity Hub / AFSIA (January 2025). Nigeria Ranked 4th in Africa for Solar Energy Adoption in 2024. theelectricityhub.com

5. Punch Nigeria / NBS. Nigeria’s Petrol Price Hits Record Highs. punchng.com

6. Vanguard Nigeria (December 2024). Blackout as National Grid Collapses 11th Time in 2024. vanguardngr.com

7. NERC Q4 2024 Quarterly Report. nerc.gov.ng (PDF)

8. IFC (2019). The Dirty Footprint of the Broken Grid. ifc.org (PDF)

2. Solar Home Systems / PAYG

9. IFC Press Release (May 2025). Sun King, IFC, and Stanbic IBTC Bank Close $80 Million Debt Facility. ifc.org

10. The Electricity Hub / REA (October 2025). Sun King, REA Unite to Power Nigeria’s Solar Future. theelectricityhub.com

11. This Day Live (October 2025). Sun King Plans $150M Investment in Local Production in 5 Years. thisdaylive.com

12. HSFG Africa / Sun King (November 2025). Sun King Announces Target of 200M People by 2030. hsfg.africa

13. Nithio Press Release (June 2025). Nithio Provides Due Diligence for Sun King’s $80M Facility. nithio.com

14. Sun King (May 2025). $80M Naira-Denominated Facility Press Release. sunking.com

15. KPMG Nigeria (March 2026). Nigeria’s Emerging Net Metering Framework. assets.kpmg.com (PDF)

3. Urban Rooftop / Arnergy

16. TechCrunch (April 2025). Arnergy to Expand Solar Access in Nigeria with $18M Series B. techcrunch.com

17. CleanTechnica (April 2025). Arnergy Raises $18M to Expand Solar Energy Access Across Nigeria. cleantechnica.com

4. Estate-Embedded Solar

18. BusinessDay Nigeria (February 2026). Nigerian Estates Turn to Solar to Lure Homebuyers Tired of Darkness. businessday.ng

5. Quality Crisis

19. DD Building Tech / NESP (2024). Why Solar Energy Systems Fail in Nigeria: Challenges and Solutions. ddbuildingtech.com

20. ENF Solar. Nigeria Installer Directory. enfsolar.com

6. Mini-Grids

21. World Bank ESMAP / 17 Actions for Mini-Grids. Mini Grids to Catalyze Electrification in Sub-Saharan Africa. sun-connect.org (PDF)

22. AMDA (2024). Benchmarking Africa’s Minigrids Report 2024. africamda.org

23. AMDA BAM 2024 / LinkedIn analysis. Mini-Grid CAPEX and Financing Trends. LinkedIn

24. AMDA BAM 2024 / LinkedIn. Average Mini-Grid Consumption Data. LinkedIn

25. ACAP (2026). Tariff and Regulatory Frameworks Masterclass for Mini-Grids. africanclimateactionpartnership.org (PDF)

26. World Bank (2021). Mini Grids for Half a Billion People. sdgs.un.org (PDF)

27. Allafrica / REA (March 2026). Nigeria 1,000+ Deployed Mini-Grids. allafrica.com

28. CrossBoundary Access (January 2026). CBEA Acquires ANKA Madagascar. crossboundary.com

29. Nextbillion.net (January 2026). CrossBoundary Energy Access Mini-Grid Madagascar. nextbillion.net

30. Husk Power Systems (April 2025). Olam Agri Partners with Husk Power for Sustainable Energy in Nigeria. huskpowersystems.com

7. Regulation & Policy

31. Guardian Nigeria / NERC (December 2025). NERC Issues Framework to Unlock Mini-Grids. guardian.ng

32. Allafrica / NERC (April 2026). NERC Mini-Grid Regulations 2026: 10 MW Cap. allafrica.com

33. DARES Programme / REA. Scaling Up Renewable Energy Access in Nigeria. energytransition.gov.ng

34. Seetaoe.com (March 2026). DARES 250 Interconnected Mini-Grids Analysis. seetaoe.com

35. World Bank (October 2023). DARES Project Information Document. worldbank.org (PDF)

36. Bizwatchnigeria (April 2026). 15 Nigerian States Electricity Regulation Transition 2026. bizwatchnigeria.ng

37. Vanguard Nigeria (March 2026). Sanwo-Olu Signs Lagos Electricity Law, Inaugurates LASERC. vanguardngr.com

38. PMNews Nigeria (March 2026). Sanwo-Olu Unveils Electricity Commission Board. pmnewsnigeria.com

39. NERC (September 2025). Draft Net Billing Regulations. nerc.gov.ng (PDF)

40. Nigeria. Electricity Act 2023 (Official Text). rea.gov.ng (PDF)

8. Manufacturing

41. Nairametrics / REA (April 2026). Solar Capacity in Nigeria Rises to 300MW. nairametrics.com

9. Structural Interventions

42. The Africa Report / Chapel Hill Denham / GEAPP. $50 Million Renewable Energy Fund in Nigeria. theafricareport.com

43. IEX India. Annual Report / Market Data. iexindia.com

10. Academic Frameworks

44. Perez, Carlota (2002). Technological Revolutions and Financial Capital. cambridge.org

45. Mazzucato, Mariana (2013). The Entrepreneurial State. marianamazzucato.com

46. Williamson, Oliver E. (2009). Nobel Prize: Transaction Cost Economics. nobelprize.org

47. Khanna, Tarun & Palepu, Krishna (2005). Strategies That Fit Emerging Markets. Harvard Business Review. hbr.org

Zafiri / Inspired Evolution

• Rockefeller Foundation announcement (16 June 2026): https://www.rockefellerfoundation.org/news/zafiri-announces-usd-176m-commercial-launch-accelerate-energy-access-private-sector-s

• IFC/LinkedIn announcement (April 2026): https://www.linkedin.com/posts/ifc-infrastructure_mission300-poweringafrica-activity-7448384348959158272-P4IR

• DevDiscourse coverage: https://www.devdiscourse.com/article/international/3936623-the-last-mile-of-african-power-just-got-a-176-million-backer

• Inspired Evolution appointed as Zafiri investment manager (IFC/AfDB, October 2025): https://www.linkedin.com/posts/engineering-africa_tackling-africas-off-grid-gap-the-international-activity-7388510149772197888-8

In summary

• The Fable 5 shutdown matters because it turned a theoretical risk into a live market fact. The US government showed that a deployed frontier AI model can be switched off through export-control authority, even after launch and even when paying customers are already using it.

• The government’s stated concern was a jailbreak linked to cyber capability, but the bigger issue is the precedent: a contested safety finding was enough to disable Anthropic’s most capable model worldwide.

• The order targeted foreign nationals, but Anthropic could not realistically separate US and non-US users at inference speed across a global cloud product. So a foreign-national restriction became a global shutdown. Very elegant, if your idea of elegance is setting the house on fire because one room has ants.

• Frontier labs are valued on the assumption of global scale: train once, serve everywhere, amortise the enormous compute bill across the planet. If access can be carved up by nationality, jurisdiction or alliance bloc, that scale story becomes weaker.

• This does not mean AI is dead. AI is still likely to be a multi-trillion-dollar industry. The more interesting point is that the value may shift away from pure-play model labs and toward the companies that own compute, cloud infrastructure, chips, data centres, power contracts and government distribution. In other words, the people selling the shovels may do better than the people posing beside the gold mine.

• Sovereign AI just got its best marketing campaign. Every government worried about depending on US-hosted frontier models can now point to Fable and say: “This is why we need local compute, local hosting, national AI capacity and sovereign cloud.” One Commerce Department letter may have done more for sovereign AI than a hundred conference panels in Dubai.

• The biggest economic damage is not only lost revenue; it is worse unit economics. Frontier AI is already expensive to train and serve. If the global user base shrinks while compliance, legal, localisation and duplicated infrastructure costs rise, the cost per monetised user goes up. The denominator shrinks; the bill does not.

• The Global South is especially exposed. Rich countries can respond to an off-switch by building sovereign AI stacks. Most African markets cannot. Without local compute, data centres and cheap power, the fallback option is weak. This is the same old infrastructure lesson: do not build your future on a supply line someone else controls.

• The historical parallels are not comforting. Semiconductors, cryptography, telecoms and ITAR-controlled technologies all show the same pattern: governments can slow rivals with controls, but they also raise costs, fragment markets and encourage foreign substitutes. Control buys time, but it also creates competitors.

• The likely outcome is not collapse, but repricing. Anthropic can still be enormous. OpenAI can still be enormous. AI can still be enormous. But after Fable, frontier AI deserves a higher sovereign-risk discount. The trillion-dollar number is not dead; it just has a government-sized asterisk beside it.

• The key question going forward is whether this stays at the frontier. If governments only restrict the most capable cyber, bio or military-relevant models, the industry absorbs the shock. If controls creep into ordinary commercial models, then the “global AI platform” dream starts looking more like a fragmented, regulated, lower-margin infrastructure business.

• My Argument: the model may be intelligent, but the off-switch is sovereign.

On Friday evening, the 12th of June 2026, at 5:21pm Eastern, a letter landed in Dario Amodei’s inbox and did something no government had ever done to a piece of commercial software running live in front of the public.

It told Anthropic to switch off its two most capable models.

By around 10pm, Claude Fable 5 and Claude Mythos 5 were dark. Not throttled, not geofenced, not “temporarily unavailable in your region.” Gone. For everyone, everywhere, including the company’s own engineers, three days after Fable 5 launched as the most capable publicly available model the company had ever shipped (Fortune, Bloomberg).

I write mostly about power plants, biomass boilers and solar leases, so you would be forgiven for asking why I am opening a clean-energy newsletter with an AI export control story. The answer is simple, and it is the spine of this whole piece: AI has become infrastructure, and the thing that just happened to it is the oldest story in my industry.

It is the story of building your entire operation on a single supply line that someone else can shut at will. We have lived that story with gas-to-power in Nigeria. Silicon Valley just got its first taste of it. And the bill, when it comes, lands hardest on people like us.

So let me pre-empt the question that every investor, founder and procurement officer may now asking, and then let me actually try to answer it with numbers as I mentally explain this to myself as well.

Anthropic was just valued at 965 billion dollars. After Friday, is that still a real number? Is frontier AI still a multi-trillion-dollar bet, or did the off-switch just put a ceiling on the whole thing?

Part 1: What actually happened

The US Commerce Department, through Secretary Howard Lutnick, sent Anthropic an export control directive (Axios, confirmed by a US official to Bloomberg). The order said Fable 5 and Mythos 5 could not be accessed by any foreign national, whether located outside the United States or inside it, and that explicitly included Anthropic’s own foreign-born staff (Anthropic’s statement).

Here is the operational trap. You cannot check the nationality of hundreds of millions of API and chat users in real time, at the moment of inference, on a shared cloud service. There is no toggle for “block foreigners only.” So the only way to comply with a foreigner-specific ban was to turn the models off for everyone. The global blackout was not the government’s literal instruction. It was the only state the company could reach that satisfied the instruction (Quartz).

Every other Claude model, including Claude Opus 4.8, kept running (Fortune). The order targeted only the top tier, the new “Mythos-class” frontier that Anthropic had priced at launch as a premium product, reportedly around $10 per million input tokens and $50 per million output tokens, nothing like a cheap commodity model. (Again, hold that detail). The state reached in and switched off the most expensive, most capable layer, and left the boring, cheaper models alone. It matters enormously later.

The stated reason was a jailbreak. The government had been shown a technique that strips Fable 5’s safety controls and unlocks the cyber-offensive capabilities of Mythos, the underlying model, specifically the autonomous hunting of software vulnerabilities (Al Jazeera, TIME). Anthropic does not dispute that a jailbreak exists. It disputes that it matters. The company argued the technique is narrow rather than universal, that it surfaces a handful of already-known minor flaws, and that the same trick works on other public models including OpenAI’s GPT-5.5 (Anthropic). Then it issued a warning that, read carefully, is a small bombshell. If this becomes the standard, the company said, it would “essentially halt all new model deployments for all frontier model providers.”

That is a billion-dollar firm telling the regulator: do this consistently and you stop the entire industry from shipping.

The twist in the Whole story that is as incomprehensible as it is uncomfortable

Now the part that turns a policy story into a thriller. The party that demonstrated the jailbreak to Washington was, according to the Wall Street Journal as reported across Axios and others, Amazon. Amazon’s own research team ran the prompts, and Amazon CEO Andy Jassy reportedly called administration officials on the Thursday night to flag the findings (Bitcoin.com News).

Read that again with the cap table open. Amazon has poured roughly 13 billion dollars into Anthropic and holds a roughly 100 billion dollar AWS infrastructure commitment from the company. Amazon is simultaneously Anthropic’s largest investor, its primary cloud host, and a direct competitor in the model market (MLQ). The entity that triggered a federal takedown of Anthropic’s flagship product had material financial and competitive interests in the outcome.

I am not going to tell you Amazon acted in bad faith, because I cannot prove motive and neither can anyone writing about this today. Security researcher Katie Moussouris of Luta Security, who reviewed the findings, called the government’s response a “complete overreaction” and noted the capability helps defenders more than attackers (Bitcoin.com News).

What I will say is this: the structure of the incident now contains a permanent fact. A commercial rival can be the party that pulls the trigger on a sovereign off-switch. That is a new variable in the game, and you do not need anyone to be a villain for it to reprice the entire board.

Three things are now established, and one is not.

• Established: the government has and will use a mechanism to disable a deployed model; the trigger can be a contested safety finding rather than an actual harm that occurred; and a competitor can be the one who supplies it.

• Not established: how far the precedent reaches. The whole valuation question lives inside that uncertainty.

For what it is worth, the betting markets think this specific episode gets walked back. Polymarket and Kalshi traders are pricing roughly a 68 to 71 percent chance that Fable 5 access is restored before the 1st of July (Bitcoin.com News). Anthropic itself called the order a “misunderstanding” and said it is working to restore access (Anthropic). The incident may well be temporary. The capability it demonstrated is permanent. Those are two completely different things, and most of the commentary I have read is confusing them.

Part 2: This did not come out of nowhere

To understand why this is structural rather than a one-off tantrum, you have to know the four months that preceded it.

This is the second front in a war that has been running since February. Defence Secretary Pete Hegseth met Amodei in February over a contract dispute. Anthropic had drawn two red lines: it would not let Claude be used for fully autonomous lethal weapons, or for mass surveillance of US citizens. The Pentagon wanted access for “all lawful purposes” and argued a private company cannot dictate how the state uses a tool in a national security emergency (CNN).

Talks collapsed. On the 27th of February, Trump and Hegseth ordered federal agencies to stop using Anthropic and designated the company a “supply chain risk,” a label normally reserved for firms tied to foreign adversaries and never before applied to an American technology company (NPR). On the 9th of March, Anthropic sued, alleging retaliation and violations of its First Amendment and due process rights (Al Jazeera). On the 26th of March, a federal judge blocked the designation, writing that the measures “do not appear to be directed at the government’s stated national security interests” (CNN). That is a judge using polite language to say this looks like punishment.

So by Friday, Anthropic was in a remarkable position, captured perfectly by Axios: simultaneously on a Pentagon blacklist that deems it too dangerous for the government to use, and inside a Commerce licensing regime that deems it too dangerous for foreigners to use (Axios). Same company. Same week. Opposite rationales. That contradiction is the tell. When the justification flips depending on which lever the state wants to pull, you are looking at a relationship that has become adversarial, with safety as the available pretext.

And here is the timing that can’t be ignored: Anthropic filed confidentially for its IPO around the start of June, and the off-switch came two weeks later (Fortune). I am not asserting cause and effect. I am noting that any honest valuation model now has to price the possibility that the trigger gets pulled again, for reasons that may have nothing to do with cybersecurity.

Part 3: The money, before we even get to fragmentation

Let us anchor the valuation, because the headline number is genuinely staggering and genuinely earned, up to a point.

On the 28th of May, Anthropic closed a 65 billion dollar Series H at a 965 billion dollar post-money valuation, led by Altimeter, Dragoneer, Greenoaks and Sequoia (CNBC, Anthropic). That vaulted it past OpenAI’s 852 billion dollar mark from March (Bloomberg). An October IPO window has been floated (Tech Insider).

What justifies a near-trillion-dollar mark is the slope. Run-rate revenue crossed 47 billion dollars in May, up from roughly 9 billion at the end of 2025 (Anthropic, via Digital Applied). Note that “run-rate” means a recent month multiplied by twelve, not audited annual revenue, so treat the trajectory as real and the precise level as an estimate. Still, going from a 380 billion dollar valuation in February to 965 billion in May is one of the steepest re-ratings in the history of private capital. Investors are not pricing the company. They are pricing the derivative.

Now the other half of the ledger, the part the slope obscures.

Frontier AI is, today, a structurally loss-making business held up by equity. Anthropic reportedly generated around 918 million dollars in revenue in 2024 while burning around 5.6 billion (investor data reported by Reuters, cited in the analyses underpinning this piece). The pattern across the field is the same. Training costs for frontier models are climbing roughly 2.4 times a year, with the largest runs projected to break a billion dollars by 2027. Inference costs are falling roughly tenfold a year per unit of performance, which sounds like salvation until you realise the fixed training bill does not shrink in proportion. The result, by several accounts, is that frontier inference is priced below true cost to win market share. OpenAI has been estimated to spend somewhere near 2.25 dollars to earn each dollar of revenue once you load in everything.

The hard evidence for this is now visible in margins. ICONIQ’s January 2026 survey of around 300 software executives put AI-native gross margins at roughly 52 percent in 2026, up from 41 percent in 2024, against the 75 to 85 percent that traditional software enjoys. Inference alone runs at about 23 percent of revenue for scaling AI firms. Anthropic’s own inference costs reportedly overran internal expectations by about 23 percent, with gross margin around 40 percent, which is why it started charging enterprise subscribers the full compute cost in April. That 30-point gap between AI margins and software margins is an inference-cost gap. Keep it in mind, because sovereign fragmentation widens it directly.

So before Friday, the picture was already this: a phenomenal revenue ramp, sitting on top of an economic model that has not yet proven it can make money, funded by the belief that scale will eventually fix the unit economics. Everything depends on scale. Which is exactly what the off-switch attacks.

Why a foreigner ban is a margin event, explained with arithmetic

Here is the cleanest way I can show you why geographic restriction is not just a top-line problem but a unit-cost problem. This is the bit my industry understands in its bones.

A frontier training run is a fixed cost. You pay it once, whether you serve ten users or ten billion. So imagine a one billion dollar training run.

The cost did not change. The denominator shrank. Your unit cost just rose 67 percent while you were sleeping. And you absorbed that hit in an industry that was already pricing below cost to begin with.

This is why fragmentation is so corrosive to the trillion-dollar thesis. It does not merely cap how much revenue you can earn. It simultaneously degrades the denominator of your cost problem. You lose the high-volume international traffic that was helping you amortise the fixed bill, and you take on new costs: nationality-verification systems, jurisdiction-aware routing, duplicated serving stacks, compliance teams, legal overhead. The ITAR compliance industry for traditional defence technology is a multi-billion-dollar overhead sector. AI’s version will not be cheaper.

There is a genuine counter-force, and intellectual honesty demands I name it. Inference cost per token really is falling fast, thanks to new hardware generations and tricks like prompt caching and attention reuse. That lowers the absolute cost base even as restriction raises the cost per user. The net effect on margins is a tug-of-war between those two trends, and theory does not settle who wins. That tug-of-war is the single most important number to watch in this whole story. Anyone who tells you they know the answer is selling something.

Part 5: Why AI is a stickier trap than anything before it

Now to the intellectual core, the idea that makes this more than a news cycle. There is a body of political-economy work, most cleanly stated by Henry Farrell and Abraham Newman in their theory of weaponized interdependence, that explains exactly what Friday was.

The argument runs like this. Modern economic networks are not flat. They funnel through a few central nodes. A state that has legal jurisdiction over one of those nodes, plus the institutions to act on it, can weaponise that position in two ways. It can watch everything flowing through the node (Farrell and Newman call this the panopticon effect), and it can deny access to the network entirely (the chokepoint effect). Their classic example is the US using its jurisdiction over dollar clearing and the SWIFT messaging system to enforce sanctions on Iran. Cut a target off from the financial plumbing and you do not need to fire a shot.

Friday was a textbook chokepoint exercise. A state with jurisdiction over the node (a US-incorporated lab, running on US-controlled cloud and US-controlled chips) denied network access to a class of users (foreign nationals) for a strategic end (containing autonomous cyber capability). Both enabling conditions were satisfied.

The decisive question the framework forces is: how durable is the node? And this is where AI turns out to be a far nastier trap than its closest historical cousin.

Compare it to the Crypto Wars. From the late 1970s to 2000, the US treated strong encryption as a munition, controlled its export, and even tried to mandate a government backdoor (the Clipper Chip). That regime eventually collapsed and liberalised. Why? Because once you export a piece of cryptographic software, it runs anywhere, on any machine, beyond the reach of the controller. The chokepoint was leaky. Foreign substitutes appeared, US firms lost sales, a court ruled that source code was protected speech, and by 2000 the controls were gutted. The whole thing took roughly two decades and required real commercial pain plus a favourable judge.

Hosted frontier AI is the opposite kind of node, and this is the part people miss. A model served from a data centre is not a shipped artifact. It is an ongoing service. It stays tethered to scarce, observable, controllable compute. It can be watched and interrupted at the node in real time. The panopticon and the chokepoint coincide. On top of that, the chips themselves are already under a separate, mature US export-control regime that has been tightening since October 2022, and the frontier is held by a handful of firms inside US or allied jurisdiction.

The conclusion is stark. Frontier AI is a more durable chokepoint than encryption ever was. The leak that eventually freed cryptography (it runs anywhere once exported) does not exist for a hosted model. That single fact is why the ceiling on frontier valuations is real, and simultaneously why the picture is not pure catastrophe. The same logic that makes hosted models controllable makes open-weight models uncontrollable. Once a model’s weights are released, they behave like exported crypto: diffuse, unobservable, beyond the node. The policy will therefore bifurcate. Tight, effective control over hosted frontier services. Leaky, ineffective control over open weights. We can already see this in the carve-outs for open models in existing US rules.

The other historical analogues all rhyme with the same warning, that control over a diffusible technology has a price and seeds its own competitors:

The TikTok ban established the legal and political machinery for forcing a digital platform offline. But TikTok was foreign-owned. Friday’s action was aimed at a domestically owned American firm. That is a more expansive use of state power, not a less one.

Part 6: The valuation math, made human

So can the 965 billion dollar number survive? Let us do the arithmetic the way a sober analyst would, not the way a hype cycle does.

A valuation is, at bottom, a claim about future cash flows discounted back to today. There is a simple identity for what a one-trillion-dollar enterprise value actually requires you to earn, forever, in steady state. At a 10 percent cost of capital and 3.5 percent long-run growth, you need roughly 63 billion dollars of free cash flow every year, in perpetuity. Nudge the cost of capital to 11 or 12 percent (which is exactly what rising risk does) and the requirement climbs to 78 to 87 billion dollars.

Translate that into revenue, and the gap with reality opens up:

Anthropic’s run-rate revenue is 47 billion dollars and its gross margin is around 40 percent, with free cash flow still negative. The table is asking for 200 to 440 billion dollars of sustainable revenue at healthy margins. The growth has been astonishing, and the gap is not unbridgeable in a vacuum. But now layer Friday on top, and watch what happens to the discount rate.

Sovereign risk widens the cost of capital. A common rule of thumb adds 150 to 250 basis points to the equity risk premium when policy and export risk are real. In a high-multiple business like this, that is not a rounding error. A 200-basis-point rise in the cost of equity compresses terminal value by roughly 15 to 25 percent in a standard two-stage discounting model. And there is a deeper mechanism, also from the asset-pricing literature (Pástor and Veronesi): as a revolutionary technology becomes pervasive, its risk migrates from idiosyncratic (specific to one firm) to systematic (affecting the whole sector). A demonstrated off-switch accelerates exactly that migration. Systematic risk gets a higher discount rate, full stop.

This gives us the single cleanest test of how the market is actually reading Friday, and it is something you and I can watch in the coming months. The political-uncertainty model makes a sharp prediction. If the market sees this as Anthropic’s idiosyncratic stumble, then competitors should be flat or should rise on relative advantage. If the market sees it as systematic regulatory risk, then every frontier lab should reprice downward together, in tandem. Correlated repricing across the whole sector is the signature of the dangerous reading. Watch the next funding rounds. Watch the Anthropic IPO. That cross-section will tell you more than a thousand op-eds.

Part 7: Why this hits the Global South hardest (the part I came here to write)

If you are building anything in Africa, South Asia or Latin America on top of frontier AI, Friday should have changed how you sleep.

Start with the brutal asymmetry. Most of the African continent, South Africa aside, is what researchers at New America bluntly call a compute desert. The continent holds about 2 percent of the world’s data centres. The US data-centre sector consumes over 540 kWh per person in electricity for AI workloads. Africa uses under 1 kWh per person. That is a roughly 600-fold gap in raw AI capacity.

Here is why that number is a catastrophe in the specific context of an off-switch. The natural defence against “my frontier model can be switched off” is “then I will self-host an open-weight model on local compute.” That is precisely the escape hatch that the open-weight leak provides, the one that eventually freed cryptography. But that escape hatch requires compute that does not exist here at scale. A developer in Accra or Abuja cannot meaningfully respond to a sovereign restriction by spinning up their own frontier-class inference, because the data centres, the power and the chips are not on the ground. So we get the worst of both worlds: excluded from the hosted frontier by the scope of a US export control, and unable to substitute because of the compute desert.

This is the exact shape of a problem I deal with every week in energy, and the analogy is structurally identical.

Consider the Nigerian power sector. A huge share of grid generation depends on gas. When gas supply is disrupted, by pipeline vandalism, by payment disputes, by upstream constraints, plants that are physically fine sit idle because the fuel stopped arriving. The asset is sound. The supply line failed. You built a power system on a single input with an off-switch you do not control, and one day someone, somewhere, flips it. This is why every serious developer I know now obsesses over fuel security and supply diversification. It is why my own work leans so hard into hybridisation and contracted, multi-country supply rather than betting a whole plant on one fuel from one source. You do not put a single point of failure at the base of critical infrastructure if you can possibly avoid it.

Frontier AI, embedded into engineering workflows, research, commercial intelligence and increasingly into the productivity of an entire economy, is now exactly that kind of input. Functional under normal conditions. Brittle under supply shock. And the supply shock just got demonstrated.

The macro numbers make the development stakes concrete. The World Economic Forum’s fragmentation work puts the cost of geopolitical fragmentation at 213 to 307 billion dollars a year in current conditions, rising to as much as 6.9 trillion dollars, about 6.4 percent of global GDP, in a severe scenario. And emerging markets outside the major blocs do not take the average hit. They take an estimated 10.7 percent hit, roughly 40 percent worse than the global average. The European Central Bank adds the data-quality dimension: when data is walled off by localisation rules, models become “parochial and brittle,” good at the familiar and bad at the edge cases of a global economy. The edge cases, of course, are us.

So my read for anyone building in our markets is unsentimental. Do not treat frontier AI as stable infrastructure. Treat it as imported fuel. Use it, absolutely. It is too powerful to ignore and the productivity gains are real. But architect for its removal. Keep an open-weight fallback in your designs even if it is less capable. Avoid hard-wiring a single foreign frontier model into anything mission-critical. Diversify your providers the way you would diversify fuel supply. The cost of that optionality used to look like paranoia. After Friday it looks like basic engineering discipline.

Part 8: Sovereign AI, the response that splits and grows the market at once

Friday is also the best advertisement the sovereign-AI movement has ever received, and that movement is the demand-side mirror image of the chokepoint.

The pitch, pushed hardest by Nvidia since 2023, is simple: a nation cannot depend on a model whose host country can watch, condition or sever its access, so it must build or buy its own stack on its own soil. Every sovereign-AI programme is, in effect, a bet that the chokepoint is real and that the buyer is the spoke. Friday converted that bet from hypothetical to demonstrated.

And the capital is genuinely enormous, though I would caution that most of these figures are analyst projections with promotional incentives baked in, so read them as direction rather than gospel:

Here is the crucial economic subtlety, the one that keeps this from being a simple doom story. Sovereign fragmentation splits demand and simultaneously enlarges it. The same capability now has to be deployed many times, once per jurisdiction, rather than served once globally. The total number of deployments goes up. But each one is sub-scale, locally compliant and runs at lower utilisation. So demand grows, which is good for the sellers of capability, while margin compresses, which is bad for whoever has to operate all those duplicated stacks. Walls create captive markets as well as excluded ones. Gartner already projects that 35 percent of countries will be locked into region-specific AI platforms by 2027, up from 5 percent today.

There is even a partial escape for the incumbents. If those sovereign deployments consolidate onto a few US-jurisdiction infrastructure stacks (the emerging Nvidia-and-Palantir “sovereign AI operating system” template is the model), then the winner-take-most economics survive one layer down, at the infrastructure level, even as the model layer fragments. The real question is not whether demand fragments. It will. The question is at which layer of the stack the scale economics survive. That, too, is something we can watch.

And one more pressure worth flagging, because it deepens the capital problem at the worst possible moment. A whole second frontier is opening up: world models, the spatial-intelligence push from Fei-Fei Li’s World Labs (which has raised over 1.2 billion dollars), Nvidia’s Cosmos, Google DeepMind’s Genie 3, and Yann LeCun’s new venture. These shift compute from text toward video and physics simulation, which is more compute-hungry, not less. The spend is additive to the existing LLM bill, not a substitute for it. Goldman Sachs models roughly 7.6 trillion dollars of cumulative AI capital between 2026 and 2031. And world models carry the very same dual-use exposure (robotics, synthetic sensor data) that grounded Friday’s cyber-capability action. So the second frontier is not a refuge from the first frontier’s sovereign risk. It inherits it.

Part 9: Who actually captures the value

So if the frontier lab is the exposed tier, the next question is: within AI, who is actually safe? Because the value does not vanish on June 12. It moves. And it moves in a direction my industry would recognise on sight.

In energy, the merchant power plant with one offtaker and one fuel line is the asset that dies in a shock. The vertically integrated utility that owns the generation, the wires, the fuel and the customer keeps paying dividends through the same shock. The off-switch is a body blow to the merchant frontier lab. It is a scratch on the vertically integrated giant.

Look at Alphabet. On the day Fable went dark, it was worth more than four trillion dollars on the public markets. Its Q1 2026 results, filed with the SEC, showed quarterly revenue of 109.9 billion dollars, up 22 percent, at a 36.1 percent operating margin (Alphabet 8-K, Q1 2026). Google Cloud grew 63 percent to 20.0 billion dollars in the quarter, with backlog nearly doubling to over 460 billion dollars, having already crossed a 70-billion-dollar annual run-rate at the end of 2025 (Alphabet Q4 2025). And it is funding its entire AI buildout, guided at 175 to 185 billion dollars of capex in 2026, out of its own cash flow rather than someone else’s equity cheques.

Now hold that against the off-switch. Alphabet has Search, Android, YouTube, Workspace, its own TPU chips, its own data centres, its own global distribution, decades of cash flow, and a sovereign-cloud product line built around data residency and local control. AI is one engine in a machine that has a dozen of them. If frontier-model margins get compressed by fragmentation, Alphabet shrugs. A pure-play lab whose entire valuation assumes frictionless global scale on a single premium tier cannot shrug. It has nowhere to put the hit.

That is the more useful reframing than “is AI doomed.” AI stays multi-trillion. But the multi-trillion value is safer in the hands of the diversified infrastructure owners, the hyperscalers, the chip companies and the sovereign-cloud sellers, than in the hands of the pure-play frontier labs the venture market has marked most aggressively. In a fragmented world the boring, vertically integrated incumbent beats the glamorous demo more often than not. That does not make Anthropic or OpenAI un-investable. It makes them the higher-beta bet on the same wave, the place where the sovereign risk concentrates instead of dissipating. If you want exposure to the AI build-out without the off-switch sitting at the base of your thesis, you buy the picks, the shovels and the land, not the prospector.

Part 10: The verdict, in three scenarios

Let me put my cards on the table with explicit probabilities, because hedging without numbers is cowardice.

Scenario 1: Contained incident (roughly 35 percent). Anthropic’s rebuttal wins, the order is reversed within weeks, the jailbreak is patched, and Commerce builds a proper, transparent review process with due-process safeguards. The IPO proceeds near plan. The sovereign risk premium reprices modestly, maybe 50 to 75 basis points, but does not break the model. This is what the prediction markets and the muted overall reaction are currently betting on.

Scenario 2: Escalating restriction (roughly 40 percent, and my base case). The order is resolved but the precedent sticks. Future frontier releases face mandatory pre-deployment review. Labs build nationality verification and capability routing. AI capability gets tiered by geography: full frontier access for citizens, version-controlled access for allies, degraded or no access for everyone else. This is the semiconductor playbook applied to inference. Valuations compress by 20 to 30 percent as the addressable market is geographically gated and compliance costs inflate. The Global South is effectively fenced out of the frontier.

Scenario 3: Systemic fragmentation (roughly 25 percent). Controls expand to more capability categories. The EU nationalises its data flows for frontier inference. China’s independent stack accelerates. Open-weight models close the gap fast. The global market bifurcates and US frontier labs get valued on US-only revenue multiples, maybe 40 percent of the global pie, while the rest goes to sovereign and open-source alternatives. This is the WEF’s severe fragmentation outcome, the one that knocks up to 7 percent off global GDP.

So, the headline question: Is AI still a multi-trillion-dollar bet?

My honest answer is bounded yes, not a death sentence, but priced wrong as it stands. A frontier lab is not made un-viable by a sovereign off-switch. Anthropic can plausibly still be a trillion-dollar company by headline valuation. But Friday attacked all three pillars of the trillion-dollar story at once: global reach (now revocable at the node), frontier leadership (now the locus of the highest regulatory risk), and margin via scale (now exposed to fragmentation). A higher cost of capital is warranted even in the optimistic scenario. The late-stage private marks almost certainly do not yet reflect it. The trillion-dollar number is not dead. It just got a structural discount that the cap table has not caught up to.

And the question underneath, whether the net extends from the frontier to the everyday commodity models the rest of the world runs on? The lay intuition that “nothing stops it from spreading” is correct about the government’s legal capability and wrong about the likely outcome. Nothing in the statute confines control to the frontier. There is even a textual brake, an exception in the law for a “finished item generally made available” to the public, which is exactly what a broadly deployed commercial model is. But almost everything in the political economy resists the spread: the sheer domestic cost of broad controls (the same brake that eventually freed cryptography), the existence of open-weight and foreign alternatives, and the friction of courts, Congress and allies. The broad-fragmentation scenario is a genuine tail risk that got materially fatter on Friday. It is not the central case on today’s facts.

My Watchlist

Because this is a story still being written, here is the short list of signals that will tell us which scenario we are heading into. You can track these yourself.

• The target of the next action. If a future restriction hits a broadly deployed, mid-tier commercial model rather than a top cyber or bio capability, the finished-item brake is breaking and we are sliding toward fragmentation.

• Cross-lab repricing. Correlated downward marks across all frontier labs at the next funding rounds, and at the Anthropic IPO, means the market reads this as systematic. That is the dangerous reading.

• Rules versus discretion. Published, predictable capability thresholds would be reassuring. Opaque, case-by-case, competitor-triggerable action is the opposite.

• The courts. A First Amendment challenge treating model weights or outputs as protected speech, echoing the case that freed cryptography, would brake everything. Its absence leaves the discretion fully intact.

• A non-US frontier champion. This is the single most decisive one. A credible Gulf, EU or Asian frontier lab, exactly what HUMAIN and InvestAI are reaching for, would manufacture the “foreign availability” that broke the cryptography chokepoint, and would shift the whole system toward fragmentation for any single US incumbent.

Coda: The off-switch is real now

I keep coming back to the thing my industry learned the hard way, long before Silicon Valley had to.

You do not build the base load of critical infrastructure on a single supply line that someone else can shut. We learned it from gas. We are still learning it from grids. And on Friday the 12th of June 2026, at 5:21pm Eastern, the most sophisticated technology companies on the planet got their first proper lesson in it, delivered by one letter.

The off-switch is real now. It has been pulled once. The markets, the founders, and every sovereign that was on the fence about building its own stack all saw it happen. The trillion-dollar story is not over. But it is no longer a story about frictionless global scale and ever-falling costs. It is a story about access that can be revoked, capability that draws the most fire precisely because it is the most capable, and margins that depend on a global market the state has just shown it can carve up.

For those of us building in the parts of the world most exposed to that carving, the instruction could not be clearer. Use the tools. Marvel at them. And never, ever, wire your future to a fuel you cannot replace.

This is analysis, not investment advice, and certainly not legal advice. The event facts are sourced to public reporting current to 14 June 2026 and may be overtaken by events; the underlying government directive is not public, so the precise legal basis is known only through Anthropic’s account. Several economic figures are estimates or forward projections from named secondary sources and are flagged as such in the text. If access to Fable 5 is restored next week, none of the structural argument changes. That is rather the point.

Sources and further reading

The event and the immediate fallout: Anthropic’s official statement; Axios and its follow-up on Amazon’s role; Bloomberg; TIME; Fortune; Al Jazeera; Quartz; prediction-market pricing via Bitcoin.com News.

The valuation: Anthropic Series H announcement; CNBC; Bloomberg; TechCrunch.

The hyperscaler comparison: Alphabet’s own SEC filings, Q1 2026 results (revenue, Google Cloud, margin, cloud backlog) and Q4 and full-year 2025 results (cloud run-rate, 2026 capex guidance).

The prior Pentagon conflict: NPR; CNN on the lawsuit and the injunction; Al Jazeera.

The frameworks behind the analysis: Farrell and Newman on weaponized interdependence (International Security, 2019); Pástor and Veronesi on political-uncertainty risk premia (Journal of Finance, 2012; Journal of Financial Economics, 2013) and on technological revolutions and stock prices (American Economic Review, 2009); the BCG/SIA semiconductor supply-chain study; New York Fed research on the market-cap cost of export controls; CSIS on ITAR-free satellites; ICONIQ’s State of AI survey (January 2026); McKinsey on the sovereign-AI agenda; Goldman Sachs Global Institute on AI capital expenditure; and the World Economic Forum and ECB on the macroeconomics of fragmentation.

In Summary

In May 2026, the Nigerian Electricity Regulatory Commission (NERC) signed the Net Billing Regulations 2026 (Regulation No. NERC-R-002-2026), made under section 226 of the Electricity Act 2023. For the first time, there is a clear, national rulebook for connecting your solar system to the grid and getting paid for the surplus you push back.

Here is the whole thing in five lines:

• It is net billing, not net metering. You pay full price for what you take from the grid, and you get a separate, lower price for what you send back.

• It covers solar systems from 50 kWp up to 1.5 MWp, one premises at a time (s5, s6(2)).

• Your export credit can only ever reduce your bill, never turn into cash, and your bill can never go below zero (s19(8)).

• The export price is built from the grid’s wholesale cost and then deliberately discounted: a factor of 0.55 for daytime (off-peak) and 0.75 for the 6pm to 9pm peak window (s19(5)).

• And my view on this is that, because of all that, the real money is in using your own solar, not in selling it. In the worked example below, self-consumption delivers about 92% of the value and grid exports about 8%.

There is a companion Excel calculator with this article. Plug in your system size, your load, and your DisCo’s tariff numbers, and it shows your savings, your export credits, an eligibility check, and where the value sits. Link at the end.

Thanks for reading! Subscribe for free to receive new posts and support my work.

First, the one word that changes everything

There are two ways a country can pay you for rooftop solar, and they are worlds apart.

Net metering lets you bank kilowatt-hours. Your meter spins backwards. A unit you export in the afternoon cancels a unit you import at night, one for one, at the same retail price. It is simple and generous, and it is what most people picture when they hear “selling power back to the grid.”

Net billing, which is what Nigeria has chosen, is different. Think of it like a shop credit. You bring the store your extra tomatoes; the store does not hand you cash, it records a credit on your account, and that credit knocks money off your next purchase of rice. Under the regulation, you buy from the grid at the full Retail Tariff (s19(1)) and you sell to the grid at a separate, lower Export Tariff that becomes a bill credit (s19(2)). The two are netted in naira on your monthly bill.

This is not a small accounting nuance but the entire design philosophy, and every other rule points the same way: self-consumption is where the value is, and export is a relief valve, not a revenue line. NERC has engineered it so you can never turn the grid into a profitable buyer of cheap daytime solar. Whether you think that is fair or stingy, it is the reality you have to plan around.

Why the rules are cautious: a quick look at the grid

It helps to know why the regulation is so conservative, because once you see the grid it sits on, the design makes sense.

Across 2025, the grid had on average only around 5,400 MW of available generation capacity, and actually delivered closer to 4,500 MWh per hour, according to NERC’s quarterly reports. That is very little power for a country of Nigeria’s size and ambition. The commercial picture is just as strained: in the fourth quarter of 2025, NERC reported aggregate technical, commercial and collection (ATC&C) losses of 34.90 per cent, and collection efficiency that dropped slightly to 79.36 per cent from 80.70 per cent in the previous quarter. On top of that, the metering gap is huge: as of late 2025 only about 55% of registered customers were metered, leaving roughly 5.36 million on estimated bills.

So NERC is balancing two things at once. Customers genuinely need cheaper, cleaner, more reliable power. But the DisCos cannot be turned into unpaid battery banks for uncontrolled exports, and weak feeders cannot absorb unlimited reverse flow. The result is a scheme that relieves demand (good) without creating an open-ended payment liability or letting any one feeder be overwhelmed (cautious). Hold that in mind, because it explains the discount on exports, the no-cash rule, and the feeder cap you are about to meet.

Who this is actually for

The regulation applies to (s5, s6):

• Solar only, for now. Small wind and hydro can be added later once NERC publishes technical standards for them.

• System size from 50 kWp to 1.5 MWp per user. Below 50 kWp you are out of scope. Above 1.5 MWp you are out of scope and into the world of bilateral and wheeled arrangements instead.

• One premises, for your own use. You must own or legally hold the system, and you must consume the power on the same site where it is installed (Schedule 3, clauses 1.1 and 1.2).

• Connected at 0.4kV, 11kV or 33kV (s6(3)).

So this is a framework for homes (the larger ones), and for small-to-mid commercial and industrial sites: a factory annex, a cold store, a shopping plaza, a hospital, a campus. If your site is bigger than 1.5 MWp, net billing is a complement to your other options, not a replacement.

One boundary quietly matters a great deal. The feeder limit in s6(3) says the total surplus that all prosumers push into any one network asset cannot exceed 30% of that asset’s average load, and access is first-come, first-served (s6(1)). On a busy industrial feeder, that headroom can fill up. The early applicant gets in; the next one can be blocked, no matter how good the project is. Feeder headroom is, in effect, a commercial asset, so lodge early.

The money: how your export price is built

Here is the formula that decides what you earn for every unit you export (s19(4)):

Export Tariff (ET) = Avoided Cost Delivered (ACD)  x  Export Tariff Factor (ETF)

                          GC + TC
where      ACD  =  -------------------
                          1 - TLF

In plain English, two steps.

Step one: work out the grid’s avoided cost. When you export a unit, you save the DisCo from buying that unit through the normal supply chain. That saving is the Avoided Cost Delivered (ACD). It is built from the Generation Cost (GC) plus the Transmission and admin cost (TC), then grossed up for energy lost in transmission using the Transmission Loss Factor (TLF) (s19(4)). All three come from your DisCo’s MYTO, the Multi-Year Tariff Order.

Notice what ACD is, and what it is not. It is the wholesale cost of getting energy to the edge of the distribution network. It deliberately leaves out the distribution wires, the losses inside the distribution network, and the DisCo’s retail margin. So even before any discount, the grid values your export at the wholesale gate price, not the retail price you pay at your meter.

Step two: apply the discount. NERC multiplies that ACD by the Export Tariff Factor (ETF), fixed at (s19(5)):

• 0.55 for off-peak (daytime, when solar actually generates), and

• 0.75 for peak (the 6pm to 9pm window, s4).

So the DisCo keeps 45% of the wholesale avoided cost on daytime exports, and 25% on peak exports. Put the two steps together and you get the punchline: you are paid less than it costs the grid to buy the same unit at the wholesale gate, and far less than the grid charges you at your meter. You are, in effect, a discounted upstream supplier.

Let us put real numbers on it

Two of the three inputs (Generation Cost and Transmission cost) live inside each DisCo’s MYTO and change with the dollar, gas and inflation, so I will be honest and label them illustrative: you must replace them with your DisCo’s values. The third input, the Transmission Loss Factor, we can pin down from a real source: NERC’s Order No. NERC/2026/026 (8 April 2026) records the national average TLF as 8.71 per cent in 2024, reduced to 7.24 per cent in 2025, against a 7% benchmark in the MYTO. I will use 7.24%.

Run those through the formula:

• ACD = (100 + 18) / (1 - 0.0724) = NGN 127.21 / kWh

• Off-peak export tariff = 127.21 x 0.55 = NGN 69.97 / kWh

• Peak export tariff = 127.21 x 0.75 = NGN 95.41 / kWh

Now compare those to what you pay. Through 2024 and 2025 the Band A retail tariff has sat at roughly NGN 209 per kWh (NERC’s cap is NGN 225, and Ikeja Electric reduced its own Band A rate to NGN 206.80). Commercial and industrial maximum-demand customers pay even more.

Read that table twice. A unit you export in the afternoon is worth about a third of a unit you import. Even in the best case, a peak export is worth less than half. That single fact is the foundation of every sensible decision under this regulation.

There is also a safety catch, s19(6): if your off-peak export tariff ever works out higher than your own retail tariff, it is capped at your retail tariff times the factor. For Band A and commercial customers this never bites, because your retail tariff is well above the avoided-cost-based export price. It is really there to stop arbitrage by heavily subsidised low-band customers. (The peak clause, s19(7), is drafted clumsily, but the evident intent is that the peak export tariff is uniform across all bands and is not subject to that retail cap.)

A worked example: a 500 kWp factory

Let us make it concrete with one site, using the same illustrative tariffs. Imagine a commercial site with:

• A 500 kWp solar system, producing about 800,000 kWh a year (at an illustrative yield of 1,600 kWh per kWp).

• Annual electricity use of 1,200,000 kWh, so the site is comfortably bigger than its solar (it stays a net importer, which is the healthy, normal case).

• An 80% self-consumption rate: four out of every five solar units are used on site, and one is exported.

Here is how the year shakes out (these are the calculator’s outputs):

Energy Naira value Solar used on site (self-consumed) 640,000 kWh NGN 133.76m saved (at NGN 209 avoided import) Solar exported to the grid 160,000 kWh NGN 11.2m earned (at NGN 69.97 off-peak) Total annual benefit NGN 144.95m

And the before-and-after on the bill:

• Annual bill without solar: 1,200,000 x NGN 209 = NGN 250.8m

• Annual bill with solar (net of export credits): NGN 105.85m

• Total annual saving: NGN 144.95m

Now look at the split, because it is the whole point:

Self-consumption: NGN 133.76m, which is 92% of the benefit. Export credits: NGN 11.2m, which is 8% of the benefit.

The grid exports are real money and worth having. But they are the garnish, not the meal. If you designed this system around selling to the grid, you would have designed the wrong system. You design around using your own power; export is what you do with the leftovers.

The trap

The marketing instinct will be to sell “earn money from the grid.” The regulation makes that a weak pitch, and here is the cleanest way to see why.

Every unit you use yourself is worth the full retail tariff you would otherwise have paid: about NGN 209 in the example. Every unit you export is worth about NGN 70. So a self-consumed unit is worth roughly three exported units.

This has two practical consequences.

First, right-size the system to your daytime load. The regulation lets you install export capacity up to 120% of your Eligible Load Demand (s6(4)), and it even lets your panels (DC) exceed your approved export (AC) so long as you cap the export with certified inverter settings (s6(8)). That is a gift for solar design: you can oversize the array to squeeze out more yield and clip the rare peaks. But oversizing purely to export more is value-destruction, because every extra exported unit earns a third of what a self-consumed unit saves.

Second, the regulation will not reward a system that produces more than the site consumes. And if you build one anyway, the next rule bites.

The exception worth knowing: buildings that sit empty at weekends

There is one important exception to “self-consumption is everything”, and it is easy to miss because it turns on the calendar, not the clock.

A great many commercial buildings run Monday to Friday and sit nearly empty at the weekend: offices, banks, a lot of schools, government buildings, some warehouses. Their solar does not know it is Saturday. The panels keep generating right through the weekend, but there is almost no load to absorb it. Before net billing, that weekend energy was simply lost, either spilled by the inverter or never harvested at all. Two days out of seven, more than a quarter of the system’s annual output, producing nothing of value.

Net billing changes that. On weekdays the building behaves like any other daytime site and uses most of its solar directly. At the weekend, with the building idle, almost all of the generation is exported and earns a credit. An exported unit is still worth only about a third of a self-consumed one, so this is not a licence to oversize. But it converts power that used to be thrown away into a real line on the bill.

In the calculator, this is the “Weekdays only, idle weekends” usage pattern. Switch a 500 kWp office to it and the exported share roughly doubles, from about 8% to around 18% of generation, and the annual export credit climbs from roughly NGN 11m to about NGN 22m. That extra NGN 11m is money the building could never capture before this regulation existed. For a weekday-only site, the export credit stops being a rounding error and becomes a genuine, if still secondary, part of the case.

if your building is dark at the weekend, net billing is worth more to you than it is to a 24/7 factory, because a bigger slice of your generation has nowhere else to go. Size for your weekday daytime load as usual, then let the weekend surplus earn its keep instead of spilling it.

The annual reset, and why over-building is punished

Here is one of the sharpest clauses in the whole document, and it is easy to miss.

When your monthly export credit (plus any credit carried over) is bigger than your monthly import bill, you pay nothing that month, and the surplus rolls forward as carried-forward credit (s19(7)). So far, so reasonable. But s21(2) says that carried-forward credit is netted off at the anniversary of your connection. In plain terms: use it within twelve months or lose it. You get 30 days’ written notice before it expires (s21(4)), and that is it.

For a properly sized self-consumption system, this never hurts you, because you are a net importer every single month: your import bill always swallows your export credit, nothing accumulates, and nothing is forfeited. The calculator shows this clearly: in the 500 kWp example, the carried-forward balance sits at zero all year.

But if you over-build for export, you will spend part of the year banking credits you cannot use, and watch them evaporate at the anniversary. The regulation is, once again, telling you the same thing in a different way: do not build a merchant export plant under this scheme.

A few more credit rules worth knowing:

• Credits are tied to the premises, not to you. If you sell or hand over the property and transfer the agreement, the credits go with it (s22(2)).

• If you move the solar system to a new site, the credits are zeroed and you start a fresh application (s21(3), s22(3)).

• The DisCo must keep a monthly credit ledger for every prosumer, in a format NERC approves, and show it to you on request (s21(5), s20(4)). Ask to see it.

Read this before you buy a battery

The regulation dangles a premium: peak exports earn a factor of 0.75 instead of 0.55, a roughly 36% uplift. But that premium is gated behind a battery. To qualify for the peak rate, you need a NEMSA-verified Battery Energy Storage System (BESS) with usable capacity of at least two hours at 50% of your installed solar capacity, able to charge from the solar and discharge to the grid (s17(6)). Without a qualifying battery, your inspection certificate is stamped “Off-Peak Rate Only” and every export is paid at 0.55. There is a second gate: if a time-of-use meter is not available when you commission, all exports default to off-peak until one is fitted, within twelve months (s18(5)).

Now here is the catch the regulation quietly creates. The peak window is 6pm to 9pm (s4), which is after sunset. Solar generates in the daytime, which is off-peak, and produces nothing during the peak window. So a solar-only system can only ever earn the 0.55 rate. The only way to touch the 0.75 premium is to store daytime energy and discharge it to the grid between 6 and 9pm.

For a 500 kWp system, the qualifying battery is roughly 500 kW for two hours, so about 1 MWh of usable storage. That is a serious capital cost, and the premium it unlocks is small: in our example, shifting a quarter of exports into the peak window adds only about NGN 1m a year. A 36% uplift, applied to a three-hour window, on energy worth NGN 95 a unit, does not pay for a megawatt-hour of batteries on its own.

The deeper point: a battery’s most valuable job is almost never peak export. A stored unit discharged in the evening to power your own site saves you the full retail tariff, about NGN 209. The same stored unit exported at the peak rate earns about NGN 95. So any rational operator uses the battery for evening self-consumption first, which is worth more than twice as much. The peak export premium only matters for whatever surplus is left after the site’s own evening needs are met, which for most sites is very little.

The honest read: NERC is signalling that it wants storage on the grid, not that it is paying enough to fund it. Build the battery on the full value stack (shifting solar into the evening, managing demand charges, riding through outages) and treat the peak export premium as a small bonus, not the business case.

Metering and safety

You cannot export a single unit until the system is wired safely and metered properly. Two chapters of the regulation cover this, and a couple of details trip up real projects.

On safety and interconnection (s17): the system connects at a single point, must protect against over-voltage, under-voltage and frequency deviation, and must include anti-islanding, so it disconnects during a grid outage and only reconnects when normal grid conditions return. You must install a switching or changeover panel that can isolate the system both automatically and manually, and the isolator has to be visible, lockable in the open position, and accessible to the DisCo’s staff at any time without prior clearance (s17(3)). There is also a quiet but important rule: the DisCo’s supply neutral and your renewable system’s neutral must be kept separate, not paralleled or intermingled (s17(10) to s17(11)). That one sentence can mean real rewiring work on an existing commercial building, so get an electrical review early.

On metering (s18): once you pay the connection charge, the DisCo provides a revenue-grade import and export meter (or dual-register smart meter) with time-of-use capability, installed to the Metering Code (now in its 3rd edition, March 2026). It separately records what you import and what you export, and the DisCo is responsible for reading, validating and reconciling the data for settlement. As noted above, no time-of-use meter at commissioning means off-peak settlement until one is fitted (s18(5)).

The process and the timeline

The journey from application to export is laid out across Chapter II and summarised in the process flow in Schedule 9. In order:

Two things dominate your planning.

First, the 30-versus-120-day fork for connection works (s12). If your site needs major reinforcement at 11kV or 33kV, the build can take four months on its own, and the whole process can run past six months end to end. A clean site with no reinforcement can be done in three to four months.

Second, the regulation does not tell you what the connection charge is. Schedule 4 lists the system-size bands and then leaves the naira amounts blank, to be set by each DisCo under a NERC-approved methodology. So a real cost line and the critical-path duration are both unknown until your feasibility study comes back. Do not promise a client a number you do not have yet.

Who gets the export credit: the contract question

This is the part that matters most for anyone doing a third-party deal, and it is easy to overlook.

The export credit is applied to the electricity bill of whoever holds the DisCo account, which is almost always the host or occupier (the regulation’s “Prosumer”). In a self-owned system that is fine: you own the panels, you hold the account, you keep the credit. But in a third-party-owned structure (PPA, lease, BOO, BOOT) the developer or SPV owns the asset while the host is the account holder. The DisCo credit lands on the host’s bill, not automatically with the asset owner. If your PPA or lease does not say how that credit is captured, shared or passed through, the value can simply leak to the customer.

There is a deeper legal wrinkle. The template Net Billing Agreement in Schedule 3 (and its eligibility clauses 1.1 and 1.2) reads as if the Prosumer owns the system and consumes on the same premises. That raises a genuine question for third-party ownership: can a host be the “Prosumer” while an SPV owns the asset, and can the export credits be assigned or shared? Until that is confirmed with counsel and the DisCo, do not put export credits into base-case debt sizing. Remember too that credits are tied to the premises and transfer with the agreement (s22(2)), and are zeroed if the system is relocated (s21(3)), so they do not travel with the equipment in a portable asset-finance model.

The honest bit: where the regulation is unclear

I write under a banner that promises not to pretend, so here are the rough edges. For a document this important, knowing them is part of using it well.

1. The body and the process flow disagree on timelines. The Technical Feasibility Report is “15 days” in the text (s8(1)) but “10 days” in Schedule 9. The NEMSA certificate is “5 days” in the text (s14(3)) but “7 days” in Schedule 9. Let us assume clerical error.

2. The peak-tariff clause is grammatically broken (s19(7)). The evident intent is that the peak export tariff is the same across all tariff bands and is not subject to the off-peak retail cap, but it should not be relied on without confirmation.

3. There are two different “120%” tests that sit in different places and will confuse applicants: a capacity test (export capacity within 120% of your Eligible Load Demand, in kW, at s6(4) and again s18(9)) and an energy test (if projected annual generation is more than 120% of historical annual consumption, in kWh, NERC may impose export-limitation controls, at s18(10)). They are reconcilable, but tangled.

4. The tariff is only fixed for twelve months. The settlement parameters, including the Export Tariff Factor and the underlying GC, TC and TLF, are locked for just twelve months from your connection date, then reviewable (s23(3)), and sooner if avoided cost moves by more than 20% (s23(1)). Section 28 protects rights already accrued, but the export tariff itself is a “settlement parameter” subject to review, so it is not guaranteed for the 20-year life of your asset. Anyone financing a project on the strength of export revenue is carrying real tariff risk. It is yet another reason to keep export out of your base case.

A note for the engineers: the feeder limit is pegged to average load (s6(3)), not minimum daytime load. Reverse-flow risk is actually highest when load is lowest and the sun is highest, so pegging to average arguably overstates the true hosting capacity. Do not be surprised if this metric gets revisited, possibly tightened.

So what should you actually do?

If you are a homeowner or a small business thinking about solar: this regulation is good news, but adjust your expectations. The win is cutting your bill by powering yourself during the day. The export credit is a modest top-up for spillover, worth roughly a third of what you pay to import. Size the system to your daytime use, not to “selling to the grid.”

If you are a commercial or industrial energy buyer: net billing is now a sanctioned third option alongside behind-the-meter self-consumption and wheeled PPAs, but only for sites at or below 1.5 MWp. Lead with self-consumption value. Treat net billing as a way to give residual monetary value to daytime spill and to right-size up to 120% of demand with a credit backstop. On batteries, anchor the decision on the full value stack, not the peak export premium.

If you are a developer or installer: oversize the array (DC), clip the export (AC), maximise self-consumption, and lodge your feeder application early before the 30% headroom fills. Get the feasibility study done before you quote a connection charge, and settle the export-credit question in the contract before anyone signs.

If you are a DisCo: the regulation hands you a register, a quarterly reporting duty, and a prescribed monthly credit ledger (s20, s21(5), s25). It is, frankly, a software problem waiting to be solved: application intake, feasibility tracking, NERC registration, NEMSA scheduling, connection-charge billing, and monthly settlement, repeated across thousands of prosumers, with a metering base that is still only about half complete. Reach out to me, we have solved this.

Before you promise anything: a checklist

If you are scoping a real project, gather these before you put numbers in front of a client or a credit committee:

• Bills and demand: at least 12 months of electricity bills and maximum-demand history.

• Load profile: daytime, weekend and holiday patterns, so you can estimate self-consumption honestly.

• Site control: Certificate of Occupancy, lease, or tenancy agreement (proof of ownership or occupation).

• Technical design: single-line diagram, inverter and protection specs, isolation and export-control scheme, certified by a COREN-registered engineer.

• Grid data: DisCo, feeder, transformer, voltage level, feeder average load, and any existing net-billing export on the same asset (for the 30% test).

• Solar study: a yield estimate (PVSyst, Helioscope or similar) with self-consumption and export split.

• Storage, if any: usable capacity and whether it meets the s17(6) basis for the peak premium.

• Metering: current meter type, and the need for a TOU-capable revenue-grade net meter.

• Commercial allocation: who holds the DisCo account, who receives the export credit, and how the PPA or lease treats it.

• Approvals path: DisCo feasibility, Net Billing Agreement, NERC registration, NEMSA certificate, DisCo commissioning.

The bigger picture

Step back, and the export-tariff design tells you what NERC is really optimising for. By paying you a discount to wholesale avoided cost, capping your bill at zero, and refusing to pay cash, the Commission has built a scheme that delivers demand-side relief without creating a new class of merchant generators or handing strained DisCos an open-ended payment liability. Against a grid averaging around 5,400 MW and losing close to 35% of its energy to technical, commercial and collection losses, that caution is rational. The scheme eases pressure by letting big consumers shave their own load and put their spill to modest use. What it does not do, and does not try to do, is fix the structural shortage of generation and the foundry-level reliability that Nigerian industry actually needs. That is a different, larger problem. Net billing is a sensible, conservative tool for the demand side. It is not the answer to the supply side, and it was never meant to be.

Used for what it is, this regulation is a genuine step forward: clear rules, defined timelines, real protections, and a workable path to connect distributed solar. Mistaken for what it is not, a way to get rich selling power to the grid, it will disappoint. The number to remember is the one we started with. Roughly 92% of the value is in using your own power. About 8% is in selling it. Build accordingly.

Free Excel Calculator

The companion NERC Net Billing 2026 Calculator below turns everything above into a model you can use. Open the Inputs sheet, edit the blue cells, and replace the illustrative tariff numbers with your DisCo’s MYTO values. Then read your result across the tabs: an Eligibility check against the size, 120% and 30% limits; a Monthly settlement with the carry-forward and anniversary logic; a Summary with the self-consumption-versus-export split; a Scenarios comparison (no solar, self-consumption only, off-peak, BESS peak, and a low-daytime-load case); and a Sensitivity table showing how savings move with the retail tariff and the export share. The yellow cells flag exactly which inputs you must confirm before trusting the output.

Nerc Net Billing 2026 Calculator V1

38.4KB ∙ XLSX file

Download

Download

Sources and references

The regulation

• NERC, Net Billing Regulations 2026, Regulation No. NERC-R-002-2026, signed by the Chairman and dated May 2026. Section references (s4 to s28, Schedules 1 to 9) are to this instrument throughout.

• Electricity Act 2023, section 226 (the enabling power).

Other NERC instruments and reports

• NERC Order No. NERC/2026/026, Regional Transmission Loss Factor Reporting (dated 8 April 2026): national average TLF of 7.24% in 2025 (down from 8.71% in 2024), against the 7% MYTO benchmark, with TCN directed to reach 6.5% by December 2026. Reported by The Guardian, ThisDay and The Punch, April 2026.

• NERC Q4 2025 Industry Report: ATC&C losses of 34.90%, collection efficiency of 79.36%, billing efficiency of 82.03%, and billing losses of about NGN 174.12bn. Reported by Legit.ng, April 2026.

• NERC Q3 2025 Report: metering coverage of about 55%, with roughly 5.36 million customers unmetered. Reported via AllAfrica, January 2026.

• Metering Code for the Nigerian Electricity Supply Industry, 3rd edition (March 2026): basis for the time-of-use net-metering requirement.

Tariffs and MYTO data (used for the illustrative figures; verify your own on your DisCo’s current MYTO at nerc.gov.ng and on your bill)

• Band A retail tariff around NGN 209/kWh and the NERC cap of NGN 225/kWh: Nairametrics (February 2025); Proshare (April 2024). Ikeja Electric’s reduction of its Band A rate to NGN 206.80/kWh: Channels Television / Ikeja Electric notice (May 2024).

• Commercial and industrial (maximum-demand) Band A tariffs run higher than residential (distribution-company tariff applications, 2024 and 2025).

A note on method The export-tariff figures (ACD of NGN 127.21, off-peak NGN 69.97, peak NGN 95.41) and the worked example are calculated using the regulation’s own formula (s19(4) to s19(5)). The Transmission Loss Factor input uses the 2025 national average of 7.24% from NERC Order NERC/2026/026; the Generation Cost and Transmission cost remain illustrative MYTO placeholders, because the official, DisCo-specific values change with macro conditions. They are clearly labelled and are fully editable in the companion calculator. Where the regulation is silent (for example, the connection-charge amount in Schedule 4), this article says so rather than guessing.

Disclaimer

This article and the companion calculator are for educational and analytical purposes only. They are not financial, legal, engineering, tax, regulatory, investment, or procurement advice, and they should not be treated as a substitute for advice from qualified professionals.

I have tried to interpret the NERC Net Billing Regulations 2026 carefully and to make the assumptions in the calculator transparent, but electricity regulation, MYTO parameters, tariffs, DisCo procedures, metering requirements, connection charges, and technical standards can change. The numbers used in the worked examples are illustrative unless expressly stated otherwise. Before committing capital, signing a contract, sizing a system, promising savings to a client, or submitting an application, confirm the binding figures and procedures directly with your Distribution Licensee, NERC, NEMSA, your current MYTO order, and your own legal, commercial, and engineering advisers.

The Excel calculator is a simplified decision-support tool. It is designed to help you understand the logic of the regulation and test scenarios, not to produce a bankable financial model, final engineering design, tariff approval, interconnection approval, or investment recommendation. You are responsible for checking all inputs, assumptions, formulas, and outputs before relying on them.

Where the regulation is unclear, internally inconsistent, or silent, I have said so and offered my best reading. That reading may be wrong, incomplete, or overtaken by later regulatory guidance, DisCo practice, NERC orders, or legal interpretation. Use the article and model as a starting point for better questions, not as the final answer.

It is not because I had nothing to say. I had several drafts. Some half-angry, some half-useful, some better left in the drafts folder because not every thought deserves daylight. But the honest reason is much dumber and far less intellectual.

Arsenal.

Subscribe now

I apologise in advance if this is not what you expect on a blog that pretends to be about clean energy, infrastructure, and the general comedy of trying to sound serious while figuring life out in real time. But Arsenal is an undeniable part of me. It has been since before I knew what a passion was. Before I could explain a balance sheet, a levelised cost of energy, or why “just add batteries” is not a project development strategy, I knew Arsenal. I have been with this club through the lows, the almosts, the banter years, the “next season” sermons, the painful hope, the rebuilds, the false dawns, the late winners, the collapses, and the strange emotional violence of believing again.

So the last few weeks have been consumed by Arsenal. I have a defence. A very good one.

Ten days ago, after twenty-two years, Arsenal were confirmed Premier League champions for the first time since the Invincibles in 2004. Not the FA Cup. Not a Community Shield argument on Twitter. The league. The thing. The one I had waited for since childhood. And then last night happened. We reached our first Champions League final since 2006, took the lead through Kai Havertz in the sixth minute, and lost it to Paris Saint-Germain on penalties after Eze and Gabriel both missed from the spot. PSG won the coin toss to take the shootout in front of their own supporters, and you could feel the air change. Twenty years of waiting and a season of brilliance, and at the very end it comes down to which way a coin lands and whether two penalties go in.

I am still sitting with that.

But this post is not about Arsenal.

The point is the other thing, the one percent of my brain that kept ticking away underneath while the other ninety-nine percent calculated Champions League final permutations like a deranged actuary.

That one percent has been chewing on a single idea:

Thinking and communication may become the most valuable skillset of this century.

Thinking. Communication. And the ability to fuse both with real knowledge of a field that matters.

That sounds obvious until you sit with it. Because we are living through a moment where intelligence itself is being industrialised, and somewhere inside that, a seventeen-year-old, a university fresher, a junior engineer, and a parent trying to advise their child are all asking the most reasonable question in the world. So what should I actually learn?

I have been asked some version of this more times than I can count, and I want to be honest about something. The more confidently I answered, the less I believed myself.

Let me build the stage first.

The biggest pile of money in the history of money

We are living through the largest capital mobilisation of our lifetimes, possibly of any lifetime. I do not say that for effect. The numbers are genuinely hard to hold in your head.

Three days ago, Anthropic closed a round at a 965 billion dollar valuation, eclipsing OpenAI and arriving, almost theatrically, at the doorstep of a trillion dollars. Let us not insult mathematics by pretending 965 billion is far from a trillion. This is a company that did not exist six years ago, founded by people who walked out of OpenAI because they thought it was moving too fast. For perspective, it took Alphabet roughly sixteen years from its 2004 IPO to cross a one trillion dollar market cap, which it finally did in January 2020. One is a private valuation and one is a public market cap, so it is not a clean comparison, but the direction of travel is the whole story.

Anthropic is approaching that neighbourhood in five years.

Five.

I have cooked rice longer than some AI companies have existed.

Zoom out from one company to the system. According to Stanford’s 2026 AI Index, United States private AI investment reached 285.9 billion dollars in 2025, more than twenty-three times China’s tracked figure, while global corporate AI investment hit 581.7 billion dollars, up roughly 130 percent in a single year. Then look at the concrete and copper. In 2026, the big five hyperscalers are on track to spend around 725 billion dollars on AI infrastructure, a figure that exceeds the entire annual GDP of Switzerland. And here is the stat that should reframe how you see all of it: in the first half of 2025, AI-related capital spending contributed more to United States GDP growth than all consumer spending combined. More than every shopper, every car, every dinner out, in the largest consumer economy on earth.

This is a capital war. Larry Page reportedly said he is willing to go bankrupt rather than lose this race. When founders talk like that, you are no longer in a tech cycle. You are in something closer to a national mobilisation.

And the ceiling people are imagining is absurd. Elon Musk’s SpaceX, having absorbed his AI company xAI, is targeting around 1.75 trillion dollars in an IPO that would be the largest in history. The investor Ron Baron, whose firm holds a roughly 15 billion dollar stake, went on CNBC and said SpaceX could be worth “10 trillion, 20 trillion, 30 trillion” over the next ten to fifteen years, and that he “could be very low.” His reasoning is the tell. He thinks the future is data centres in orbit, because, in his words, you get “free electricity and free cooling once you get into space.”

Read that last part again, because it is not a throwaway. The most bullish space investor on the planet is dreaming about leaving the planet, and the reason he gives is energy. Even the people imagining a thirty trillion dollar company cannot escape the constraint.

This is the world we are in. A five-year-old AI company valued like a nation. A data-centre build-out that commands more capital than the GDP of most countries. A model update that moves labour markets.

So, exciting. More money than ever, moving faster than ever, into the most powerful technology any of us has touched.

Here is the problem.

The money is not the jobs

The thing about all that capital is where it goes. Servers, data centres, networks, power systems, cooling. These assets are extraordinarily capital intensive. They are not labour intensive the way building a railway or a factory used to be. You do not need a hundred thousand people to run a hyperscale facility. You need a few hundred, and a lot of electrons.

So the biggest investment boom in history is happening at the exact same moment as one of the strangest layoff cycles in history. In the first five months of 2026 alone, tech layoffs hit 142,000, putting the year on pace to approach 370,000, close to the post-pandemic record. And the part that makes your skin crawl is why. These are not struggling companies. Oracle cut around 30,000 roles shortly after strong earnings. Meta cut 8,000 people and closed 6,000 open roles specifically to free up budget for AI infrastructure. By March, AI was the single most cited reason for job cuts in the United States, accounting for a quarter of them. Profitable companies are firing people to pay for the machines that will let them need fewer people.

That is the loop. And it is real, and it is not slowing down.

So what do you actually tell a seventeen-year-old?

Here is where my easy answers fall apart in my hands.

For years my advice was simple and, I thought, bulletproof. Study engineering. Study mathematics. Study physics, economics, computer science. Anything that forces you to solve problems and builds your thinking muscles. Pick a hard technical discipline and the world will always have a use for you.

I still believe most of that. But the bottom of the ladder, the rungs I climbed, are the ones being sawn off first.

Dario Amodei, who runs the company whose product I am literally using to research this essay, spent last year warning that AI could wipe out half of all entry-level white-collar jobs and push unemployment to between 10 and 20 percent within one to five years, with finance, law, consulting and tech most exposed. This is not an outsider throwing stones. This is the man building the thing, telling you what it does. The IMF’s Kristalina Georgieva put it even more bluntly at Davos, calling AI’s impact “a tsunami hitting the labour market,” with about 60 percent of jobs in advanced economies and 40 percent globally set to be enhanced, transformed or eliminated.

And the early data is not arguing with any of them. Stanford’s AI Index found that employment for software developers aged 22 to 25 has fallen nearly 20 percent since 2024, even as their older colleagues’ headcount grew, with one in three surveyed organisations expecting to shrink their workforce over the coming year. The junior analyst job, the first-year associate job, the entry-level modelling job, the roles that were the training ground for everyone who is now senior, those are the ones being compressed hardest.

Which leaves a genuinely brutal question hanging in the air.

Nobody becomes a senior project finance professional by reading “Project Finance for Dummies” and asking a chatbot to explain DSCR. You become good by building messy models, getting yelled at politely by investment committees, realising your assumptions make no sense, fixing them, listening on calls, watching seniors negotiate, writing bad memos, then less bad memos, then eventually something a client can read without losing faith in humanity.

Nobody becomes a good energy engineer only by passing exams. You become good by visiting sites, smelling diesel fumes, watching a roof you modelled beautifully turn out to be structurally suspicious, and learning that commissioning is where Excel optimism goes to meet reality.

Entry-level work is not just labour. It is apprenticeship. If AI deletes the apprenticeship layer faster than we build a new one, we do not just lose jobs. We lose the pipeline through which judgment is formed.

That is the hidden crisis. Not unemployment alone. The quiet collapse of skill formation. I do not have a clean answer to it. Nobody does. Anyone who tells you they do is selling a course.

So I sat with it. And the more I sat with it, the more my instinct flipped into something that, on the surface, sounds like heresy.

The heresy: every serious field becomes more important, not less

Hear me out before you close the tab.

The standard story is that AI eats jobs one field at a time, and one by one they shrink and vanish. I think the truth is closer to the opposite, and oddly, the cleanest version of the argument comes from the same Dario Amodei who scared everyone last year. A few weeks ago, sitting on a stage next to Jamie Dimon, he reframed the whole thing around the Jevons Paradox. His line was roughly this: if you automate ninety percent of a job, then everyone does the remaining ten percent, and that ten percent expands to become a hundred percent of what people do, and it makes them ten times more productive.

Read that again, because it is the whole essay.

The machine takes the commoditised part. The spreadsheet mechanics, the first draft, the boilerplate contract, the irradiance forecast, the literature review. What is left, the residue, is the part that was always the actual job and that we never had time to do properly. The judgment. The taste. The trust-building. The translation between a technical truth and a human decision. The deciding what is even worth doing. That residue does not shrink. It expands to fill the time the machine just freed up. The ten percent becomes everything.

So the question is not which field is safe. Nothing is safe. The better question is: in this field, what is the highest-value human judgment that survives after automation?

That question changes everything.

AI will not make medicine irrelevant. It will make basic diagnostic support cheaper, and make clinical judgment, patient trust, ethics and complex care coordination more valuable. It will not make law irrelevant. It will draft the first contract, and make negotiation, strategy, advocacy and knowing when technically correct language creates a commercially stupid outcome more valuable. It will not make finance irrelevant. It will build the spreadsheet faster, and make capital judgment, risk sense, structuring and storytelling more valuable. It will not make engineering irrelevant. It will run the routine calculation, and make systems thinking, safety, constructability, field judgment and stakeholder management more valuable. It will not make teaching irrelevant. It can explain calculus at 2am without judging you, and that makes motivation, mentorship and character formation more valuable, not less.

And it will not make communication irrelevant. It will flood the world with synthetic words, and make genuine clarity rarer and more precious than it has ever been.

Now look at what the data says about which skills are rising, because this is not my vibe, it is the most boring, respectable source in the room. The World Economic Forum’s Future of Jobs Report puts analytical thinking as the single most sought-after core skill on earth, with seven in ten companies calling it essential, followed by resilience, leadership and creative thinking. Yes, AI and big data top the fastest-growing list. But the report is blunt that human skills, analytical thinking, communication, collaboration, will remain critical, and that the most valuable profile combines both. The thing employers say they cannot find is not someone who can use the tool. It is someone who can use the tool, think clearly, and explain it to another human being.

And the net picture is not the apocalypse. The same report projects 170 million new jobs created and 92 million displaced by 2030, a net gain of about 78 million. Both the terror and the optimism are true at once, which is exactly why it feels so disorienting.

Look at the median, not the mean. And look at the map

There is a trap in this whole conversation, and Silicon Valley falls into it constantly. We talk about “jobs” as if they all live in San Francisco, pay 300,000 dollars, and involve writing code. We obsess over the top of the salary distribution and the most automatable knowledge work, then generalise from it to the entire human species.

That is looking at the mean. You should look at the median. And you should look at the map.

When the WEF actually lists the fastest-growing jobs by raw numbers, the top of the list is farmworkers, delivery drivers, software developers, construction workers and shop assistants. Care work, nursing and teaching are all growing fast on the back of ageing populations. The green transition alone is expected to create 34 million additional jobs by 2030. You cannot run any of these from a data centre in Virginia.

Spread it across geographies and the case gets stronger, not weaker. The messier the market, the larger the human residue, and the longer it stays unautomatable. AI performs best where processes are standardised, data is clean, rules are clear, and feedback loops are fast. African infrastructure markets often offer the exact opposite: patchy data, fragmented regulation, constrained grids, currency volatility, land complexity, informal decision pathways, and deep trust gaps between developers, offtakers, lenders, governments and communities. That is frustrating. It is also a moat. In frontier markets, value is not just knowing the answer. Value is knowing why the official answer will not work.

The strange return of communication

Here is the part people underestimate.

AI is making writing cheaper. It is not making communication easier. Those are different things.

We are about to drown in grammatically correct nonsense. Every company will have newsletters. Every founder will have thought leadership. Every analyst will submit polished memos. Every student will produce essays with suspiciously balanced paragraphs and exactly three recommendations.

The average sentence will improve. The average meaning may collapse.

This is precisely why communication becomes more valuable, not less. When everyone can generate words, the scarce skill becomes knowing what must be said, what must not be said, what matters, what is noise, what the audience actually fears, and how to move a person from confusion to clarity. Communication is not grammar. Communication is judgment under social conditions. It is translation between worlds. Technical to commercial. Commercial to legal. Engineer to client. Boardroom to site. Local reality to global capital.

AI can summarise a meeting. It cannot always tell you that the client’s silence after slide fourteen meant the FX escalation clause just killed the deal. It can write a proposal. It cannot sit across from a CFO in Accra, Lagos, Abidjan or Nairobi and feel the room shift when currency risk enters the conversation. It can generate a schedule. It cannot understand that “we will revert shortly” means seven different things depending on who said it, in what tone, on what day, and whether procurement was copied.

That skill is going to be priceless.

Energy is not safe. It is necessary

I do not want to romanticise my sector. Energy is not immune. A great deal of what junior analysts do today will be compressed. Solar yield estimates, first-pass financial models, tariff sensitivities, PPA summaries, market scans, policy briefs, due diligence checklists, even parts of technical review. AI will touch all of it. Some tasks will disappear. Some teams will shrink. Some mediocre work will no longer justify a salary. That is uncomfortable, and it is true.

But energy has one enormous advantage. It lives in the physical world.

You cannot prompt a transmission line into existence. You cannot fine-tune a substation. You cannot run a factory on a slide deck. You cannot hallucinate electrons into a smelter, a data centre, a hospital, a brewery, a cement plant or a cold room.

And the demand is structural. Every single thing AI does runs on electricity, and the industry is desperately, structurally short of it. The IEA projects that global data-centre electricity consumption will more than double to around 945 terawatt-hours by 2030, slightly more than Japan’s entire consumption today, with AI as the most important driver. In the United States, data centres are on course to account for nearly half of all electricity demand growth between now and 2030, consuming more power for processing data than for manufacturing aluminium, steel, cement and every other energy-intensive good combined. The most valuable companies in the world, sitting on the most capital in history, are bottlenecked by the thing my industry builds.

Energy is the binding constraint on the entire intelligence economy. That is the small, enormous win for my sector.

And the human residue here is huge. The IEA’s latest count puts global energy employment at 76 million people in 2024, growing nearly twice as fast as the wider economy, with the electricity sector now the single largest employer in energy for the first time, overtaking fuel supply, and solar PV the principal driver. Renewable energy employment specifically reached 16.6 million in 2024. And here is the kicker for a young African engineer: more than half of the energy companies the IEA surveyed reported critical hiring shortages that are already delaying projects, with the worst gaps in applied technical roles. The sector is not short of demand. It is short of people. And of those millions of jobs, Africa held only about 324,000, on a continent where nearly 600 million people still have no electricity and Mission 300 is racing to connect 300 million by 2030.

The gap between the work that must be done and the people who can do it is the single best place to plant a career I can think of.

This is the paradox I keep coming back to. AI may reduce labour demand in parts of the economy, but it increases electricity demand in the physical one. And electricity demand creates real work. Not just coding work. Grid planners. Protection engineers. Solar installers. Battery engineers. SCADA specialists. Cooling engineers. Project financiers. Permitting experts. Community engagement leads. O&M technicians. Energy lawyers. People who can stand between capital, technology, regulation and dust, and somehow produce a working asset.

That last sentence is basically the job.

A practical guide for the young engineer who wants to come into energy

I promised something more useful than a TED talk. So here is the guide I wish someone had handed me earlier. Ten things.

1. Build a real engineering base, then refuse to stop there. Do the electrical, mechanical or chemical degree. Understand the difference between kW and kWh so deeply that it irritates you when people misuse them. Learn AC and DC, three-phase power, transformers, inverters, switchgear, protection, harmonics, power factor, fault levels and load profiles. Learn solar PV properly: irradiation, degradation, clipping, losses, yield, performance ratio. Learn batteries: C-rate, depth of discharge, round-trip efficiency, augmentation, why the economics are always messier than the LinkedIn posts. Learn thermal energy too, because industry runs on heat, not just electricity. And learn grids, because transmission and distribution are becoming the real bottleneck. AI can help you learn faster. It cannot give you intuition unless you wrestle with the fundamentals yourself.

2. Become dangerous with Excel, Python, and AI tools, in that order. I know you want to skip straight to AI. Do not. Excel is still the native language of infrastructure finance. Learn to build a clean model: inputs, assumptions, calculations, outputs, sensitivities, debt sizing, DSCR, IRR, NPV, tariff build-up, FX. Then learn Python for scale: data cleaning, simulation, optimisation, forecasting. Then layer AI on top as pure leverage. Use it like a brilliant intern who sometimes lies confidently, because that is basically what it is. Let it draft, summarise and check. Never let it own the judgment.

3. Learn project finance earlier than you think you need to. Energy projects do not happen because the technology is beautiful. They happen because someone can pay for them. Learn how money enters a project and how it leaves. Equity, debt, development capital, construction risk, offtaker risk, currency risk, security packages, guarantees, termination payments. Understand why a project with great engineering can be unbankable, and why a project with average technology but strong risk allocation closes. If you can speak engineering and finance, you become useful fast. Add law, and people start inviting you to meetings they probably should not.

4. Go to site as early as humanly possible. Do not become a PowerPoint engineer. The site will humble you. You will learn that the satellite image lied. That the roof has leaks nobody mentioned. That the “24/7 operation” has seasonal shutdowns. That the generator logs are handwritten by someone who has run that plant for seventeen years and knows it better than your model ever will. That “available land” can mean land that is available politically, legally, culturally or spiritually, and these are not the same thing. Wear the boots. Ask the stupid questions without arrogance. Touch reality.

5. Pick a technical-commercial niche before you go broad. “Interested in renewable energy” is not a skill. Choose a wedge. Commercial and industrial solar, where factories are paying three to five times too much for diesel and the economics are survival, not virtue. Or mini-grids and the off-grid frontier. Or storage. Or industrial heat. Or data-centre power strategy. Go deep enough to explain the technology, the economics, the risks, the business model, the regulatory barriers and the main players. Before depth, breadth is just vibes. After depth, breadth is power.

6. Learn to write like someone’s money depends on it, because it does. A confusing memo can delay an approval. A vague risk register can hide a fatal issue. A lazy proposal can make a serious company look unserious. Write clearly. Short sentences when needed. State your assumptions. Separate facts from estimates. Explain uncertainty instead of hiding it. Tell the reader what decision they need to make, and make it easy for a busy person to understand a complex thing without feeling stupid. That skill alone will carry you absurdly far.

7. Develop commercial empathy. Engineers often think clients buy the best technical solution. They do not. They buy the solution they understand, trust, can afford, can approve internally, can defend to their board, and can operate without embarrassment. A CFO sees risk differently from an engineer. A plant manager sees downtime differently from a sustainability officer. A lender sees every ambiguity as a future default. A community sees land, jobs and promises. Learn to translate between those realities. It is not manipulation. It is respect for other people’s constraints.

8. Understand AI’s energy demand specifically. If you want to be relevant in energy this decade, understand data centres. Uptime tiers. Redundancy: N, N+1, 2N. Power usage effectiveness. Cooling loads. Grid connection queues. Behind-the-meter generation. Hourly matching versus annual matching, and why “100 percent renewable” can mean very different things depending on the accounting. The data-centre client does not just want cheap power. They want reliable, scalable, clean, bankable, redundant power that supports absurd load growth without becoming a regulatory scandal. That is a career.

9. Study regulation like it is the deal, because it is. In energy, regulation is not paperwork. It is market architecture. Can you sell power directly to a customer? Can you wheel electricity across the grid? Can tariffs be denominated in dollars? Can a captive plant serve multiple offtakers? Can a distribution company block your project? A single rule can decide whether a business model exists. Engineers who understand regulation outperform engineers who only understand equipment.

10. Build public proof of thought. This is one reason I write, even when I am clearly unqualified and emotionally compromised by football. Writing forces thinking, and public writing creates proof. You do not need to go viral. Please do not make going viral your strategy; it is spiritually dangerous and usually embarrassing. But explain one concept a week. Break down a project. Review a policy. Share what a site visit taught you. Over time, people learn how you think, and in a world where everyone can claim skills, visible thinking becomes a credential.

The uncomfortable bit for schools

All of this means our schools may need to change faster than they want to. We keep training young people for a world where information scarcity is the main problem. Information is no longer scarce.

Attention is scarce. Judgment is scarce. Taste is scarce. Courage is scarce. Clarity is scarce. The ability to ask a good question is scarce. The ability to work with people across difference is scarce. The ability to sit with ambiguity without becoming useless is scarce.

So yes, learn mathematics, physics, engineering, economics, programming, writing and history. But more importantly, learn to think across them. The world’s hardest problems do not arrive neatly labelled. Climate change is not just environmental science. It is energy, finance, politics, justice, technology, land and time. Energy access is not just generation capacity. It is grids, tariffs, governance, affordability, fuel logistics, metering, currency and trust. The people who can connect the dots will matter. The people who can explain the dots will lead.

If I had to compress everything into one line, it would be this:

Domain depth, plus AI leverage, plus communication, plus judgment, plus trust, equals durable value.

Not perfectly future-proof. Nothing is. But resilient. Domain depth means you know something real. AI leverage means you move faster than the people refusing the tools. Communication means others can use your thinking. Judgment means you know when the model is wrong, when the client is afraid, and when the answer is technically correct but practically useless. And trust, the one that matters most, is earned through repeated usefulness, honesty and clarity under pressure. AI can generate information. It cannot automatically generate trust. The future will not belong to people who simply “know AI.” Everyone will know AI. It will belong to people who can combine AI with real-world competence and human trust.

The part where I do not wrap this up neatly

Maybe this is where Arsenal comes back in.

Football is a strange teacher because it is both irrational and brutally data-driven. You can dominate the expected goals and still lose. You can build for years and still watch one penalty fly over the bar. You can end a twenty-two year wait to be champions of England and still go to bed grieving a Champions League final eleven days later.

Progress and pain can arrive in the same week.

That feels like the labour market right now. AI is progress and AI is pain, both at once. It will create companies, tools, medicines and possibilities we cannot yet imagine. It will also delete roles, compress careers, expose shallow competence, and make some people feel like the ladder was pulled up just as they reached it. Both things are true. Anyone selling only optimism is lying. Anyone selling only doom is also lying. The honest place is harder. It says: learn the tools but do not worship them. Build technical skill but do not hide inside it. Study your field but do not become narrow. Communicate clearly but do not become performative. Go fast but touch reality.

I genuinely do not know how long “for now” lasts. I do not know what happens to GDP if the intelligence layer keeps thinning out the middle of the labour market faster than new work appears. I do not know whether the net 78 million jobs show up where the displaced 92 million used to be, or somewhere else entirely, leaving a generation stranded in the gap. I do not know if my own kids, hypothetical and otherwise, will one day read this essay the way I now read “just learn to code.”

But I will tell you the one thing I have stopped doubting. The people who navigate this, both the energy transition and the intelligence one, will not be the ones who found the safe field. There is no safe field. They will be the ones who went deep on something real, learned to think clearly and explain themselves, picked up the new tools without being seduced by them, and refused to look away from the hard questions even when the answers were not ready yet.

Twenty-two years we waited. Then we won, and then we lost on penalties after a coin toss, both inside eleven days, and the lesson was the same in the triumph and the heartbreak. You do the deep work for years. You control what you can control. You make your peace with the margins you cannot. Then you keep going.

Think. Build. Negotiate. Explain. Judge. Earn trust. Stand on site. Read the room. Connect the dots. Turn complexity into action.

That may be the last human advantage. And for now, it is still ours.

It is past five somewhere. Pick your hard thing. Do not go back to sleep.

I’m Kay. These are my honest, sleep-deprived, slightly heartbroken thoughts from someone who does not know enough but refuses to stop asking. I would love to hear yours.

Wrong about energy so you don’t have to be.

P.S. I researched this piece using google and AI tools, including the one that helped me check the layoff numbers, which is a bit like asking the arsonist to hold your fire extinguisher. The irony is never lost on me.

P.P.S. To the young engineers specifically: learn the fundamentals, learn the tools, learn to write, go to site, and please, for the love of all that is holy, know the difference between kW and kWh.

P.P.P.S. To the Arsenal fan who is also asking me for career advice, here is a free combined lesson. Build a squad with depth, master your set pieces, and accept that sometimes you do everything right and still lose to a German scoring in the sixth minute. Wait. He scored for us. I am very tired. Good night. Good morning.

Sources

• Arsenal Premier League champions, first title in 22 years: Premier League; ESPN

• Champions League final, PSG beat Arsenal on penalties: ESPN match report; Olympics.com

• Anthropic 965 billion dollar valuation, eclipsing OpenAI: Bloomberg; founding and growth: VentureBeat

• Alphabet ~16 years from 2004 IPO to a 1 trillion dollar market cap (Jan 2020): CNBC

• US private AI investment 285.9 billion dollars, global corporate AI investment 581.7 billion dollars (+130%): Stanford HAI 2026 AI Index

• 2026 hyperscaler AI capex ~725 billion dollars, exceeding Switzerland’s GDP: AL Capital Advisory

• AI capex outpacing all consumer spending in US GDP growth (H1 2025): KKR

• Larry Page “willing to go bankrupt”: IEEE ComSoc Technology Blog

• SpaceX targeting ~1.75 trillion dollar IPO; Ron Baron’s 10 to 30 trillion dollar projection and orbital data-centre reasoning: Invezz; Benzinga

• AI capex is capital-intensive, not labour-intensive: Real Investment Advice

• 142,000 tech layoffs in 2026, profitable companies cutting to fund AI: Tech Times; Oracle cuts: Tech Insider; Meta cuts: TechSpot

• AI as the leading cited reason for US job cuts: Trading Economics / Challenger, Gray & Christmas

• Amodei on entry-level white-collar jobs: Axios; Jevons Paradox reframing: Fortune

• IMF’s Georgieva on AI as a labour-market “tsunami” (60% advanced economies, 40% globally): Business Today

• Software developers aged 22 to 25 employment down nearly 20% since 2024; one-third of organisations expect workforce reductions: Stanford HAI 2026 AI Index, Economy chapter

• WEF Future of Jobs Report 2025, skills and net job projections: WEF press release; fastest-growing jobs and skills; digest

• Global data-centre electricity demand to ~945 TWh by 2030, nearly half of US electricity demand growth: IEA, Energy and AI; executive summary

• Global energy employment 76 million in 2024, electricity now the largest employer, critical hiring shortages: IEA, World Energy Employment 2025

• Renewable energy jobs 16.6 million in 2024: IRENA; Africa’s ~324,000 renewable jobs: IRENA

• ~600 million Africans without electricity and Mission 300 progress: Mission 300 / fundsforNGOs

For years, the difference is invisible. Going back and coming home look exactly the same. Same airport. Same suitcase. Same mother waiting at arrivals. Same heat pressing against your face before you even leave the terminal. Same food on the table when you get in. Same old bed. Same family shouting your name like you have only been gone for a weekend.

And because it feels familiar, you believe it.
You tell yourself, I am home.

Then one day, usually over something small, the lie breaks.
Someone mentions a name you should know.
You smile like you know it.
You do not know it.
Everyone else does.
That is when you realise home has been happening without you.

Subscribe now

The airport lies

The first few days back are always dangerous because they convince you nothing has changed.
You land at Murtala Muhammed at night. Lagos is loud before you even reach the car. The air is hot, wet, rude, alive. The immigration officer looks at your passport, looks at you, and says, “Welcome, sir.”
Something in your chest relaxes.

Outside, the city is exactly dramatic enough to make you believe in belonging again. Third Mainland Bridge. Oworonshoki. Brake lights. Petrol. Pepper. Grilled fish. Dust. Someone selling plantain chips between lanes. A billboard for some new bank, some new estate, some new version of Nigeria promising to fix itself by December.
You crack the window.
You breathe it in.
You think, yes.
This is me.

At home, your mother has cooked like you are returning from war. Jollof. Fried plantain. Stew. Pounded yam wrapped and waiting. More food than any reasonable person could eat, because mothers do not cook for appetite. They cook for absence.
She hugs you too long.
Your sister pretends not to cry.
Your cousins are loud. Someone is playing music. Someone has brought drinks. A generator is humming in the background like it, too, remembers you.

For a few days, you are restored.
You sleep in your old bed. You eat food you did not have to explain to anyone. You hear your name pronounced properly. You sit in rooms where nobody needs your origin story. You feel, for the first time in months, like a complete person.

This is the trap.
Because what you are feeling is not proof that you are home.
It is proof that you miss it.
Those are not the same thing.

The moment it breaks

It usually happens casually.
Your cousin says, “You remember Tunde now? He just moved his mother to Lekki.”
You nod.
You smile.
You say, “Ah, that’s big.”
You have absolutely no idea who Tunde is.

So you start searching your memory. School? Church? Family friend? Football? Someone’s younger brother? One of the boys from the street?
Nothing.

But your cousin remembers him. Your sister remembers him. Your mother remembers him. Everyone remembers him.
And suddenly you understand.

There is an entire version of home that continued without your permission.
People got married. People died. Babies were born. Friends became strangers. Strangers became family. Your parents built routines without you inside them. Your siblings became adults in ways you did not witness. The street changed. The jokes changed. The names changed.

Nobody pushed you out.
Time did.

That is the part that hurts.
There was no argument. No betrayal. No dramatic exile. No one said you no longer belonged.
You simply left.
And while you were building a life somewhere else, home built one too.

Achebe knew

Chinua Achebe wrote this wound before most of us knew we would inherit it.
In No Longer at Ease, Obi Okonkwo leaves Nigeria for England. He studies. He returns educated, polished, changed. On paper, he has done everything right. He is the success story. The son who left and came back with something to show for it.

But success does not save him.
Because the real problem is not that Obi went abroad.
The real problem is that when he returns, he no longer fits cleanly anywhere.

England does not belong to him.
Nigeria does not receive him unchanged.
He is too foreign for home and too African for abroad. Too local there, too altered here. Too much of both places. Fully at ease in neither.

Obi’s problem is not unique to colonial Lagos; it is the permanent condition of many Africans who leave and try to come back.
That is the oldest diaspora problem.
Not poverty.
Not distance.
Mismatch.

The strange feeling of becoming impressive in one world while becoming slightly unfamiliar in the one that raised you.
Most of us do not want to admit we are Obi.
But we are.

We left to become more useful, more powerful, more educated, more mobile, more free.
And then one day we came back and realised freedom had a cost.
The cost was ease.

The first thing leaving takes

The first thing leaving takes is not your accent.
It takes your timing.

You start missing things.
At first, small things. A birthday dinner. A Sunday lunch. A cousin’s engagement. A random evening where everyone ended up at your mother’s house and laughed about something you will only hear about later.

Then bigger things.
Weddings. Funerals. Births. Surgeries. Graduations. The kind of moments families use to remember who showed up.

You always have a reason.
The flight was expensive.
The client meeting moved.
The visa was delayed.
The project was closing.
The timing was impossible.

And most of the time, the reason is real.
That is what makes it worse.

You are not lying.
You are not careless.
You are not wicked.
You are simply absent.

And absence, even justified absence, still accumulates.

Then it takes the voice

After timing, it takes the accent.
Not all at once. Just a little. A vowel changes. A rhythm flattens. You start saying things the way people say them where you live now because repeating yourself became tiring.

Then one day an aunty laughs and says, “Ah-ah, you don turn oyinbo.”
Everyone laughs.
You laugh too.
But something has been marked.

Then it takes the language.
Not the whole language. Just enough to humiliate you privately. You pause in the middle of a sentence, reaching for a word you used to know without thinking. The word is there, but behind glass. You can see it. You cannot reach it fast enough.

That pause is the grief.
Everybody hears it.
Nobody says anything.

Then it takes the names.
The cousin you used to see every Christmas is now someone you recognise from Instagram. Your uncle’s new wife has a name you have asked for twice. The baby born two years ago is still “the new baby” in your head. Your friend got married and you watched the reception from stories. Someone from the old street died and you found out three days later in a family group chat.

This is how leaving works.
It does not steal your home in one dramatic moment.
It makes you slightly late to everything until lateness becomes your role in the family.

WhatsApp is not presence

This is where the diaspora lies to itself.
We think because we call, we are there.
We think because we send money, we are there.
We think because we reply in the group chat, comment under the pictures, send the birthday voice note, and wire something for the funeral, we are there.

We are not.
We are involved.
That is different.

Money is not presence.
A video call is not a chair at the table.
A voice note is not your hand on your mother’s shoulder.
A bank transfer is not the same thing as standing beside your brother at the burial.

These things matter. Of course they matter. Sometimes they are all we can do. Sometimes they are love doing its best across distance.
But they are not the same.

And some part of your family knows they are not the same.
They may never say it.
Because they love you.
Because they are proud of you.
Because they know you left for a reason.
Because they do not want to make your life harder.

But they know.
And so do you.

The actual bill

The bill is not dramatic.
That is why it is easy to ignore.

You missed your sister’s thirtieth because of a site visit in Abidjan. She said it was fine. You told yourself it was fine.
It was not fine.

You watched your mother recover through a phone screen because your flight was the next morning. Your aunty held up the phone. Your mother smiled. You smiled back. You told her you loved her.
Then you hung up and went to the meeting.
The meeting went well.
That is the part that makes it cruel.

Your father called at ten in the morning on a Tuesday. You were in a pitch. You texted, “In a meeting, call you later.”
You called back four hours later.
He said it was nothing.
It was nothing.

But he is older now. And you were in a pitch. And the next pitch is always coming. And one day the call will not be nothing.

This is the bill.
Not one great tragedy.
A slow leak.

Birthdays you joined by FaceTime. Weddings you watched through someone else’s shaky phone. Funerals you sent money to. Children who know you first as a voice message. Parents whose ageing you witness in pixels.
The diaspora does not only separate you from a place.
It teaches the people you love how to need you less.
They are adjusting to your reality the only way they know how: by learning to hold their crises, their joys, their boredom, without expecting you in the room.

The version of you that stayed

There is another person in this story.
The version of you who never left.

You think about him sometimes.
He speaks the language better than you. He knows the streets better than you. He knows the current slang. He knows which road to take when traffic misbehaves. He can enter a market without being priced like someone who just landed. He does not have to perform localness because he never stopped being local.

He is not better than you.
But he belongs with less effort.
And effort is the thing you notice now.

You notice yourself trying.
Trying to sound normal.
Trying to remember names.
Trying to follow jokes.
Trying not to ask stupid questions.
Trying not to look too impressed by things everyone else considers ordinary.
Trying not to become the person who says, “Back in London…”
Trying not to be foreign in your own house.

That is when you understand what leaving has done.
It has not made you an outsider.
That would be simpler.
It has made you almost an insider.
Almost is worse.

The trade was real

Now, be honest.
Leaving gave you things.

It gave you rooms you might never have entered. It gave you language for power. It gave you discipline. It gave you perspective. It gave you a way to see your country clearly, sometimes more clearly than you could when you were inside it.

It gave you access.
Capital.
Confidence.
Range.

It taught you how systems work elsewhere. It taught you what is possible. It taught you what home deserves. It may even give you the tools to build something back home one day.

So this is not an argument against leaving.
Leaving may have saved you.
Leaving may have fed your family.
Leaving may have made the future possible.

But we have to stop pretending it was free.
You gained things that mattered.
You lost things that mattered.
Both are true.

The maturity is not choosing one truth and deleting the other.
The maturity is carrying both without lying.

How to know if home has become a place you visit

You cannot answer this by feeling bad for five minutes.
Guilt is lazy.
You need evidence.

So write down the last five major family events.
Weddings. Funerals. Births. Surgeries. Graduations. Milestone birthdays. The moments people remember.

Now write what actually happened.
Not what you planned.
Not what almost happened.
Not what would have happened if the ticket was cheaper or the meeting moved or the visa came through or the timing was better.
What actually happened.

How many did you attend in person?
How many did you join by phone?
How many did you miss and apologise for later?

Do not round up.
Do not give yourself credit for intentions.
Intentions do not appear in family photos.

If the number is low, it does not mean you are a bad son. It does not mean you are a bad daughter. It does not mean you do not love your people.
It means your life has a cost.
And the cost has a name.
Absence.

The question is not whether you are paying it.
You are already paying it.
The question is whether you are paying it on purpose.

Because sacrifice and drift can look very similar from the outside.
Same airport pictures.
Same calendar.
Same “just landed” posts.
Same career updates.

But sacrifice is chosen.
Drift just happens.
One is a life.
The other is a current.

Home is not a memory

Here is the thing Achebe understood.
Home is not something you own forever because you were born there.

Home is not a certificate.
Home is not a passport.
Home is not your mother’s address.
Home is not the fact that you still know the chorus to an old song or can name the street you grew up on.

Home is a relationship.
And relationships weaken when they are not tended.

The country you left will not wait for you. Not because it is cruel. Because it is alive.

Lagos will keep moving.
Accra will keep moving.
Nairobi will keep moving.
Abuja, Kampala, Kigali, Johannesburg, Freetown, Dakar.

They will all keep moving.

The streets will change. The people will change. The jokes will change. Your parents will change. Your siblings will change. You will change.

And one day you will return to a place that still loves you but no longer needs you in quite the same way.
That is not betrayal.
That is time.

You can go back.
You can visit.
You can invest.
You can build.
You can even move back.
But you cannot unleave.

No longer at ease

Obi Okonkwo did not know how to be at ease in either of his worlds.
Achebe called that a tragedy.
We call it a career.

Maybe we are wrong.
Maybe success abroad is not the problem. Maybe distance is not even the problem. Maybe the problem is pretending nothing has changed when everything has.

We left home to become something.
Some of us became it.
But while we were becoming, home became something too.

And now the work is not simply to return.
The work is to rebuild ease.

Call before there is bad news.
Visit before there is a burial.
Learn the names before the children become adults.
Ask the boring questions before they become impossible.

Send the money, yes.
But do not confuse money with yourself.
Your people do not only need what you earn.
They need the person earning it.

The diaspora gives.
The diaspora takes.
The least we can do is stop acting surprised when the bill arrives.

So go home

Go home.
Not forever.
That is a different conversation.

Go home more deliberately.
Go home before guilt is the only thing strong enough to move you.
Go home when nobody is getting married.
Go home when nobody is sick.
Go home when there is no funeral, no emergency, no December performance, no big event to optimise around.

Go home for ordinary days.
Because ordinary days are where belonging lives.

Sit in your mother’s kitchen. Let her tell you stories about people whose names you have forgotten. Ask your father the questions you keep postponing. Learn the children’s names. Let your language come back badly before it comes back well. Stay long enough for the visit to stop feeling like a visit.

Go home for a week, not a weekend.
Go home and be bored.
Boredom is part of belonging.

You cannot maintain home only through crisis.
You cannot keep a relationship alive only by appearing for the dramatic scenes.
You have to show up for the quiet ones too.

Talk soon,

- Kaykluz

P.S. I wrote most of this on a flight. The fact that it took a flight to write about not flying home enough is either deeply ironic or exactly what I deserve.

P.P.S. My mother is fine. She asked me to tell you that. She also wants me to remind you to call your own.

P.P.P.S. If you are reading this on a layover, close the laptop. Call somebody. This piece will still be here when you land.

And honestly, I am still processing it.

Subscribe now

Let me tell you what Anthropic has shipped since February. Somewhere between 74 and 120 product releases in roughly ninety days, depending on whose list you trust.

A new flagship model. Then a better one. Then a cybersecurity model that found a bug hiding in OpenBSD since the year Windows 98 came out, which is to say since before some of you were potty-trained. Then Claude Managed Agents, a product that turned the business plans of several hundred AI agent startups into a checkbox in somebody else’s pricing tier, practically overnight. Then a desktop agent called Cowork that reads your files, drafts your emails, and now pairs with your phone so you can text it tasks from the train. Then an Excel and PowerPoint integration that builds your decks and pivot tables for you. Then Claude Design, which does most of what Figma does from a sentence.

Anthropic’s annualised revenue run rate went from $9 billion last December to over $30 billion this April. Their valuation is $380 billion. Their IPO is rumoured for October. Their chief product officer resigned from Figma’s board on 14 April, three days before the tool dropped, which is the corporate equivalent of breaking up with someone via their LinkedIn profile.

I read all this on the train and laughed. Then I read it in the shower and did not laugh. Then I read it a third time on Sunday with coffee, and realised I had been smiling nervously at my phone for about forty minutes.

The Forbes 50 is genuinely impressive. Lovable. Cursor. Harvey. Runway. Perplexity. Suno. Decagon. Cognition. $305.6 billion raised between fifty companies. OpenAI alone accounts for $182.6 billion of that, which is nearly four times the entire annual federal budget of Nigeria for 2026. For one company. To build one thing. Which also happens to write better emails than most of my cousins.

Every company on that list is doing something that would have looked like witchcraft in 2019, for the very reasonable price of putting about a hundred million people out of a job.

And here is what bothers me about the list.

Nobody on it looks particularly happy.

Something is going quietly wrong with the startup dream right now and nobody wants to say it at dinner.

A $100 million exit used to be the dream. Paul Graham sold Viaweb for $49 million and became Paul Graham. Wordle sold to the New York Times for a low seven-figure sum and the founder went home to the rest of his life. Most of you reading this could retire on those numbers and still buy your mother a house.

In today’s market, a $100 million exit is described by VCs as a “consolation prize”. The pattern is predictable. A founder exits for $1 million and the first thing they think is, I need to start something that sells for $10 million. The one who exits for $10 million is already aiming at $100 million. The ladder climbs itself.

We do not need $10 million.

We need a reason to do something at all.

Meanwhile the ground keeps moving. A year ago your startup was building a feature. Today Anthropic shipped that feature as a checkbox. Your investors are now on a call explaining that actually $500 million is the new floor, which is a bit like being told the game has moved from chess to chess-but-the-board-is-a-helicopter. The rules changed. You did not. The goalposts moved because somewhere in San Francisco a man named Dario shipped something on a Tuesday.

This is not greed. It is not the economy. It is physics.

The ground under the numbers got softer, and we all kept measuring.

Here is the harder thing. For most of human history there was a wall. On one side, people who made things. On the other side, people who used the things. Designers made the website. Coders made the app. Writers made the book. Lawyers drafted the contract. Everybody else paid for the output.

That wall is being dismantled in real time. Lovable lets anyone build an app by describing it. Cursor turned coding into a conversation. Runway turned video into a prompt. Suno turned songwriting into a sentence. Claude Design turned prototyping into a paragraph your aunt could write. Literally. Somebody’s aunt is about to prototype a catering app this weekend, call it Mama Something Kitchen, and refuse to accept edits.

For the person who always wanted to make something and did not know how, this is a gift.

For the person who spent twenty years learning to code or design or draft or compose, it is something more complicated. Call it grief. Call it opportunity. Call it whatever helps you sleep. The wall was a moat. The moat is gone. The ducks have notes.

Will I still have a job?

Tens of thousands of tech workers have been cut this year already, and we are not yet at Easter. Meta announced 10% layoffs for May 20. Block, Snap and Atlassian all blamed AI efficiency like it was a weather event. Figma has lost more than 80% of its post-IPO peak value. Atlassian, a company whose entire business is collaboration software, just posted its first-ever decline in enterprise seat count. Workday, whose entire business is managing other people’s workforces, quietly laid off 8.5% of its own. Intuit is down 33%. Salesforce is down 40%. Adobe is down 30%. Thomson Reuters fell 16% in a single day when Claude Cowork launched legal plugins. IBM had its worst single-day drop since 2000 the week Anthropic released Claude Code for COBOL modernisation.

The SaaS industry ran on per-seat pricing for twenty years. AI agents do not need seats.

Here is the interesting counter-story. When Anthropic launched its HR plugin in February, the tech world waited for Deel to sweat. Instead Alex Bouaziz, Deel’s CEO, posted something on X that I have not been able to stop thinking about. Anthropic’s engineers, he noted, could probably build an HR system in three weeks. And yet Anthropic itself uses Workday. Why? Because the hard part of HR is not the software. It is carrying the legal and compliance liability when somebody in Romania gets their payroll wrong and the labour inspector shows up. Anthropic is not going to become an employment law firm in 150 countries. The moat is boring.

Boring, it turns out, is often the moat.

The honest answer to will-I-still-have-a-job is: probably yes, if you are willing to become something adjacent to what you currently are. The jobs that survive will not be the ones that were hardest to automate. They will be the ones that were hardest to explain. If you can write down exactly what you do in a single document, a model will probably do a decent version of it by Christmas. If you cannot, you are safer than you think.

That is not a comfortable truth. It is just the one I keep landing on.

So what do you actually do on Monday morning.

Here is what I have landed on, after a week of staring at stock charts like a mildly concussed seagull. You cannot out-ship Anthropic. You cannot out-raise OpenAI. You cannot be more efficient than a model that works while you sleep and does not ask for PTO. If you spend your life trying, you will lose, because the scoreboard is not yours, the game is not yours, and the numbers on it are decided by a 26-year-old in Menlo Park whose therapist has not seen him in three months.

The move, as far as I can tell, is to stop measuring yourself with somebody else’s ruler. Pick a small thing. Do it well. Do it tomorrow. Do it the day after.

Write the paragraph. Finish the meeting properly. Call the friend you have been avoiding. Read the book instead of the feed. Cook the meal. Put your phone in another room. Say no to the thing you do not want to do. Say yes to the thing you have been putting off for six months.

None of this will tank anyone’s stock.

None of this will make Forbes.

But it is the only work that compounds in a direction that matters, which is toward a version of yourself you might actually like when you are sixty.

The next time somebody asks me what I think about AI eating the world, I am going to tell them this. The world is not being eaten. It is being reorganised. A lot of people are losing jobs. A lot of people are getting superpowers. Most of us are doing both at once and pretending neither is happening. The companies on that Forbes list will keep breaking things faster than we can adapt. The stocks will keep moving. The goalposts will keep moving. The wall will keep falling.

And tomorrow morning, somebody is going to wake up, ignore the three news alerts about which billion-dollar company just bought which billion-dollar company, and spend an hour on something that nobody is going to pay them for. A book. A letter. A walk. A conversation. A half-decent idea about how to be a slightly better version of themselves next Tuesday.

They will be winning.

Most of us would be, if we noticed what the game actually was.

• • •

P.S. I wrote most of this while watching Figma’s stock chart do a waterfall impression, and I realised halfway through that I was checking the chart more than I was writing the essay. I put the phone in another room. The essay got finished. That, I think, is the whole point.

P.P.S. I promised myself I would not use Claude to write a single word of this. Not as a principled stand. More like: if I am going to sit in judgement of an industry, the absolute minimum is that I type my own nonsense.

This piece is about what history says to do about it.

Subscribe now

Here is something I find genuinely clarifying: almost every structural problem this sector faces has been solved somewhere else, in a different industry or a different era, with a different set of actors. The solutions are not hidden. They are sitting in case studies and corporate histories that the African energy sector largely ignores because it is too busy attending conferences about blended finance instruments.

There is a name for what we are experiencing. The economist Carlota Perez studied five technological revolutions over more than 200 years and found a recurring pattern: an installation phase where infrastructure is built through bubbles, speculation, and financial experimentation, followed by a deployment phase where broad societal adoption happens and the real economic gains materialise. The transition between the two, what Perez calls the turning point, requires institutional innovation. Policy. Shared standards. Market-making infrastructure.

The global renewable energy sector is in that transition right now. Solar costs have collapsed. Deployment is beginning at scale in India, Brazil, China, the US. But Africa is attempting to enter the deployment phase without having built the installation-phase institutions. Each African developer must individually traverse both phases, which is structurally impossible for small firms.

Mariana Mazzucato’s complementary framework sharpens the point further. Mazzucato distinguishes between “market fixing” and “market shaping”. The conventional DFI approach to African energy is market fixing: de-risk individual projects, provide concessional capital, correct specific market failures. What the sector actually needs is market shaping: building the institutions that create a functioning market where one does not yet exist. The foundry is not a risk mitigation instrument. It is a market creation instrument.

* * *

The Cleantech 1.0 Autopsy: We Have Seen This Movie Before

Between 2006 and 2011, US venture capital firms poured over $25 billion into clean energy startups. They lost more than half of it. MIT’s Energy Initiative found that more than 90% of cleantech companies funded after 2007 failed to return initial capital to investors.

The post-mortem was devastating. VC firms had expected clean energy to follow the software playbook. Rapid iteration, hockey-stick growth, quick exit. What they got instead was capital-intensive manufacturing, ten-year development timelines, commodity products with razor-thin margins, and dependence on government subsidies that could evaporate overnight.

Van den Heuvel and Popp, in NBER Working Paper 29919, found something that should be tattooed on the forehead of every energy VC: clean energy firms struggled to generate outsized profits because of difficulties differentiating products and increasing market power.

Translation: solar panels are solar panels. You cannot 10x your price because your brand is cooler.

Bessemer Venture Partners did the retrospective in November 2022, analysing 394 cleantech companies via Pitchbook. During Cleantech 1.0, 90% of companies failed to return capital to investors. The survivors added software and services layers. Opower survived by wrapping energy efficiency analytics around utility relationships. First Solar survived by being disciplined: low leverage, strong balance sheet, proprietary cadmium telluride thin-film technology that actually differentiated it from commodity silicon.

SunEdison did the opposite. Aggressive growth. A yieldco model that loaded the parent with $214 million in interest expenses in a single quarter while spinning off operating assets. It filed for bankruptcy in April 2016. Solyndra received a $535 million Department of Energy loan guarantee and still went bankrupt when Chinese manufacturers drove down crystalline silicon costs. Technological novelty is not a moat in commoditised energy hardware. Neither is government-backed capital, by itself.

The lesson is not subtle: clean energy is an infrastructure category, not a venture capital category.

VC works for technology risk. Will this thing work?

Clean energy’s primary risk is deployment and financing. Can we build and sell it at scale?

Those are different risk profiles requiring different capital structures.

And yet here we are in 2026, watching African solar developers try to raise equity from investors who expect software returns on infrastructure timelines. The African tech funding boom saw startups raise approximately $4 to $5 billion in 2021. Then the correction hit. The naira collapsed. Many died. But the ones that survived, Moniepoint, Flutterwave, Paystack, did so because they had built recurring revenue on infrastructure, not because they had the best pitch decks. The lesson for energy is the same.

* * *

The Semiconductor Analogy

I keep coming back to semiconductors because the structural analogy is almost perfect.

Before the 1990s, the semiconductor industry was dominated by Integrated Device Manufacturers. Companies like Intel, Texas Instruments, and Motorola that did everything: designed chips, built the fabrication facilities to manufacture them, sold the finished product. This worked when fabs cost a few hundred million dollars. It stopped working when fabs crossed a billion. Today, a state-of-the-art TSMC fab costs $20 billion or more. The cost trajectory only makes the foundry model more important over time, not less.

In 1987, Morris Chang founded TSMC with a radical idea: separate chip design from manufacturing. Instead of every company building its own fab, TSMC would manufacture chips for anyone. This single innovation created an entirely new category: the fabless semiconductor company. Qualcomm, NVIDIA, Broadcom, and eventually thousands of others could focus entirely on chip design without needing billions in manufacturing capital.

TSMC now holds nearly 70% of global foundry revenue. The scale effect of the dedicated foundry model in semiconductors demonstrates what one well-designed shared infrastructure institution can achieve. IDM market growth stagnated at 1% to 4% CAGR between 2007 and 2024, while the fabless market is projected to grow at 9.6% annually through 2032 per SNS Insider. A McKinsey analysis by Naeher, Suzuki and Wiseman covering 1996 to 2009 found that semiconductor players destroyed a combined $140 billion in value, with most IDMs netting out at zero. The winners were Intel, Texas Instruments, and TSMC. Everyone in the middle got crushed.

Sound familiar?

Right now, most African renewable energy companies are trying to be IDMs. They originate projects, develop them, arrange financing, manage EPC, then operate and maintain the assets. The economist Oliver Williamson, in his 2009 Nobel lecture, explained why this happens. When market-based transactions are too costly, due to incomplete contracts, asset specificity, and information asymmetries, firms vertically integrate. In developed energy markets, a developer can outsource legal counsel, financial advisory, environmental assessment, and technical review to specialist markets. In Africa, these specialist markets barely exist. Every project requires bespoke legal structures. Lenders cannot assess risk without comparable-transaction track records. Regulatory uncertainty makes contracts incomplete by design.

The Williamson-consistent response is vertical integration. But vertical integration has overhead costs prohibitive for small firms. This is the IDM trap.

The contrast with Samsung is instructive. Samsung operates as both chip designer and chip manufacturer, creating inherent conflicts of interest. Fabless customers like NVIDIA and Qualcomm have expressed concern about IP leakage when manufacturing at Samsung Foundry. Samsung is under sustained pressure to spin off its foundry business precisely because trust is critical in the foundry business. The mapping to African energy is direct: state utilities that both buy power as offtaker and compete with IPPs as developers are Samsung, not TSMC. They create the same conflict-of-interest trap. The foundry model requires institutional separation.

Here is the part most people miss about the TSMC story: government capital was decisive. Taiwan’s National Development Fund took a 48.3% stake. Philips invested approximately 27.5%. The government recruited Morris Chang from the US, built the research infrastructure through ITRI, and offered tax benefits that de-risked the entire venture. Five factors drove Taiwan’s semiconductor vertical disintegration: industrial clustering, fast technology change, rising development costs, emergence of design firms, and government support. Remove any one and TSMC probably does not happen.

The African RE sector has the first four. It emphatically lacks the fifth.

* * *

The Seplat Story

The question I get most often is some version of: Seplat and Oando started small and became major players in oil and gas. Why can’t renewable energy developers do the same?

It is the right question. And the answer is instructive precisely because most people cite it incorrectly.

Seplat was formed in 2009 as an SPV. Its founders, through Platform Petroleum and Shebah Petroleum, had spent years building smaller companies first. Platform Petroleum had been developing and operating its Egbaoma marginal field since the 2002/2003 licensing round. That operational track record was the credential that allowed them to participate in the much larger Shell divestment.

And that marginal field opportunity existed because of a deliberate government policy intervention. The 2002/2003 Marginal Field Bid Round created a protected entry lane for small indigenous operators. Fields that IOCs had discovered oil in but left undeveloped for ten-plus years. Confirmed discoveries with known reserves and reduced royalty rates, 2.5% for onshore fields below 5,000 bopd, versus the then-standard approximately 20% under older terms. Not exploration gambles. Starter assets.

Think about that. The government did not say “go find oil and compete with Shell.” It said: “Here are proven reserves. Reduced royalties. Simplified terms. This is your entry lane.” And then small companies used those starter assets to build track records that qualified them for bigger opportunities.

By late 2024, confirmed in 2025, indigenous companies accounted for over 50% of Nigeria’s total oil and gas output for the first time in history. Shell, ExxonMobil, Eni, TotalEnergies, all sold onshore assets to local companies. Seplat completed its $1.28 billion ExxonMobil (MPNU) acquisition.

Now compare this to what RE developers get. DARES is a $750 million World Bank initiative targeting 17.5 million Nigerians. Real and significant. But grants are paid after commissioning, not before. Developers must pre-finance everything. There is no farm-in/farm-out mechanism. Revenue depends on community willingness-to-pay. No secondary market exists for operating assets.

Africa GreenCo is the closest thing to a renewable energy equivalent of the marginal field programme, and it is already operational. GreenCo is a creditworthy intermediary offtaker and energy trader in the Southern African Power Pool. It purchases power from renewable IPPs and re-sells to utilities and commercial buyers, absorbing the offtaker credit risk that makes IPP investment impossible when the counterparty is an insolvent state utility. The US Development Finance Corporation committed a $40 million facility to GreenCo in October 2024, backing over 350 MW of renewable capacity across Zambia, South Africa, and Namibia. GreenCo explicitly addresses “structural and market weaknesses rather than just financial symptoms.”

* * *

Africa’s Own Proof of Concept: Telecom Towers and Geothermal Wells

You do not even need to look outside the continent. Africa has already built two foundry models.

Telecom towers. Mobile operators, MTN, Airtel, Glo, originally built and owned their own towers. Capital-intensive and non-core. The solution: sell to specialised towercos and lease back. IHS Towers, founded in Lagos in 2001, grew to 39,229 towers across seven operating markets by end of 2024. The model separated infrastructure ownership from service delivery. PwC found that operators that separated infrastructure assets enjoyed valuation premiums of 30% to 50%.

Yes, MTN announced a $6.2 billion deal to reacquire IHS in February 2026. The “disaggregation failed” takes flooded my timeline. They are wrong. MTN is reacquiring because it accounts for roughly 70% of IHS’s revenue. At that concentration, the independent model is economically irrational for this specific case. The reacquisition is the exit, not the failure. For renewable energy, where demand is far more fragmented, no single offtaker would ever reach 70%. The RE towerco model would have a much longer independent lifecycle.

Geothermal in Kenya. This is the parallel nobody in the solar conversation knows about, and it is the most direct proof that the foundry model works in African energy. Kenya’s Geothermal Development Company, established in 2008, absorbs the riskiest phase of geothermal development: exploration and drilling. GDC drills wells and develops steam fields. Private operators then sign agreements to purchase steam and build power plants. At the Menengai field, GDC drilled wells producing 105 MW of steam capacity. Sosian Energy’s 35 MW plant started supplying power to the grid in 2023.

GDC is the geothermal foundry. It processes the raw resource so that “fabless” IPPs can build power plants without needing to develop geothermal expertise or absorb resource risk. The CIF/AfDB Menengai case study states it directly: “The GDC model shifts exploration and resource risks away from private investors.”

This is not a future concept. It is operational. In Africa. Today. The question is why nobody has built the equivalent for solar.

* * *

The Countries That Built the Foundry: India, Brazil, and Morocco

The parallels above are from other industries and one African sector. But three countries have built the actual energy foundry, the shared market infrastructure for renewables, and the contrast with the rest of Africa is so stark it should end every conversation about bankability.

India. In 1987, the same year Morris Chang founded TSMC, India established IREDA. A government-owned non-banking financial institution with one job: finance renewable energy projects from conception to post-commissioning. IREDA did not just lend money. It standardised loan structures. It built technical appraisal capacity so banks did not have to evaluate solar technology from scratch on every deal. It took long-tenor risks that commercial banks refused. By FY2026, IREDA’s loan book stood at ₹93,075 crore, approximately $10 billion. Loan sanctions in FY2025 hit ₹47,453 crore, a 27% year-on-year increase.

Then India added SECI, a procurement intermediary that aggregates demand through centralised auctions and on-sells to distribution companies, providing a creditworthy central offtaker. India’s IEX energy exchange has over 8,100 participants including more than 2,100 renewable generators. The virtuous cycle: auction certainty generates financing appetite. IREDA can lend against a predictable forward project pipeline because SECI creates it.

Result: India installed 24.5 GW of solar in 2024. Nigeria installed 803 MW in 2025, its strongest year ever. A 25x difference in annual deployment. Architecture, not sunshine.

Brazil. BNDES, Brazil’s development bank, has financed roughly 70% of all renewable energy projects in Brazil since 2000, with documented financing of $33.1 billion for the 2004-2020 period. BNDES standardised project finance structures. It imposed local content requirements that drove manufacturers to build Brazilian factories. And competitive federal auctions provided revenue certainty that made ordinary bank lending sufficient.

Result: Brazil surpassed 55 GW of installed solar capacity in 2025, more than doubling in two years. Federal auction prices have cleared around $32 per MWh. Private corporate PPAs have gone even lower. Solar is now on track to overtake hydropower as Brazil’s second-largest power source by 2032.

Morocco. This is the example that eliminates the “that will not work in Africa” objection. Morocco’s MASEN, the Moroccan Agency for Sustainable Energy, functions simultaneously as procurement authority, project co-developer taking 25% equity stakes in project SPVs, financing intermediary receiving DFI loans and on-lending to project companies, and strategic policy actor. Its Noor Ouarzazate complex, 580 MW of concentrated solar and PV, was financed with $1.6 billion from seven DFIs. Morocco now has over 4,680 MW of operational renewable capacity and targets 52% renewables by 2030. MASEN has Green Climate Fund accreditation, enabling it to access GCF financing directly.

MASEN absorbs sovereign and regulatory risk for private investors. It runs transparent competitive procurement. It uses each bidding round to stimulate local manufacturing capacity. It is the most complete integrated state renewable energy development institution in Africa. And it exists. Right now.

* * *

The Consumer Foundry: What M-KOPA and Sun King Built

There is one more parallel that most C&I energy people dismiss. They should not. Because it solves the same problem at a different scale.

M-KOPA solved the consumer solar bankability problem not by finding already-bankable customers but by creating bankability. An embedded GSM chip enables remote monitoring and disabling, solving the collateral problem with IoT instead of legal contracts. Daily M-Pesa payments create a credit track record that substitutes for a credit bureau.

Sun King scaled this further. In May 2023, Sun King completed a $130 million securitisation, the first of its kind in Sub-Saharan Africa for off-grid solar. Then in July 2025, Sun King closed a $156 million securitisation of approximately 1.4 million solar products, backed by five commercial banks: Absa, Citi, Co-operative Bank of Kenya, KCB, and Stanbic Bank Kenya, plus three DFIs: British International Investment, FMO, and Norfund. The largest securitisation in Sub-Saharan Africa outside South Africa. D.light has done $490 million in total securitised financing since 2020.

The progression from $130 million to $156 million in two years illustrates market maturation. Sun King and M-KOPA did not lower the cost of capital by finding better borrowers. They lowered it by building the information infrastructure that lenders require. This is the foundry in its most elegant form: shared technology infrastructure that converts un-bankable transactions into bankable ones.

CrossBoundary Energy Access is building this in mini-grids. CBEA finances construction and owns assets. The developer focuses on origination and operations. The developer retains minority equity and the O&M contract. In January 2026, CBEA completed the first acquisition of an operational mini-grid platform, ANKA in Madagascar. The mini-grid sector quietly built the foundry that C&I solar has not.

* * *

The Architecture, Not the Finance

Let me pull all seven parallels together.

1. In Cleantech 1.0, the survivors built services layers and recurring revenue.

2. In semiconductors, TSMC separated capital-intensive manufacturing from knowledge-intensive design, and government capital underwrote the infrastructure layer.

3. In Nigerian oil and gas, government-created entry lanes gave indigenous companies a structural pathway that raw competition never would have.

4. In telecoms, tower separation unlocked capital.

5. In Kenyan geothermal, GDC absorbed resource risk so private operators could deploy.

6. In India, Brazil, and Morocco, specialised green lenders, procurement intermediaries, and standardised auctions created the rails that made deployment possible at scale.

7. At the consumer level, M-KOPA and Sun King built information infrastructure that manufactured bankability from un-bankable customer bases.

Every time, the intervention that worked was architectural. Every time, the intervention that failed was financial tweaking on a broken structure.

The German Energiewende did not work because German banks invented sophisticated blended finance. It worked because a 20-year feed-in tariff guarantee made ordinary bank lending sufficient. A German Sparkasse could look at 20 years of revenue certainty and extend credit to a farmer with solar panels on a barn roof. The instrument was not sophisticated. The guarantee made sophistication unnecessary. By 2012, nearly half of Germany’s installed renewable capacity was owned by citizens, farmers, cooperatives, and small businesses. There are over 350 Sparkassen in Germany.

The Rural Electrification Administration, created by executive order in 1935 and given permanent legislative mandate through the Rural Electrification Act of 1936, did not work because rural Americans got better financial products. In the mid-1930s, nine out of ten rural farms had no electricity. The response was to reorganise the market. Create rural electric cooperatives. Change the ownership model entirely. Provide 2% loans of up to 35 years to those cooperatives. Not to individual projects. To the organisations themselves. By 1953, over 90% of US farms had electricity, at roughly half the cost private utilities had estimated.

Khanna and Palepu, the Harvard economists who coined “institutional voids,” described precisely this. In emerging markets, the absence of credit bureaus, contract enforcement, quality certification, professional intermediaries, forces firms to internalise costs that markets would otherwise absorb. Each firm fills the same voids individually, at enormous cost.

The foundry fills those voids for all participants simultaneously.

Drawing from all seven parallels, the foundry performs four specific functions. Risk absorption: GDC absorbs exploration risk, GreenCo absorbs offtake credit risk, MASEN absorbs financing intermediation risk. Standardisation: SECI and Brazil’s auctions standardise procurement, enabling price discovery and capital formation. Scale aggregation: IREDA aggregates small project demand into financeable portfolios. Market signalling: Germany’s feed-in tariff provides the long-dated revenue certainty against which commercial banks can lend.

The African energy sector needs an institution, or a set of coordinated institutions, that performs all four functions. Some pieces already exist. InfraCredit provides naira-denominated credit guarantees. GreenCo intermediates offtake risk. ETAFA Nigeria, a $50 million local currency subordinated debt fund launched by GEAPP and Chapel Hill Denham, has deployed $10 million in concessional capital to catalyse $40 million in commercial finance for distributed renewable energy. Each is a partial foundry. None covers the full stack.

The question is not whether this architecture is needed. It is whether it will be built fast enough.

This is an architecture problem.

* * *

P.S. I spent three weeks reading about semiconductor fabrication plants for this piece. My brother asked me what I was working on and I said “a blog post about solar panels.” She looked at my screen, which had a 40-page McKinsey report about chip foundries open, and said “you need to go outside.” She is not wrong. But also, the fact that a 1987 Taiwanese chip factory explains why a solar developer in Lagos cannot close a deal in 2026 is either the most important insight in this series or proof that I have lost the plot entirely. I am genuinely unsure which.

P.P.S. Nigeria’s government just approved a ₦4 trillion bond to settle power sector debts and signed a ₦100 billion bank facility for mini-grid developers. The political will is not zero. What is missing is the institutional design that converts political will into market architecture. Morocco had a King who decided renewable energy was a national priority. Nigeria has a federal system where twelve states are simultaneously building their own electricity markets. Different governance. Same foundry problem. Arguably harder. Definitely more interesting.

P.P.P.S. The German Energiewende section required me to learn what a Sparkasse is. It is a German savings bank. There are over 350 of them. They funded more renewable energy than any venture capital firm in history, not because they were innovative, but because a 20-year government tariff guarantee made the lending decision boring. The most transformative energy policy in modern history worked because it made finance boring. I think about this constantly. The African energy sector is addicted to innovative finance. What it needs is boring finance backed by credible policy. Nobody will fund a conference panel on “making energy finance more boring.” Somebody should.

P.P.P.P.S. I cited Williamson, Perez, Mazzucato, Khanna, and Palepu in a blog post about solar panels. My PhD supervisor will either be proud that I am applying academic frameworks to real-world problems or concerned that I am using a Substack to avoid writing my actual thesis. Both interpretations are correct. A full reference list for all data cited across this series is available below. If you made it to the fifth P.S. section, you are either my mother or someone who genuinely cares about African energy market structure. Either way, thank you.

Kay is Manager for West Africa at Berkeley Energy Corporate Solutions, where he structures and finances renewable energy projects across 16 countries in West and Central Africa. He is a PhD researcher in multi-physics modelling of solar-biomass-hydrogen hybrid systems at the University of North Dakota, and writes The Impostor’s Guide to Clean Energy at kaykluz.com. The views expressed here are his own and do not represent the views of his employer.

References

1. Cost of Capital & Opening Frame

1. Clean Air Task Force (October 2024). High Capital Costs Are Stalling Clean Energy Investment Across Africa. catf.us

2. International Energy Agency (June 2024). Clean Energy Investment for Development in Africa. iea.blob.core.windows.net (PDF)

2. Carlota Perez Framework

3. Perez, Carlota (2002). Technological Revolutions and Financial Capital. Edward Elgar. e-tcs.org (PDF)

4. Wikipedia. Technological Revolutions and Financial Capital. en.wikipedia.org

5. Perez, Carlota. TRFC Chapter 4. carlotaperez.org (PDF)

3. Mazzucato Framework

6. Mazzucato, Mariana & Penna, Caetano. Beyond Market Failures: The Market Creating and Shaping Roles of State Investment Banks. INET Working Paper. ineteconomics.org (PDF)

7. Mazzucato, Mariana (2015). From Market Fixing to Market Creating. ISIGrowth Working Paper. isigrowth.eu (PDF)

4. Cleantech 1.0

8. Gaddy, Sivaram, Jones, Wayman (July 2016). Venture Capital and Cleantech: The Wrong Model for Clean Energy Innovation. MIT Energy Initiative MITEI-WP-2016-06. energy.mit.edu (PDF)

9. Van den Heuvel, Matthias & Popp, David (2022). The Role of Venture Capital and Governments in Clean Energy. NBER Working Paper 29919. nber.org (PDF)

10. Bessemer Venture Partners (November 2022). Eight Lessons from the First Climate Tech Boom and Bust. bvp.com

11. Utility Dive (2013). How Opower Sells Energy Efficiency to Utilities. utilitydive.com

12. MIT Sloan. First Solar Case Study. mitsloan.mit.edu (PDF)

13. Kunz v. SunEdison Class Action Documents. classaction.org (PDF)

14. Bloomberg (April 2016). SunEdison Files for Bankruptcy After Acquisition Binge. bloomberg.com

15. U.S. Department of Energy Inspector General. Solyndra Special Report 11-0078-I. energy.gov

16. Fortune (August 2015). Why the Solyndra Mistake Is Still Important to Remember. fortune.com

17. TechCrunch (February 2022). African Startups Raised Record-Smashing $4.3B to $5B in 2021. techcrunch.com

5. Semiconductor / TSMC Analogy

18. Wikipedia. Morris Chang. en.wikipedia.org

19. SemiWiki. How Philips Saved TSMC. semiwiki.com

20. Construction Physics (May 2024). How to Build a $20 Billion Semiconductor Fab. construction-physics.com

21. Taipei Times (March 2026). TSMC Nets Nearly 70% of 2025 Foundry Market. taipeitimes.com

22. The Edge Malaysia (March 2025). IDM Market Growth 2007-2024. theedgemalaysia.com

23. SNS Insider / GlobeNewswire (October 2024). Semiconductor Fabless Market Report 2024-2032. globenewswire.com

24. Naeher, Suzuki & Wiseman (Autumn 2011). The Evolution of Business Models in a Disrupted Value Chain. McKinsey on Semiconductors. mckinsey.com (PDF)

25. National Central University (2004). The Vertical Disintegration of Taiwan’s Semiconductor Industries. scholars.ncu.edu.tw

26. Asterisk Magazine (March 2026). The Institute Behind Taiwan’s Chip Dominance. asteriskmag.com

27. The Korea Herald (2025). Samsung Foundry and IP Leakage Concerns. koreaherald.com

28. Williamson, Oliver E. (2009). Nobel Prize Lecture: Transaction Cost Economics. nobelprize.org

6. Seplat / Nigerian Oil & Gas

29. Wikipedia. Seplat Petroleum Development Company. en.wikipedia.org

30. Platform Petroleum. Official History. platformpet.com

31. NUPRC. Guidelines for the Award and Operations of Marginal Fields in Nigeria. nuprc.gov.ng (PDF)

32. Mondaq. Marginal Fields and the Petroleum Industry Act 2021. mondaq.com

33. Morgan Lewis. Nigeria Overhauls Its Oil and Gas Laws with Petroleum Industry Act. morganlewis.com

34. BusinessDay Nigeria (February 2025). We Control 50% of Nigeria’s Oil Production. businessday.ng

35. Punch Newspapers (December 2024). Seplat Completes $1.28bn Purchase of Mobil Nigeria. punchng.com

7. Africa GreenCo

36. Africa GreenCo. Official website. africagreenco.com

37. DFC / Africa GreenCo. DFC Commits a USD 40m Facility to GreenCo. africagreenco.com (PDF)

38. Engineering News (October 2024). USD 40m Facility to GreenCo. engineeringnews.co.za

8. Telecom Towers / IHS

39. IHS Towers. Official History. ihstowers.com

40. IHS Towers 2024 Annual Report (Form 20-F). ihstowers.com (PDF)

41. IHS Towers Press Release (February 2026). IHS Towers Announces Proposed Sale to MTN Group. ihstowers.com

42. MarketForces Africa (February 2026). Highlights of MTN-IHS Towers $6.2bn Acquisition Deal. dmarketforces.com

43. PwC (October 2025). Telecom Operators: Reinventing Their Business Models. pwc.com

9. Kenya Geothermal / GDC

44. Wikipedia. Geothermal Development Company. en.wikipedia.org

45. GDC. 15 Years Milestones Blog. gdc.co.ke

46. CIF/GDI. Menengai Case Study Kenya. effectivecooperation.org (PDF)

47. ThinkGeoEnergy. 35-MW Menengai Geothermal Power Plant Starts Grid Supply. thinkgeoenergy.com

48. Green Building Africa (August 2023). Sosian Energy Commissions 35MW Geothermal Power Plant in Kenya. greenbuildingafrica.co.za

49. CIF/AfDB. Kenya Menengai Geothermal Development Project Supplementary Document. cif.org (PDF)

10. India (IREDA, SECI, IEX)

50. Wikipedia. Indian Renewable Energy Development Agency. en.wikipedia.org

51. Business Standard (April 2026). IREDA Loan Book Climbs 22% YoY in March 26. business-standard.com

52. Economic Times (March 2025). IREDA Reports 27% Rise in Loan Sanctions in FY25. economictimes.com

53. SECI. What We Do. seci.co.in

54. Indian Energy Exchange (IEX). iexindia.com

55. PV Magazine (January 2025). India Adds Record 24.5 GW of Solar in 2024. pv-magazine.com

56. Leadership Nigeria. Nigeria Installs 803MW of Solar Capacity in 2025. leadership.ng

11. Brazil (BNDES)

56. New Development Bank (2024). BNDES Project Performance Evaluation Report. ndb.int (PDF)

57. Energy Tracker Asia. Solar Energy in Brazil. energytracker.asia

58. EIA (September 2025). Brazil Solar vs Hydropower Trajectory. eia.gov

12. Morocco (MASEN)

59. World Bank / PPIAF. Morocco Noor Ouarzazate Solar Complex Case Study. ppp.worldbank.org (PDF)

60. Climate Investment Funds (February 2016). World’s Largest Concentrated Solar Plant Opened in Morocco. cif.org

61. Energy Partnership Morocco. MASEN News. energypartnership.ma

62. Green Climate Fund. MASEN Accredited Entity. greenclimate.fund

13. Nigeria — DARES

63. World Bank Press Release (December 2023). Nigeria to Expand Access to Clean Energy for 17.5 Million People. worldbank.org

64. Nigeria Energy Transition Office. DARES Programme. energytransition.gov.ng

14. M-KOPA / Sun King / Consumer Foundry

65. GSMA Mobile for Development. M-KOPA. gsma.com

66. Sun King (2023). $130 Million First-of-its-Kind Securitisation. sunking.com

67. Sun King (July 2025). $156M Securitisation to Deliver Solar for Over a Million Kenyans. sunking.com

68. Citi Press Release (July 2025). Citi Sun King Securitisation Deliver Solar Million Kenyans. citigroup.com

69. Kenyan Wall Street (July 2025). Sun King Secures Record-Breaking US$156 Million. kenyanwallstreet.com

70. D.light / PR Newswire. D.light Closes USD125M Securitization Facility. prnewswire.com

15. CrossBoundary Energy Access / Mini-Grids

71. CrossBoundary (January 2026). CrossBoundary Energy Access ANKA Mini-Grid Acquisition Madagascar. crossboundary.com

72. CrossBoundary. Energy Access Open-Sources Project Financing Approach for Mini-Grids. crossboundary.com

16. ETAFA / InfraCredit

73. Global Energy Alliance for People and Planet / Chapel Hill Denham. $50M Energy Transition Africa Fund Nigeria. energyalliance.org

74. InfraCredit Nigeria. Financial Guarantee Product. infracredit.ng

17. German Energiewende

75. Clean Energy Wire. Defining Features of the Renewable Energy Act (EEG). cleanenergywire.org

76. EnergyTransition.org (October 2013). Citizens Own Half of German Renewables. energytransition.org

77. EnergyTransition.org (January 2014). The Hidden Power of Local Finance. energytransition.org

78. Deutscher Sparkassen- und Giroverband (DSGV). dsgv.de

18. US Rural Electrification Administration

79. WECI. Powering Rural America: The REA of 1936. weci.net

80. Wikipedia. Rural Electrification Act. en.wikipedia.org

81. University of Wisconsin Rural Electric Research. reic.uwcc.wisc.edu

82. National Park Service. Rural Electrification. nps.gov

83. U.S. Code Title 7 Chapter 31 Subchapter III. Rural Electrification Act Statute. uscode.house.gov

84. Investopedia. Rural Electrification Act. investopedia.com

85. Richmond Fed. Economic History — Rural Electrification. richmondfed.org

19. Academic Frameworks

86. Khanna, Tarun & Palepu, Krishna (July 1997). Why Focused Strategies May Be Wrong for Emerging Markets. Harvard Business Review, Vol. 75(4). institutional voids paper (PDF)

87. Perez, Carlota (2002). Technological Revolutions and Financial Capital. PDF

88. Mazzucato, Mariana (2015). From Market Fixing to Market Creating. PDF

89. Williamson, Oliver E. (2009). Nobel Prize Lecture. nobelprize.org

And yet there I was.

Reading the whole thing.

Twice.

Because apparently I do not believe in sleep.

But buried inside the usual fortress-balance-sheet, America-needs-to-get-its-act-together, bank-CEO energy was something important. Actually, several important things. And for once, a lot of it felt uncomfortably close to the reality some of us in energy, infrastructure, and African markets have been shouting into the void about for years.

Not in exactly the same language, of course. Jamie says things like “AI, data and technology are key to the future.” I say things like “the robots are coming, and annoyingly they need a lot of electricity.” Same message. Different tailoring.

FAIR WARNING: This is going to be long. Not Dimon-long. Nobody is Dimon-long. But long. The man wrote 50 pages to shareholders who, let’s be honest, mostly just check the dividend line and close the PDF.

If you bill $1,200 an hour and are skimming for talking points, here is your TLDR:

TLDR: Jamie Dimon wrote his 2025 shareholder letter. It is 50 pages of fortress balance sheet, geopolitics, and a Rudyard Kipling quote that I am almost certain his comms team begged him to remove. Buried in there is a $1.5 trillion plan that is, if you tilt your head, an admission that America now needs to do all the things African energy developers have been doing since forever. AI is eating all the electricity. Geopolitics is chaos. Europe is quietly falling apart. Regulations are insane everywhere. Welcome to the club, Jamie. The hazing involves a diesel genset and a substation that catches fire on Tuesdays for reasons nobody can explain.

Still here? Let’s go.

Subscribe now

First, a confession

There are two kinds of people who read Jamie Dimon’s annual letter.

The first are finance people, who pretend not to care and then read it secretly while muttering words like “macro” and “resilience” into a flat white.

The second are people like me, who read it because once every year the CEO of one of the biggest banks on earth accidentally writes 50 pages that explain why your job is both important and impossible.

I read it twice. Once for the macro, and once because watching a 70-year-old banker drop “Sturm und Drang” into a shareholder letter is the most fun you can have with your trousers on.

This year’s letter was especially irritating because he is not just talking about banking. He is talking about AI, industrial policy, national resilience, critical minerals, infrastructure, supply chains, and energy systems.

In other words, he is describing a world in which all the supposedly boring sectors are suddenly back in fashion.

Funny how that works.

For years infrastructure was treated like the boring cousin at the innovation party.

Now everybody is slowly realising the party does not happen if the boring cousin does not bring electricity.

The setup

You expect banker language. Safe language. Language with the emotional range of an audited receivable. Instead, Dimon’s letter basically says: the world is unstable, AI is real, inflation is still lurking in the bushes, credit markets have started doing creative writing, and the people who think this is all abstract are about to be slapped by physical reality.

And honestly? Relatable.

Because if you work in energy, infrastructure, project development, industrial decarbonisation, or anything adjacent to “things that must exist in the real world before PowerPoint can become prophecy,” none of this sounds theoretical. It sounds like Tuesday.

The bank stuff is, as always, a flex. $185.6 billion in revenue. 20% return on tangible common equity. Eighth consecutive year of record revenue. They move $12 trillion a day. Per day. Not a typo. Twelve. Trillion. Dollars. Per day. For context: the GDP of Nigeria, one of the largest economy in Africa, is about $400 billion a year. JPMorganChase moves that in roughly 1 hour before the first cup of cappuccino is done.

Yes Jamie, we get it, you’re winning, please put the receipts down.

The interesting part is everything else. Because Dimon, very politely and in a very expensive accent, has just written the African energy thesis. He has no idea that’s what he did. But that’s what he did.

Allow me.

Exhibit A: The $1.5 trillion industrial policy that dare not speak its name

JPMorgan launched something called the Security and Resiliency Initiative. $1.5 trillion over 10 years. $10 billion of direct equity to start. Five pillars:

1. Supply chain and advanced manufacturing (critical minerals, shipbuilding, robotics)

2. Defense and aerospace

3. Energy independence and resilience, including battery storage, grid resilience and distributed energy

4. Frontier and strategic technologies (AI, cyber, quantum)

5. Pharma and health tech

Read pillar three again. Slowly. Battery storage. Grid resilience. Distributed energy.

Friends. That is a C&I solar developer’s pitch deck. That is, word for word, what every Energy-as-a-Service shop in Africa wakes up and does. We just don’t call it “national security.” We call it “your factory cannot run on diesel anymore because the FX is eating your children.” Same energy. Different vocabulary. Way worse PowerPoint templates on our side, admittedly.

Then comes the quietest sentence in the whole letter:

“Unfortunately, we need industrial policy to guarantee our national security and resiliency.”

Sorry, what.

The Chairman and CEO of JPMorgan Chase. The high priest of free markets. The man who has spent two decades telling everyone the government should kindly shut up and sit down. Has just written the words “we need industrial policy” in a shareholder letter. In English. With his actual name on it. The word “unfortunately” in that sentence is doing about ten years of therapy.

In Lagos we call this Tuesday. The bankability of any African renewable project is a Jenga tower of MIGA, EAAIF, GuarantCo, IFC, FMO, AFC and DFC, held together with prayer and a credit committee that has been awake for eleven hours. The thing Dimon is now proposing for America is the financial duct tape we have been improvising with since 2003.

Plot twist: we were ahead the whole time. We just thought we were behind because nobody told us we were doing the future.

Someone owes us back pay.

Exhibit B: “We have a lot to catch up on and not much time”

That sentence appears twice in the SRI section. It is also, word for word, the unofficial motto of every African energy developer I have ever met. It should be printed on a mug at AEF in Cape Town. It should be tattooed on the Energy Minister of Nigeria. It is the slogan of my entire career, my PhD, and approximately 80% of my WhatsApp messages after 9pm.

Dimon is talking about America’s critical minerals dependency, semiconductor exposure, and merchant marine atrophy. I am talking about 600 million Africans without electricity, 700 GW of installed deficit, and the fact that Nigeria’s grid trips more often than my router. Same vibes. Different timeline.

When the most powerful banker on Earth starts talking about “catching up,” the rest of us get to enjoy a rare and beautiful moment: the frontier is no longer a place you visit on a fact-finding mission with a per diem and a malaria pill. The frontier is now a strategic asset class. We are the precedent. We are what you do when you cannot rely on a working grid and have to build the future anyway.

Please update your LinkedIn headlines accordingly. I have already updated mine. Twice.

Exhibit C: AI is not in the cloud, it is in someone else’s substation

For the last few years, AI has been discussed like it descended from heaven fully formed, floating six inches above the ground, untouched by boring earthly limitations like transmission capacity, land rights, cooling systems, diesel backup, political risk, customs delays, and the fact that somewhere, somehow, one procurement manager is still asking for three stamped copies and a company profile in PDF.

But physical reality is undefeated.

AI is not floating six inches above the earth, untouched by material reality.

AI is about to be slapped in the face by material reality.

And I say this with love.

Because every time somebody says AI lives “in the cloud,” a part of me dies a little.

My brother in Christ, the cloud is just someone else’s power bill. And their land issue. And their water issue. And their transmission issue. And somewhere in the background, their exhausted energy team trying to explain for the seventh time that compute does not run on vibes.

That is why one of the most important things Dimon gets right is that he takes AI seriously as a real structural force. Not a toy. Not a side feature. Not something that only matters to people who use words like “wrapper” without embarrassment.

He treats it like infrastructure.

Good.

Because that is what it is becoming.

That is where my industry enters the chat. Because AI is not just software. AI is electricity with good PR.

And yet vibes remain a major asset class.

Here is the number that should chill you, or excite you, depending on which side of the trade you sleep on:

“Huge increase in AI-driven capital spending and construction by the five hyperscalers. In 2025, this number was $450 billion, and in 2026, it will be approximately $725 billion.”

$725 billion. From five companies. In one year. On AI infrastructure. That is roughly 1.5x the total annual capex of the entire global oil and gas industry. It is approximately the GDP of Switzerland, give or take a chocolate factory. It is also, and this is the part nobody is willing to say into a hot mic, almost entirely a bet on cheap, abundant, reliable electricity that does not currently exist anywhere near where they’re trying to build the data centers.

I wrote about this last August in Energy is how Africa wins at AI. The thesis was simple: the people spending $4 trillion on AI infrastructure forgot to do the thermodynamics homework. They are putting data centers in places where electricity costs 17x what it should and the cooling physics actively hate them. I made a lot of jokes about thermometers. None of those jokes have aged.

Dimon does not say any of this. Dimon says, in his most carefully manicured voice:

“The landscape will change rapidly, with shifting assumptions about power consumption, costs, chip technologies and the speed at which data centers are deployed.”

That, my friends, is JPMorgan-speak for “we have absolutely no idea where these things are going to end up but we know it isn’t where they currently are.” Translation: the great migration is coming. The only question is whether African energy ministries understand they are about to be on the most expensive blind date in human history.

If you work in siting decisions in Lagos, Mombasa, Casablanca, Accra, or Cape Town, please for the love of God check your inbox. The next hyperscaler deal is in there somewhere, probably in your spam folder, probably from a Gmail address that ends in numbers.

Exhibit D: Europe is cooked, and Jamie says so out loud with full Catholic guilt

This is the part of the letter where Dimon stopped being a banker and became a strategist who has clearly been holding this in for years. EU GDP relative to America has gone from 90% in 2000 to 70% today. Internal EU market barriers, per Mario Draghi, function as “hard tariffs” of 45% for goods and 110% for services. A hundred and ten percent. That is not a market. That is a polite suggestion to please not trade with each other.

If you are an African developer who has spent the last decade chasing European DFIs, this should make you blink. European money is not going away. EIB, AFD, KfW, Proparco, BII and the rest will keep showing up to events with branded tote bags. But the relative weight is shrinking. The new money is American, Gulf, still some Chinese, possibly Indian, possibly Japanese if METI ever finishes its meeting.

A significant portion of the development finance that flows into African renewable energy deals comes from European institutions. I have been in deal processes where the timeline from term sheet to disbursement stretches past two years. Not because the project is bad. Not because the offtaker is bad. Because the institutional machinery on the funder side requires seventeen rounds of due diligence, three independent consultants who all say the same thing, two board approvals in different currencies and a partridge in a pear tree.

Africa needs hundreds of billions in clean energy investment every year for the next decade. It cannot afford a funding architecture that runs at the speed of European committee culture.

Update your capital stack template. The 2018 vintage, where you assumed a European senior tranche, an Anglo mezzanine, and a local equity sliver, is being quietly replaced by something that looks more like an American DFI, a Gulf sovereign, and a corporate offtaker with an SRI mandate. Get there first. Bring snacks.

Exhibit E: Regulation

There is a section in this letter where Dimon uses phrases like “convoluted and distorted,” “intensely inaccurate,” and, my personal favourite, “frankly, it’s not right, and it’s un-American.”

Reader, I felt this.

Not about banking regulations specifically. About permitting, tariff regulation, grid code compliance, environmental impact assessment requirements and the seventeen-step approval process for connecting a solar installation to a grid that barely works to begin with.

Dimon’s critique of bad regulation is that it creates the appearance of safety while generating real costs, locks up productive capital in compliance theatre, and produces outcomes that are the opposite of what was intended. He could be describing the regulatory environment for renewable energy project development in West Africa, word for word, sentence for sentence.

He also has a line I want to put on a t-shirt:

“You probably need to have real-life experience in dealing with regulations to understand this.”

Yes, Jamie. Yes, you do. After 12 years of trying to close renewable energy deals across markets where every week brings a new requirement from a different regulator who has not spoken to the previous regulator, I understand this deeply and personally.

Exhibit F: The line that is going on my fridge

Buried in the section on city competition, where Dimon is mostly grumbling about his New York property tax bill, there is this:

“No city, or company or country, has a divine right to success.”

I need every African leader, every infrastructure minister, every regulator who sat on a license application for nine months while it accumulated dust and supernatural powers, every utility CEO who thinks the capital will keep showing up because it always has, to write that on a Post-it and stick it to their monitor. Then take a photograph of it. Then frame the photograph.

The continent does not have a divine right to capital. We are competing for it. With America, which now has industrial policy. With Europe, which is finally panicking. With China, which never stopped. With India, which just discovered ambition and is ordering shoes.

The good news: we have what they need. Sun. Land. Cooling. Biomass. Critical minerals. A young workforce. A demographic bulge that is, depending on your priors, either the next great opportunity or the next great catastrophe.

The bad news: having it is not the same as monetising it. Diamonds in the ground are also just rocks until somebody digs them up, cuts them, certifies them, and convinces a 28-year-old in Antwerp to buy one. Nothing about this happens on autopilot.

In Other news - The cloud, sadly, still needs steel

This is the part I think too many people miss. Technologies do not arrive in neutral markets. They arrive in structures. And structures decide who wins.

If the rails are owned elsewhere, the chips are made elsewhere, the data centres are financed elsewhere, the standards are written elsewhere, and the power is more stable elsewhere, then “AI opportunity” can quickly become a very elegant way of saying:

Congratulations. You may now consume the future from the comfort of your own underpowered economy.

I am sorry, but no.

That is not a strategy. That is premium dependency. And this is why I keep coming back to energy. Because when people talk about AI in Africa, they often go straight to talent, software, policy, apps, training, startup ecosystems, and all the other shiny bits.

All good things. All necessary.

But underneath all of it sits the more irritating question nobody can joke away: What powers the stack? Not metaphorically. Literally. Because a data centre cannot run on optimism.

Inference does not happen because a minister said “innovation” with confidence. A GPU cluster does not care how inspiring your panel session was. A model does not train on narrative.

It trains on compute.

And compute, in a cruel act of physical realism, runs on electricity.

This is also why the physical world is becoming more valuable, not less

A lot of people seem to think AI will make the physical world less important.

I think the opposite.

I think AI is about to make the physical world brutally, embarrassingly, gloriously important again.

More power generation. More storage. More transmission. More cooling. More data infrastructure. More real assets. More industrial policy. More project finance. More grid planning. More actual grown-ups in the room.

Not fewer.

Because AI is not escaping the old economy. AI is marrying it. Messily. With expensive catering. And a prenup written by lawyers billing in six-minute increments.

That is why I find this moment so funny.

For years the physical industries were treated like the old world. Slow. Unsexy. Complicated. Full of hard hats and CAPEX and men named Peter who say things like “let us revisit the assumptions.”

Now the future’s hottest sector has discovered it is entirely dependent on all of that.

Delicious.

The productivity conversation, which gets morally serious fast

Dimon also says something that people should not skip past: AI will eliminate some jobs, and deployment may move faster than workforce adaptation. He argues that business and government need plans for retraining, reskilling, relocation support, and income assistance.

Yes. Exactly.

Because one of the most irritating habits in tech discourse is that people love disruption right up until they have to talk about displaced humans in full sentences. Then suddenly everyone becomes very abstract. “New categories will emerge.” “History shows labor adapts.” “Net productivity gains.”

Lovely. Tell that to the 29-year-old analyst, call-center worker, junior designer, or ops staff member whose function just got atomized by a model trained on the internet and several billion dollars of compute.

Even in advanced economies, adaptation will be messy. In African economies, where informality is high, social protection is thin, energy access is inconsistent, and labor absorption is already structurally weak, messy can become dangerous very quickly.

So no, I do not think the right African AI strategy is to cosplay Silicon Valley and pray. It has to ask: where will real demand come from? What local problems justify deployment? What infrastructure exists beneath the shiny demo? Who captures value? Who gets displaced? Who gets retrained? Who gets financed? Who gets left holding the motivational quote?

The one thing Dimon did not write, and it is loud

Dimon spent 50 pages writing the African energy thesis without ever mentioning Africa.

The continent with the fastest-growing population on Earth. The continent that holds an extraordinary percentage of the world’s critical mineral reserves, the same minerals he describes as essential to national security and the AI infrastructure buildout. The continent that is projected to have the largest workforce in the world by mid-century.

“America” appears 89 times in the letter. “Africa” appears zero. I counted.

That is either the saddest oversight of 2026 or the greatest accidental endorsement we are ever going to get. I am choosing endorsement. It is cheaper than therapy.

This is not a personal complaint. I am not writing an invoice to JPMorganChase. This is a structural observation: the global financial system’s most articulate spokesman wrote 50 pages about the future and forgot a continent. Which means the people who will build the infrastructure that matters in these markets are the people who did not forget the continent existed.

What I am taking away, written in bullets because my edit deadline is in twelve minutes

1. Industrial policy is now bipartisan and trans-Atlantic. Argument over. African policymakers have a 25-year head start in living with industrial policy and should be exporting playbooks, not importing them. Print receipts.

2. Energy resilience is the new national security. The C&I solar, BESS, and biomass deals we do every day just got upgraded, in the mind of the most powerful banker on Earth, to “critical infrastructure.” Adjust your tariffs. Adjust your pitch decks. Adjust your hourly rate.

3. The AI energy migration is real. $725 billion of hyperscaler capex cannot find a home where the lights flicker every six minutes. Africa is not “an option.” Africa is the rational answer to a thermodynamics question that has not yet been asked at board level. Be in the room when somebody finally asks it.

4. Europe is wobbling. Make new friends in Houston and Abu Dhabi. Send Christmas cards.

5. Nobody has a divine right to anything. Not me, not you, not the developer down the road who has been “about to close” the same deal for three years and is starting to feel like a meme.

In closing

People think AI will replace the physical world. I think it will make the physical world more valuable. More grids. More generation. More storage. More transmission. More data infrastructure. More industrial capacity. More discipline around what is actually bankable. More adults in the room.

Because AI is not escaping the old economy. AI is about to marry it. Messily. With expensive catering.

For years, infrastructure was treated like the boring cousin at the innovation party. All those sectors people looked down on while throwing money at apps that delivered premium oxygen to dogs. Now everyone is slowly realising the party does not happen if the boring cousin does not bring electricity. Funny how that works.

Which is excellent news for those of us in energy. We have spent years being told we are in a “traditional” industry. Suddenly everybody has discovered that the future runs on our balance sheet.

And beneath every clean product demo, every AI wrapper, every shiny forecast of abundance, there is still a very old question waiting in the dark:

Yes, but where will the power come from?

And somewhere, an energy developer who has not slept properly in six years whispers back:

Finally. A question I know how to answer.

Now if you’ll excuse me, I have a Steam Purchase Agreement to finalise and a CHP plant in Abidjan to dispatch-model. The future, it turns out, is not waiting for permission. It also doesn’t take Sundays off, which I’m choosing not to think about.

- Kay

P.S. If you are a JPMorgan SRI team member who wandered into this post by accident: hi. The third pillar of your initiative has a natural extension to the African C&I market that I suspect nobody on your 30-person team has properly socialised yet. My DMs are open. My coffee order is reasonable. I will even let you expense it.

P.P.S. I tried to count how many times Dimon used the word “fortress.” I gave up at fourteen. The man loves a fortress. The Normans would have approved.

P.P.P.S. Also, “America” appears 89 times. “Africa” appears zero. I checked. We have work to do. Some of that work is writing better blog posts. Some of it is getting somebody at 270 Park Avenue to read this one. If you know somebody, forward it. I’ll owe you a jollof.

Standard disclaimers: I do not work for JPMorgan and they did not pay me to write this. I do work for a company that builds the things in Dimon’s third pillar, so in the grand tradition of this blog I am absolutely talking my own book and have the decency to admit it. Past performance is not indicative of future results, but past Jamie Dimon letters are extremely indicative of future Jamie Dimon letters. If I’m wrong about anything here, I warned you. If I’m right, please remember my consulting rates are very reasonable and I take payment in jollof, project finance gigs, or both.

Subscribe now

The Strait of Hormuz, the narrow passage through which roughly a quarter of the world’s seaborne oil trade flows ^[2], has been functionally closed since the US and Israel attacked Iran on February 28, 2026 ^[3]. The IEA’s Fatih Birol said it plainly: “Today, we are losing 12 million barrels per day ^[4], surpassing the combined losses of the 1973 and 1979 oil crises.” Qatar has declared force majeure on its LNG exports ^[5]. Saudi Aramco’s Ras Tanura terminal has shut down ^[6]. US gas prices hit $4 a gallon by the end of March ^[7]. Goldman Sachs has modelled a severely adverse scenario reaching $160 ^[8] a barrel. Wood Mackenzie and Macquarie analysts are warning of scenarios approaching $200 ^[9].

Africa is getting hammered.

Ghana raised petrol prices 15% ^[10]. Tanzania, 33% ^[11]. Gambia, nearly 19% ^[12]. Botswana, Mali, South Africa, all scrambling. The World Economic Forum noted ^[13] what should be obvious: poorer fuel-importing states in Africa and Asia cannot absorb these shocks the way wealthier nations can. For them, the pain arrives as higher household prices, fiscal strain, and a greater risk of rationing or unrest.

The 2026 spike is a cyclical shock. Africa’s dependence on imported fuels is structural.

And this is happening against a backdrop where the US president pulled America out of the Paris Climate Agreement for the second time in January 2026 ^[14], gutted the IRA’s solar and wind credits through the One Big Beautiful Bill ^[15], and made it clear that American climate finance for Africa is not coming. When the US steps back from Paris and hollows out the IRA, it is not symbolism. It removes one of the largest prospective anchors for concessional climate capital that African DFIs were counting on. The EU, meanwhile, is watching its own gas and oil routes destabilise and diverting money into domestic resilience: storage, interconnectors, renewables at home. Africa is last in that queue.

I am sitting in Lagos, watching diesel generators rumble to life across every industrial estate in the city, burning fuel that now costs more than it did last month, which cost more than the month before, and thinking: Africa holds roughly 40% of the world’s solar potential. How are we still this dependent on a shipping lane in the Persian Gulf?

The answer is not technology. Solar PV is the cheapest source of new electricity generation ^[16], on a levelised cost basis, in most African countries. It is not demand. Around 80% of Nigerian companies cite electrification challenges ^[17] as their most significant obstacle to doing business. It is not even money, in the aggregate. Clean energy investment on the continent reached approximately $40 billion in 2024 ^[18], more than double the 2019 figure, in a year when global clean energy investment hit $2 trillion.

The answer is market structure.

Whether this war lasts six months or three years is almost secondary to what it reveals about the way we have built our markets. And for 600 million Africans without grid access, the same states most exposed to fuel price spikes are also most exposed to climate impacts. That doubles the cost of delay.

Let me be precise about the problem this series is addressing. African renewable energy developers, the small and mid-sized companies that originate, develop, finance, build, and operate solar projects, mini-grids, and distributed energy systems, are dying. Not in dramatic bankruptcies. In the slow bleed of running out of working capital while waiting for projects to close. The market structure around them is broken in specific, diagnosable, fixable ways. And every other industry that faced the same structural problem eventually found the same kind of answer.

This article shows how the current architecture kills developers. Part 2 looks at industries that fixed the same problem. Part 3 sketches what we need to build now.

* * *

The Conversation That Never Changes

I have been having the same conversation for three years.

Different rooms. Different cities. Lagos, Accra, Casablanca, Abidjan, sometimes Nairobi. Different developers sitting across from me. But the shape of the conversation is always the same, almost word for word, like we are all reading from a script that nobody wrote down.

A developer has a pipeline. Three or four C&I solar projects, maybe a biomass opportunity. One of them is real: a real industrial client, a real energy load, real diesel bills eating into margins that were already thin before Hormuz closed. The developer has done a site visit. They have a relationship that took eighteen months to build.

Then the conversation stalls.

Not because the economics are bad. Diesel at $0.35/kWh against a solar PPA at $0.10 was already compelling. At $110 oil, with diesel heading north of $0.50/kWh in some markets, it is an emergency. Not because the technology is uncertain. Not because demand is absent.

The conversation stalls because the developer cannot get the project financed. And they cannot get it financed because they cannot prove bankability. And they cannot prove bankability because they have no track record. And they cannot build a track record because they cannot close their first project. Eighteen to thirty-six months from first conversation to commissioning is normal for a 1 to 5 MW C&I project. Almost no local developer is capitalised for that kind of runway.

This is the classic SME “missing middle” problem, dressed up in kilowatts and PPAs. The SME financing gap in Sub-Saharan Africa is estimated at $331 billion ^[19]. In African climate tech specifically, EchoVC Partners found that less than 10% of all deals fall in the $250,000 to $1 million range, the exact band where most ventures need capital to transition from proof-of-concept to commercial readiness. Across all African climate tech, just 10 companies captured more than half of all capital.

Even the IFC’s own Scaling Solar initiative, designed specifically to solve this problem with DFI backing and standardised frameworks, failed to scale in Africa ^[20], per a 2023 Devex investigation. If even a well-resourced IFC initiative could not build the rails, that tells you the problem is structural, not operational.

* * *

What the Rest of the Nigerian Economy Understood

Let me tell you about companies that solved a version of this problem.

In 2015, a company called TeamApt started building backend banking software for other people’s banks. Nobody noticed. It spent four years as a plumbing company, generating revenue from invisible infrastructure work. When they started, POS terminals were an afterthought product for incumbent banks. Then TeamApt pivoted. Rebranded as Moniepoint. Started deploying those terminals to market women in Lagos, Ibadan, Kano. By 2024, Moniepoint was processing over ₦412 trillion in transactions ^[21], handling roughly 80% of in-person payments in Nigeria ^[22]. TIME named it one of the world’s 100 most influential companies ^[23]. Valuation: over $1 billion.

Moniepoint did not start by trying to be a unicorn. It started by generating revenue from infrastructure nobody else wanted to build. It is the foundry concept in action: shared infrastructure enabling distributed operators.

Aliko Dangote built a nearly $20 billion refinery ^[24] in Ibeju-Lekki, not because Nigeria had a supportive regulatory environment for refining. Nigeria exported crude and imported refined fuel for decades. Dangote built it anyway. He vertically integrated an entire supply chain: port, pipeline, crude import deals, downstream distribution. Because the shared infrastructure for each of those functions did not exist. That is what the absence of a foundry costs at scale. No solar developer can spend $20 billion building their own institutional infrastructure.

By 2025, the Dangote Refinery was supplying 18 million litres of gasoline per day ^[25]. When the Iran war hit and European countries started buying aviation fuel from Lagos for the first time in history, the bet looked less like ambition and more like prophecy. The expansion to 1.4 million barrels per day, partnering with Honeywell ^[26], targets 2028.

Or look at Rensource. Ademola Adesina started it in 2015 as a power-as-a-service company for Nigerian market traders. When the pandemic halted that business, the team built Sabi, a B2B marketplace for informal merchants, which hit $1 billion in annualised GMV ^[27] and raised $66 million at a $300 million valuation. Sabi then pivoted again into traceable mineral supply chains through its TRACE platform ^[28], helping Africa capture more value from its critical mineral reserves. This is what a foundry-thinking company does: it builds shared infrastructure that enables an entire sector, then adapts the infrastructure as sectors evolve.

Nigeria’s top nine fintechs are worth a combined $10.6 billion ^[29]. Flutterwave at $3 billion. OPay at $2.75 billion. Moniepoint at $1 billion. Paystack, acquired by Stripe for more than $200 million ^[30], just created The Stack Group ^[31], a holding company with a payments unit, a microfinance bank, a consumer app, and a venture arm. In telecoms, towercos like IHS and Helios took network infrastructure off operator balance sheets so everyone could focus on customers and spectrum instead of steel and concrete. In data centres, MainOne and Rack Centre built the pipes so everyone else could sell services.

In the AI industry, OpenAI, Anthropic, Mistral, and others built foundry infrastructure first, letting the market form around it. Meta open-sourced Llama. Amazon invested $8 billion in Anthropic ^[32]. They were building foundries, not competing for individual chip designs.

Here is what Moniepoint, Dangote, Sabi, and OpenAI share: they built market infrastructure rather than competing within a broken market. The renewable energy sector has no equivalent. No shared rails. No common standards. No foundry. Every developer is building their own refinery from scratch, one project at a time.

This is not a technology problem. It is an architecture problem.

* * *

The Wrong Diagnosis: Bankability Versus Architecture

The dominant narrative in African energy finance is that African projects are not bankable. DFIs repeat it. Development banks repeat it. Lenders who spent thirty minutes with a project information memorandum repeat it.

It is the wrong diagnosis.

Bankability is a symptom, not the disease. The disease is a capital stack mismatch so severe that even good projects cannot find the right money at the right stage of development.

Think of the capital stack as four distinct risk buckets: idea risk, development risk, construction risk, and operating risk. In African clean energy, buckets one and four are almost empty, while three is over-subscribed.

• Stage One: Origination and Pre-Development. Tens of thousands of dollars over 6 to 18 months. Patient capital that tolerates high failure rates. Almost entirely absent. DFI grants usually pay consultants and studies, not payroll. They de-risk assets for lenders but do not extend the runway for the developers creating those assets.

• Stage Two: Late Development. Low hundreds of thousands over 12 to 24 months. Equity that funds legal, environmental, technical, and financial costs. Scarce and expensive. The developer gives away too much equity to access too little money.

• Stage Three: Financial Close. Multi-million dollar project finance debt. Available from DFIs and commercial lenders for projects with credible offtakers. This is the capital that shows up in press releases. Traditional project finance is optimised for de-risked, late-stage assets. It was never designed to fund origination at scale in fragmented markets.

• Stage Four: Company Capital. The most absent capital of all. Recurring operating lines. The entire ecosystem finances projects, not companies. But companies die between closings, not during them.

Here is the number that makes this concrete. Africa’s average weighted cost of capital for renewable energy projects is 15.6%, versus 4.2% in Western Europe ^[33], per the Clean Air Task Force. For Nigeria specifically, the WACC for renewable energy is 25 to 31%, per the Climate Policy Initiative ^[34]. An Ortelius analysis made the point viscerally: the identical solar park that costs €86/MWh to build in Europe at 5% WACC costs €135/MWh in Africa at 12% WACC ^[35].

The sun is the same. The technology is the same. Only the financing cost is different.

And then there is the currency mismatch. Most projects are financed in dollars or euros but generate revenue in naira, cedis, or shillings. The Energy for Growth Hub quantified it: currency mismatch alone adds 5 to 6 percentage points ^[36] to the effective cost of capital and pushes the LCOE for utility-scale solar in Africa to 10 to 15 cents per kilowatt-hour. Two to three times higher than in Europe or Asia. Not because the sun is weaker. Because the financial architecture is broken.

Africa’s solar installations fell from 3.07 GW in 2023 to 2.4 GW in 2024 ^[18], in a year when global clean energy investment hit $2 trillion. Clean energy investment on the continent reached approximately $40 billion in 2024 ^[18], roughly 2 to 3% of the global total, against an estimated need of $200 billion per year ^[37].

This is a textbook case of coordination failure: no single actor can justify investing in early-stage pipelines or shared standards, so everyone waits for someone else to move first.

* * *

Who Survives When Architecture Is Missing

Look at who actually survived in this market. And look at what they have in common.

Daystar Power is Shell ^[38]. Not Shell-backed. Shell. Acquired outright, operating with an oil major’s balance sheet. Offering C&I solar PPAs at half the grid price, recently announcing its expanded footprint in the Agbara industrial estate, pushing deeper into Nigeria’s manufacturing heartland.

BECS, is backed by Berkeley Energy, which manages billions of dollars renewable energy funds across Africa and Asia. Starsight recently raised $15 million in mezzanine funding from British International Investment ^[39] on top of its AIIM/Helios equity backing. Empower New Energy, backed by Climate Fund Managers (a Dutch-South African entity) and Norfund (Norway’s development finance institution) ^[40], just financed a JustRite Superstores solar installation across multiple locations. Husk Power, backed by Shell, Engie, and IFC, is raising toward an IPO targeted for 2027 ^[41].

Konexa is incubated by Shell Foundation, Rockefeller Foundation, DFID, and USAID ^[42], with MIGA guarantees. UBA recently did a solar deal with Renewvia.

The common denominator is not technology or execution. It is access to cheap, patient, foreign balance sheets that reprice risk and smooth volatility. In practice, this means control over African power assets is quietly migrating to balance sheets in London, Oslo, New York, and The Hague, even when the assets sit in Lagos or Accra. A local developer raising naira-denominated working capital at mid-teens effective interest rates simply cannot match Shell’s dollar cost of capital on a 15-year PPA.

But the market is not closed. Daystar’s own feasibility analysis identified 170,000 C&I businesses ^[43] in Nigeria as potential solar customers. They serve a few thousand. Less than 2% penetration. You know who else left 98% of their addressable market untouched? The big Nigerian banks. For decades, First Bank, UBA, and GTBank served maybe 30 million Nigerians. Then OPay, Moniepoint, and PalmPay raced to sign up the other 170 million.

The renewable energy version of this play: projects below 2 MW, regional manufacturers in tier-two cities, biomass CHP, industrial process heat, hybrid systems. A developer who understands the energy profile of a cassava processing plant is competing on knowledge, not cost of capital. The foreign-backed platforms are not even looking for those deals.

This is not a technology problem. It is an architecture problem.

* * *

The Iran Factor, Trump, and Three Triggers

The Iran war does three things to this picture simultaneously.

First, it makes the economic case for solar undeniable at the C&I level. Jeff Currie, Chief Strategy Officer ^[44] of Energy Pathways at Carlyle and a former Goldman Sachs energy analyst with a 25-year track record, said at CERAWeek: “We are going to get the energy transition forced on us in a very painful way.”

Subsidy removal in May 2023 flipped the economics for households. Grid collapses in 2024 forced SMEs to scramble. The Iran war hits C&I balance sheets directly. Three triggers, three customer segments.

Second, it exposes the structural absurdity of Africa’s energy position. Nigeria is a crude oil exporter and a refined fuel importer. Dangote Refinery is now bailing out European airlines with aviation fuel from Lagos. But Nigeria cannot deploy solar at scale on its own factory roofs. The country with the refinery that European airlines are scrambling to buy fuel from cannot figure out how to finance a 2 MW solar installation in Ibadan.

Third, and this is the part the clean energy community has not processed yet: the US is not coming to help. Trump has made that explicit. The EU is diverting capital into domestic resilience before thinking about Africa. Every assumption about where the money for African energy transition was supposed to come from is being rewritten. Yet the Iran war also creates a window: DFIs and European governments are now prioritising energy resilience as a geopolitical imperative. African distributed solar is not just a development play anymore. It is strategic infrastructure. That reframing changes who the investors are.

* * *

The Solar Import Trap and the Manufacturing Gap

Nigeria imported ₦435.52 billion worth of solar panels ^[45] in 2025, with 71.38% from China ^[46]. REA’s managing director, Abba Aliyu, put a number on the problem: over ₦200 billion spent annually importing PV panels. “We want to reverse that trend,” he said.

This is energy’s version of the old commodity trap: export raw materials, import finished goods, and let currency swings determine who survives. We are replacing dependence on Gulf oil with dependence on Chinese manufacturing.

In October 2025, the Minister of Power, Adebayo Adelabu, hosted the inaugural Nigerian Renewable Energy Innovation Forum at the Transcorp Hilton, themed “Implementing the Nigeria First Policy.” Vice President Shettima opened it. Nearly $500 million ^[47] in announced or committed investment deals were signed for solar panel assembly, battery manufacturing, and local component production. Adelabu declared that Nigeria is on track for nearly 4 GW per annum of solar manufacturing capacity ^[47]. By October, Nigeria had begun exporting locally manufactured solar panels to Ghana ^[48].

Then Adelabu resigned to contest the Oyo State governorship election. The minister who was championing local manufacturing is now running for governor. The policy momentum he created has yet to find its next champion.

The Dufil Group, makers of Indomie noodles, understood something about local manufacturing that the solar industry has not figured out. When Dufil entered Nigeria, it built factories. It localised production. Flour Mills of Nigeria followed the same playbook. Dangote followed it in cement, sugar, refining. The solar industry has no equivalent. No local panel assembly at meaningful scale. Every project is a dollar-denominated import transaction in a naira economy. The naira went from approximately ₦460 to the dollar in early 2023 to ₦1,535 by end-2024 ^[49], a collapse in dollar purchasing power of more than 70%. Every solar panel got proportionally more expensive.

If a Tier-1 platform is landing panels at 18 to 20 cents per watt and a small developer pays 23 to 25 cents, your project-level LCOE gap is locked in before you design anything.

For comparison: India installed 24 GW of solar in 2024 alone. Nigeria installed 803 MW in 2025 ^[50], its strongest year ever, and was celebrated as Africa’s second-largest market. The gap is not resource or ambition. It is architecture: India has IREDA, SECI, standardised auctions. Nigeria does not.

* * *

The EV Boom’s Connection to Solar

While the energy sector argues about bankability, another transition is happening on Nigerian roads.

Between 15,000 and 20,000 electric vehicles ^[51] now drive in Nigeria. SAGLEV has a dedicated EV assembly plant ^[52] in Imota, Lagos, scalable to 10,000 units per year. Spiro has deployed over 60,000 electric motorbikes ^[53] with more than 1,200 battery swap stations across Africa. Foltï Technologies launched eDryv in Lagos ^[54], Nigeria’s first ride-hailing platform powered by solar-charged EVs. Ecowaka, led by Prince Ojeabulu ^[55], is manufacturing electric keke napeps starting at ₦2.6 million. The Federal Executive Council approved ₦58 billion in December 2025 for 200 electric buses ^[56]. Nigeria signed a deal with South Korea for an EV manufacturing plant ^[57] targeting 300,000 vehicles annually.

My brother, mentor, and former colleague John Okoro, co-founder and Managing Director of Growth Energy (part of the Solio Group, headquartered in France), recently commissioned the largest solar-powered EV charging station in East Africa ^[58], in Burundi. Growth Energy operates across Nigeria, Burundi, Tanzania, Zanzibar, and Kenya. John’s work demonstrates something profound: the EV-solar convergence is not a future possibility. It is being built right now.

Every EV is a mobile battery. In a grid with chronic outages, that is not just a transport asset. It is potential backup storage. For a solar developer, EV charging is exactly the kind of predictable, high-load, daytime-skewed demand profile that makes mini-grids and C&I systems bankable. Yet EV and power regulators rarely talk. There is no standard tariff, no integrated planning, almost no concessional capital for solar-powered charging corridors.

The EV boom and the solar boom are the same transition, being discussed in separate conference rooms by separate industries with separate investors. That is the architecture problem in miniature.

* * *

The Ecosystem That Is Actually Forming

This is what a proto-foundry looks like: not one big company, but many specialised firms standardising messy bits of the value chain.

On February 23, 2026, the Rural Electrification Agency signed an MoU with Lotus Bank unlocking a ₦100 billion revolving credit facility ^[59] for mini-grid developers participating in the DARES programme. Up to ₦8 billion per developer. 18-month tenor tailored to the real construction cycle of sub-5 MW projects. A Nigerian bank providing project-level lending at the ticket sizes developers actually need. That is architecture changing.

Sterling Bank secured a $13 million facility from the Universal Green Energy Access Programme ^[60], managed by Cygnum Capital, to expand renewable energy lending in Nigeria. Viathan, Nigeria’s leading embedded energy company, launched Decentralised Energy Limited with $10 million in equity from the Anergi Group ^[61] and ₦8.5 billion in debt, backed by an investor community that has committed over $100 million to Viathan over the past decade. DEL is an attempt to turn one company’s investor relationships into a platform. A mini-foundry.

Rivy, a climate finance platform, has disbursed over ₦43 billion and approved 38,000 loans ^[62]. Then Rivy registered solar mini-grids for International Renewable Energy Certificates, increasing the supply of RECs from solar assets in Nigeria by 25% within a month. I-RECs turn each MWh of clean electricity into a tradable certificate. That extra revenue can raise project IRRs by 1 to 3 percentage points on small systems. Solad Integrated Power, running solar mini-grids at Iponri market in Lagos, sold its first I-RECs through Rivy’s platform, with backing from SEforALL. Carbon finance making mini-grid economics work.

SunFi, founded by Rotimi Thomas, has deployed over 1,500 solar systems managing over 4 MW ^[63] of assets, connecting Nigerians to solar through financing plans. Uwana Energy is connecting households to solar loans. Fixr, at usefixr.com, is building the installation and maintenance services layer.

InfraCredit Nigeria announced a credit guarantee for Africa’s first solar panel assembly plant ^[64], combining local currency instruments with Bank of Industry concessional capital. This attracted Nigerian pension funds and insurance companies: domestic institutional capital that previously could not participate in energy infrastructure. That is landmark. Not because of the deal size, but because it proves that local currency guarantees can unlock domestic institutional capital for renewable energy manufacturing.

The IFC and AfDB launched Zafiri in October 2025 ^[65], a dedicated equity vehicle for distributed energy companies, specifically addressing the missing middle. Their own framing: a “missing middle” problem requiring “long-term equity to these providers.” The Africa Mini-Grid Developers Association, under its new leadership, continues the coalition-building work that individual developers cannot: standardising documentation, aggregating policy advocacy.

A taxonomy of what is forming: financing rails (Rivy, SunFi, Uwana, Lotus Bank), services rails (Fixr, installers), policy rails (AMDA, REA), risk-mitigation rails (I-RECs, carbon finance, InfraCredit guarantees). The pieces exist. The integrated system does not.

* * *

Three Regulatory Shifts That Change the Terrain

NBET becomes NENEX. The Federal Government approved NBET’s transition to an energy commodity exchange modelled on India’s IEX ^[66], which cleared over 100 billion kWh in FY2024 with more than 8,100 participants ^[67] including over 2,100 renewable generators. In practice, a 1 to 10 MW solar plant near an industrial cluster can list on the exchange and sell to multiple buyers instead of begging one anchor offtaker for a 15-year PPA.

The Electricity Act 2023 devolution. Twelve Nigerian states have enacted their own electricity acts, with three fully taking charge. LASERC has assumed full regulatory authority ^[68] in Lagos. This decentralisation is a double-edged sword: more regulators can mean more market access, but also more fragmentation. Regulatory resets are painful, but they are also rare windows where nimble players can grab licences and shape rules. President Tinubu signed the Electricity Act Amendment 2025 in February 2026 ^[69], further clarifying state powers.

DARES at scale. Nigeria’s $750 million Distributed Access through Renewable Energy Scale-up ^[70] programme, targeting 17.5 million Nigerians through solar mini-grids and standalone systems, is part of Mission 300, the World Bank/AfDB joint programme to connect 300 million Africans to electricity by 2030. The first DARES disbursements were enabled by the Lotus Bank MoU. That MoU is already performing foundry functions.

Regulation is quietly shifting from vertically-integrated monopolies to market-platform logic. Developers who think like infrastructure-as-a-service providers will win.

* * *

The Question

So here is what I keep coming back to.

Nigeria has fintech companies worth a combined $10.6 billion ^[29]. It has the Dangote Refinery expanding to become among the largest on earth, exporting refined fuel during a Middle Eastern war. It has 20,000 EVs on its roads and a local assembly plant. It has solar panel exports to Ghana and carbon finance platforms registering I-RECs from market mini-grids.

And yet. The renewable energy sector, in a country where the sun beats down 300 days a year, cannot reliably finance a 2 MW rooftop installation.

The foundry problem is that we have developers and demand, but no shared infrastructure that turns bespoke projects into repeatable products.

What would the foundry be? Standardised PPA templates that any developer can use without commissioning bespoke legal work. A credit database that converts a developer’s five-year operational track record into a bankable risk profile. A local-currency refinancing vehicle that allows a developer who built a project on expensive dollar debt to refinance in naira once the asset is proven. A technical standards body whose sign-off is credible to international lenders without each developer hiring their own international advisory team.

This is not new. This exact dynamic has played out in semiconductors, Cleantech 1.0, Nigerian oil and gas, and telecoms. Every time, the industry eventually found a structural answer.

Part 2 of this series opens those history books. We will ask what an African TSMC for power might look like: a platform that standardises contracts, capital, and components so developers can specialise in origination and operations.

The war in the Gulf will not last forever. Oil prices will stabilise. But the structural vulnerability, the fact that the continent with the most solar potential on earth is still dependent on a shipping lane 8,000 kilometres away, will persist until somebody builds the infrastructure that makes it unnecessary.

The foundry problem is not about money. It is about architecture.

And the clock is ticking.

* * *

P.S. And no, I am not entirely sure that writing a blog post about solar financing that includes electric tricycles, the Strait of Hormuz, Indomie noodles, and OpenAI in the same argument is evidence of coherent thinking. But the connections are real. The structural parallels are real. And if drawing the lines between them makes even one small developer rethink their approach on Monday morning, then these were worth it.

* * *

Kay is Manager for West Africa at Berkeley Energy Corporate Solutions, where he structures and finances renewable energy projects across 16 countries in West and Central Africa. He is a PhD researcher in multi-physics modelling of solar-biomass-hydrogen hybrid systems at the University of North Dakota, and writes The Impostor’s Guide to Clean Energy at kaykluz.com. The views expressed here are his own and do not represent the views of his employer.

* * *

References

[1] Fortune, Oil Price Tracker, April 3, 2026. https://fortune.com/article/price-of-oil-04-03-2026/

[2] U.S. Energy Information Administration, Strait of Hormuz. https://www.eia.gov/todayinenergy/detail.php?id=65504

[3] Britannica, 2026 Iran War. https://www.britannica.com/event/2026-Iran-war

[4] CNBC, IEA’s Fatih Birol, April 1, 2026. https://www.cnbc.com/2026/04/01/oil-price-iea-fatih-birol-brent-iran-strait-hormuz.html

[5] Reuters, QatarEnergy Force Majeure, March 4, 2026. https://www.reuters.com/business/energy/qatarenergy-declares-force-majeure-lng-shipments-2026-03-04/

[6] Reuters, Saudi Aramco Ras Tanura, March 2, 2026. https://www.reuters.com/business/energy/saudi-aramco-shuts-ras-tanura-refinery-after-drone-strike-source-says-2026-03-02/

[7] CNN, US Gas Prices Hit $4, March 31, 2026. https://www.cnn.com/2026/03/31/business/us-gas-prices-usd4-intl

[8] Fox Business, Goldman Sachs Oil Scenarios, March 2026. https://www.foxbusiness.com/economy/iran-war-could-push-inflation-higher-year-goldman-sachs-says

[9] Al Jazeera, Could Oil Hit $200?, March 19, 2026. https://www.aljazeera.com/news/2026/3/19/could-oil-hit-200-a-barrel-analysts-no-longer-think-its-far-fetched

[10] GhanaWeb, Fuel Price Increase, April 2026. https://www.ghanaweb.com/GhanaHomePage/business/Petrol-diesel-prices-to-rise-by-15-from-April-2026-Benjamin-Nsiah-2026861

[11] The Chanzo, Tanzania Fuel Prices, April 1, 2026. https://thechanzo.com/2026/04/01/tanzanias-fuel-prices-surge-to-record-highs-as-middle-east-war-drives-global-oil-crisis/

[12] The Point, Gambia Fuel Prices, April 2026. https://thepoint.gm/africa/gambia/headlines/fuel-pump-prices-increase-in-gambia

[13] World Economic Forum, Global Price Tag of War, March 2026. https://www.weforum.org/stories/2026/03/the-global-price-tag-of-war-in-the-middle-east/

[14] New York Times, US Exits Paris Agreement, January 27, 2026. https://www.nytimes.com/2026/01/27/climate/paris-climate-agreement-withdrawal.html

[15] PV Tech, Renewable Energy Credits in One Big Beautiful Bill. https://www.pv-tech.org/us-renewable-energy-credits-face-steep-cliff-edge-in-one-big-beautiful-bill/

[16] PV Tech / IRENA, Renewable Power Generation Costs 2024. https://www.pv-tech.org/irena-global-solar-pv-lcoe-increases-by-0-6-in-2024-to-us0-043-kwh/

[17] World Bank Enterprise Surveys, Nigeria 2025. https://www.worldbank.org/en/programs/enterprise-surveys

[18] Global Solar Council, Africa Market Outlook 2025–2028. https://www.globalsolarcouncil.org/news/global-solar-council-africas-solar-market-set-to-surge-42-in-2025-but-finance-bottlenecks-threaten-growth/

[19] MIT Sloan, Financing Africa’s Missing Middle, November 2024. https://mitsloan.mit.edu/centers-initiatives/ksc/responsibly-financing-africas-missing-middle

[20] Devex, Why a Major IFC Solar Initiative Failed, May 2023. https://www.devex.com/news/why-a-major-ifc-solar-initiative-failed-to-scale-in-africa-105513

[21] BusinessDay Nigeria, Moniepoint 2025 Year in Review. https://businessday.ng/technology/article/breaking-as-banks-pull-back-moniepoints-n1trn-loans-fill-nigerias-sme-credit-gap/

[22] TechCrier, Moniepoint 80% In-Person Payments, January 2026. https://www.techcrier.com/2026/01/moniepoint-processes-80-percent-of.html

[23] TIME, TIME100 Most Influential Companies 2025. https://time.com/collections/time100-companies-2025/7289615/moniepoint/

[24] Wikipedia, Dangote Refinery. https://en.wikipedia.org/wiki/Dangote_refinery

[25] Premium Times, Dangote Refinery Output, December 2025. https://www.premiumtimesng.com/news/top-news/841988-dangote-refinery-supplied-23-5-million-litres-of-petrol-daily-in-november-nmdpra.html

[26] Honeywell, Dangote Expansion Partnership, November 2025. https://www.honeywell.com/us/en/press/2025/11/honeywell-to-help-dangote-double-production-capabilities-at-africa-s-largest-refinery

[27] TechCrunch, Sabi $300M Valuation, May 2023. https://techcrunch.com/2023/05/19/african-b2b-e-commerce-startup-sabi-tops-300m-valuation-in-new-funding/

[28] Condia, Sabi TRACE Platform Pivot, October 2025. https://thecondia.com/sabi-minerals-pivot-trace-platform/

[29] Nigeria Communications Week, Fintech Valuations, January 2026. https://www.nigeriacommunicationsweek.com.ng/nigerias-9-top-fintech-firms-valued-at-10-6bn-in-january-2026/

[30] TechCrunch, Stripe Acquires Paystack, October 2020. https://techcrunch.com/2020/10/15/stripe-acquires-nigerias-paystack-for-200m-to-expand-into-the-african-continent/

[31] FF News, Paystack Launches The Stack Group, January 2026. https://ffnews.com/newsarticle/paytech/paystack-launches-holding-company-the-stack-group-tsg/

[32] Amazon, Investment in Anthropic, November 2024. https://www.aboutamazon.com/news/aws/amazon-invests-additional-4-billion-anthropic-ai

[33] Clean Air Task Force, WACC in African Power Sector, 2024. https://www.catf.us/resource/evaluating-weighted-average-cost-capital-wacc-power-sector-african-countries/

[34] Climate Policy Initiative, Cost of Capital for Renewable Energy, 2023. https://www.climatepolicyinitiative.org/wp-content/uploads/2023/06/Discussion-Paper-EM-Cost-of-Capital-for-RE-and-GCGF-FINAL-Jun-2023.pdf

[35] Ortelius, Cost of Capital Hits Developing Countries Hardest. https://ortelius.be/cost-of-capital-hits-developing-countries-the-hardest-in-their-green-transition/

[36] Energy for Growth Hub, Local Currency Advantages, June 2025. https://energyforgrowth.org/article/the-quantifiable-advantages-of-local-currency-on-african-energy-projects/

[37] Global Solar Council, Africa Market Outlook for Solar PV 2025–2028. https://www.globalsolarcouncil.org/resources/africa-market-outlook-for-solar-pv-2025-2028/

[38] Shell, Daystar Power Acquisition, December 2022. https://www.shell.com/what-we-do/renewable-power/renewable-power-news-releases/shell-completes-acquisition-of-solar-solutions-provider-daystar-power-group.html

[39] Africa Newsroom, Starsight / BII $15M, March 2026. https://www.africa-newsroom.com/press/starsight-energy-africa-group-starsight-partners-with-british-international-investment-bii-to-advance-clean-energy-growth-in-west-africa-through-us15-million-mezzanine-funding

[40] Empower New Energy, CFM/Norfund Partnership, December 2022. https://www.empowernewenergy.com/post/empower-new-energy-partners-with-climate-fund-managers-and-norfund-to-accelerate-solar-investments

[41] Yahoo Finance, Husk Power IPO Plans, January 2025. https://finance.yahoo.com/news/husk-power-targets-400m-funding-121530024.html

[42] Shell Foundation, Konexa, December 2019. https://shellfoundation.org/news/can-an-integrated-energy-distribution-model-turn-around-africas-ailing-utilities/

[43] Daystar Power / RMI, Feasibility Study, March 2024. https://daystar-power.com/en/news-article?post_id=176354

[44] AOL Finance / S&P Global, CERAWeek, March 2026. https://www.aol.com/finance/shift-oil-isnt-just-being-120741898.html

[45] Punch Newspapers, Solar Panel Imports, March 2026. https://punchng.com/solar-panel-imports-hit-2-9m-amid-outages/

[46] PVKnowhow, Nigeria Solar Panel Imports, October 2025. https://www.pvknowhow.com/news/nigeria-solar-panel-imports-impressive-n242-68-billion-in-2025/

[47] The Electricity Hub, Nigeria $500M Solar Manufacturing, October 2025. https://theelectricityhub.com/nigeria-secures-500m-for-solar-manufacturing-plants/

[48] Guardian Nigeria, Solar Panel Exports to Ghana, October 2025. https://guardian.ng/energy/nigeria-now-exporting-solar-panels-to-ghana-says-adelabu/

[49] Nairametrics, Exchange Rate 2024. https://nairametrics.com/2024/12/31/exchange-rate-ends-2024-at-n1535-1-marking-a-40-9-depreciation/

[50] PV Tech / Global Solar Council, Africa Record Solar Capacity 2025. https://www.pv-tech.org/gsc-africa-adds-record-4-5gw-new-solar-pv-capacity-2025/

[51] Climate Scorecard, Nigeria EVs, March 2025. https://www.climatescorecard.org/2025/03/nigeria-has-pledged-to-achieve-100-zero-emission-sales-for-new-cars-and-vans-by-2040/

[52] SAGLEV, EV Assembly Plant, May 2025. https://saglev.com/saglev-targets-10000-electric-vehicles-annual-output/

[53] TechCrunch, Spiro $100M Raise, October 2025. https://techcrunch.com/2025/10/21/spiro-raises-100m-the-largest-ever-investment-in-africas-e-mobility/

[54] TechCabal, eDryv Solar EV Ride-Hailing, April 2025. https://techcabal.com/2025/04/07/edryv-nigerias-first-95-green-powered-ev-ride-hailing-service-unveils-in-lagos/

[55] Ecowaka, Electric Keke Launch, July 2025. https://ecowaka.io/ecowaka-launches-three-wheeled-electric-vehicles-to-boost-transportation-in-nigeria/

[56] Guardian Nigeria, FEC Electric Buses, December 2025. https://guardian.ng/news/fec-endorses-new-industrial-policy-okays-electric-buses-boi-hq-others/

[57] Business Insider Africa, Nigeria–South Korea EV Deal, February 2026. https://africa.businessinsider.com/local/markets/nigeria-signs-deal-with-south-korea-to-launch-africas-first-electric-vehicle-factory/xe79vqe

[58] Growth Energy, Burundi EV Charging Station, January 2026. https://growth-energy.fr/building-africas-largest-solar-powered-ev-charging-station-in-burundi-growth-energy-gem/

[59] The Nation, REA–Lotus Bank ₦100B Deal, February 2026. https://thenationonlineng.net/rea-bank-strike-n100b-energy-deal/

[60] Cygnum Capital, UGEAP / Sterling Bank $13M, March 2026. https://www.cygnumcapital.com/news/ugeap-provides-usd-13-million-facility-to-sterling-bank-to-promote-renewable-energy-lending-in-nigeria

[61] BusinessDay, Viathan / DEL Launch, March 2026. https://businessday.ng/companies/article/decentralised-energy-secures-10m-initial-equity-funding-from-anergi-group-to-kick-start-operations/

[62] TechCabal, Rivy Carbon Finance Platform, October 2025. https://techcabal.com/2025/10/13/rivy-is-growing-the-carbon-market-in-nigeria/

[63] The Realistic Optimist, SunFi Profile, May 2025. https://www.realisticoptimist.io/sunfi-managed-solar-marketplace-in-nigeria/

[64] Energy Capital Power, InfraCredit Solar Manufacturing Guarantee, March 2026. https://energycapitalpower.com/infracredit-to-issue-africas-first-solar-panel-manufacturing-guarantee/

[65] IFC, Zafiri Launch, October 2025. https://pressroom.ifc.org/all/pages/PressDetail.aspx?ID=27872

[66] ThisDay Live, NBET Energy Exchange, March 2024. https://www.thisdaylive.com/2024/03/04/like-india-nbet-begins-process-of-transmuting-to-energy-exchange-to-revamp-nigerias-electricity-market/

[67] Indian Energy Exchange, Official Website.

https://www.iexindia.com

[68] Templars Law, LASERC Regulatory Authority, July 2025. https://www.templars-law.com/knowledge-centre/new-lagos-electricity-order-what-it-means-for-current-operators-and-potential-investors/

[69] Daily Intel, Electricity Act Amendment, February 2026. https://www.dailyintelnewspaper.com/beyond-the-grid-what-the-new-electricity-act-really-changes/

[70] Solar Quarter, Nigeria DARES Programme, March 2026. https://solarquarter.com/2026/03/20/nigerias-dares-project-to-deliver-clean-power-to-17-5-million-people-with-world-bank-support/

I woke up for a scheduled call (US time zones are undefeated at ruining sleep patterns) and couldn’t fall back asleep. So I did what any highly intelligent and resourceful person would do. I doomscrolled on Twitter. (Sorry Elon, it’s still Twitter).

And then I saw it.

Jack Dorsey just cut Block’s workforce nearly in half. Over 10,000 people down to just under 6,000. More than 4,000 human beings, gone. Not because the company is struggling. No. Block reported $10.36 billion in gross profit in 2025, up 17% year-over-year. Revenue is growing. They’re serving more customers than ever. Profitability is improving.

He did it because of AI.

His words: “Intelligence tools we’re creating and using, paired with smaller and flatter teams, are enabling a new way of working which fundamentally changes what it means to build and run a company.”

The stock jumped over 20% after hours. The market cheered. Four thousand people are updating their LinkedIn profiles at 2am and the market is cheering.

Someone in the quote tweets compared this to the first reports of coronavirus breaking out as a “pneumonia-like symptom” that many people didn’t pay attention to. I don’t know if that comparison is right. But something froze inside me when I read it, and I think it is fear.

And here’s the part that actually made me sit up in bed:

“I don’t think we’re early to this realization. I think most companies are late. Within the next year, I believe the majority of companies will reach the same conclusion and make similar structural changes.”

Within the next year. The majority of companies.

Block isn’t an anomaly. Amazon laid off 16,000 employees at the end of January, having already cut 14,000 roles a few months earlier (while simultaneously ramping AI spending, naturally). Zuckerberg said he expects 2026 to be “the year that AI dramatically changes the way we work,” adding that projects that used to take big teams are now being accomplished by a single person. HP announced plans to cut up to 6,000 employees by 2028, explicitly citing AI.

This isn’t a blip. This is the pattern.

I’m not going to pretend I have answers. I’m going to tell you upfront: I don’t. I’m writing this because I can’t sleep and because the questions won’t stop looping in my head and because I think people in my industry, the energy sector, specifically renewable energy, need to be having this conversation right now instead of pretending it’s not coming for us.

Because it is coming for us.

But also (and this is the thing that’s breaking my brain at 2am) it literally cannot exist without us.

The Loop I Can’t Escape

Here’s what’s been circling in my head since I saw that tweet:

If AI accelerates and removes all tasks, what exactly does humanity do? What will drive GDP? Who even pays for the AI? If there are no jobs in the market, how does any of this stay sustainable? What does it all mean?

I keep coming back to this: the entire economic model we’ve built assumes people work, earn money, spend money, and that cycle creates demand. AI is threatening to break that cycle. Not in 20 years. Not in 10 years. Right now. Today. 4,000 people at a time.

And look, I’ll confess something. I am not some distant observer wringing my hands about the ethics of automation from a philosophical ivory tower. I am a user. A heavy user. Claude (yes, the AI I’m literally using to help research this piece, make of that what you will) is shipping new products virtually every single day. I rely on Sunclaw (sunclaw.kiisha.io) AI tools now for things I didn’t think were possible two years ago. I text an AI on Telegram and it gets things done. When I say everything, I mean everything. You don’t fully grasp what I mean by everything but I’ll leave that for another post.

Every single day, startups that launched on low friction and high speed are being killed by the same AI that birthed them, faster than they can spell ASI. The iteration speed is insane.

Many people thought AI was going to make them rich. And it did, briefly. Then the same tool made them irrelevant.

So yeah, I’m scared. I’m also complicit. We all are. The irony isn’t lost on me. Let’s keep going.

But I’m an Energy Guy, So Let Me Talk About Energy

I don’t know enough about fintech or SaaS or whatever Block’s 4,000 employees were building to tell you what their future looks like. I’m not going to pretend. (If you’ve read anything I’ve ever written, you know pretending isn’t really my thing. I literally named my blog “The Impostor’s Guide to Clean Energy.” The self-awareness is load-bearing at this point.)

What I do know is energy. Specifically, renewable energy in Africa. I spend my days structuring solar PV deals, biomass projects, battery storage solutions for industrial clients across Nigeria, Ghana, and Côte d’Ivoire. I’ve worked on 40+ renewable energy transactions. I’m doing a PhD on hybrid solar-biomass-hydrogen systems. I co-founded an AI-native platform for African energy infrastructure. I still confuse harmonic distortion with power factor correction sometimes. Twelve years in. We’ve established this.

So when I ask “what does AI mean for energy jobs?” I’m not asking from the sidelines. I’m asking from inside the house. And the house might be on fire. Or it might be fireproof. I genuinely cannot tell.

The Paradox That Should Keep Every Energy Person Up at Night (It’s Keeping Me Up, Clearly)

Here is the thing nobody in the AI discourse is talking about enough:

AI has three Achilles heels. Three things without which it simply cannot exist. Chips. Energy. Water. Take away any one of these three and AI dies. It’s a trifecta, a delicate balance upon which the entire future of artificial general intelligence rests.

This is not metaphorical. This is physics. Every large language model, every inference call, every AI agent writing code or cutting jobs or helping me research this blog post at 2am runs on electricity. And the numbers are absolutely wild.

Data centres globally consumed about 415 terawatt-hours of electricity in 2024, according to the IEA. That’s about 1.5% of global electricity consumption. By 2030, that number is projected to more than double to around 945 TWh. That is roughly equivalent to Japan’s entire electricity consumption. Japan. An entire industrialised nation’s worth of power. Just for data centres.

(I’m going to throw more numbers at you because I’m an energy guy and numbers are how I cope with existential dread.)

AI-specific servers alone consumed an estimated 53 to 76 TWh in the US in 2024. By 2028, that could climb to 165 to 326 TWh. Training a single large model like GPT-4 consumed roughly 50 GWh of energy and generated over 550 tons of CO2, equivalent to the annual carbon footprint of 121 US households. And here’s the bit that really gets me: training is actually the smaller piece. Inference, the process of actually using these models after they’re trained (every ChatGPT query, every Claude response, every AI-generated email), accounts for up to 90% of a model’s total lifecycle energy consumption.

Goldman Sachs estimates current global data centre power usage at around 55 GW, projecting that to reach 122 GW by 2030. S&P Global puts total global power demand from data centres at 860 TWh in 2025, rising to 1,587 TWh by 2030. These aren’t incremental numbers. This is a structural shift in global electricity demand of a kind we haven’t seen since, I don’t know, industrialisation?

And how fragile is all of this? Hilariously fragile. In 2024, a minor grid disturbance in Virginia’s Fairfax County caused 60 data centres to switch to backup generation simultaneously. The sudden loss was around 1,500 megawatts. That’s roughly equivalent to the entire power demand of Boston. One hiccup. One county. Nearly cascaded into widespread failure. And Virginia’s data centres already consume 26% of the state’s total electricity supply.

So here’s your paradox: the technology that is eliminating jobs across every sector is simultaneously creating the single largest surge in electricity demand in human history. And somebody has to build, finance, operate, and maintain the infrastructure to meet that demand.

That somebody is us. The energy people.

The tech companies know it too. They’re scrambling. Google opened its first UK data centre powered by a renewable portfolio managed by Shell. Amazon invested $700 million in small modular reactor technology through X-energy (SMRs! For data centres!). Plans are underway to revive retired nuclear plants, including Three Mile Island (yes, that Three Mile Island), specifically to power AI. Several US states are now weighing legislation requiring data centres to draw power from renewable sources.

The AI revolution runs on electricity. And it is desperately short of the clean kind.

What AI Is Already Doing to Energy Jobs

Let me be honest about both sides of this, because one thing I’ve learned from years of writing about energy is that people can smell BS from three continents away. So here’s the full picture.

The stuff that’s being automated or compressed (a.k.a. the part that should make you uncomfortable):

The modelling work I do? Financial models for solar projects, IRR calculations, sensitivity analyses, tariff structures? AI can already do 80% of that in minutes. Work that used to take a junior analyst a week, Claude or GPT can draft in an afternoon. The iteration speed means one person with AI tools can now do what three or four people did two years ago. I know this because I am that one person. I used to need a team for things I now do with a prompt and a coffee.

AI-driven predictive maintenance is already cutting wind turbine downtime by up to 20% and extending asset life by 15%. Fewer emergency repair crews. Fewer unplanned maintenance dispatches. More uptime. Less humans.

Site assessment? Satellite imagery analysis that used to require teams of engineers walking sites with GPS equipment can now be done with AI-powered geospatial tools in hours.

Energy forecasting? AI predicts solar irradiance and wind speeds with over 95% accuracy now. The old models, the ones that employed teams of meteorologists and data scientists to build and maintain, are being replaced by systems that learn and improve on their own.

Procurement and supply chain optimization? Automated. Contract review and due diligence? Getting there. Regulatory compliance monitoring? AI is eating that too.

If your primary value as an energy professional is sitting at a desk running spreadsheets and generating reports, you should be paying very close attention. I say this as someone whose primary value was, for a long time, sitting at a desk running spreadsheets and generating reports.

The stuff that can’t be replaced yet (a.k.a. the part that lets me sleep at night, on the nights I can actually sleep):

Nobody is sending an AI agent to negotiate a power purchase agreement with a Nigerian industrial offtaker. Trust me on this. The cultural nuance, the relationship building, the ability to read a room when a CFO is nervous about committing to a 15-year contract, that is deeply, irreducibly human. I’ve been in those rooms. The AI would get eaten alive. Politely, with kola nut and small talk, but eaten alive nonetheless.

Nobody is sending an AI to navigate the politics of getting grid connection approvals in Ghana. Or to manage the on-the-ground realities of constructing a biomass facility in a West African industrial zone where the logistics alone require a level of improvisation that no training dataset can capture. (Last month I had to solve a transport bottleneck that involved, among other things, a road that didn’t exist on any map. Good luck prompting your way through that one.)

The physical infrastructure has to be built by human hands. Solar panels don’t install themselves. Transmission lines don’t appear from the cloud. Substations need engineers who can work in 40-degree heat with unreliable supply chains.

And the deal-making. The deal-making is where I exhale a little. Structuring the first FX-hedged commercial bioenergy agreement in Ghana wasn’t something you could prompt your way through. It required understanding regulatory frameworks across multiple jurisdictions, building trust with counterparties over months, and creative problem-solving that emerged from years of pattern recognition in markets that don’t have standardized playbooks.

You want to know what energy deal-making in Africa is like? It’s jazz. It’s improvisation on top of theory. The theory is the financial model. The improvisation is everything else. And “everything else” is about 85% of the job.

The Africa-Specific Reality (Or: Why Silicon Valley’s AI Discourse Misses the Plot)

Here’s something the AI conversation completely ignores: the world is not uniform.

In markets like Nigeria, Ghana, and across West Africa, the energy transition is not a luxury. It is not a nice-to-have sustainability initiative for the ESG section of an annual report. It is an infrastructure imperative. These are markets where grid reliability is a daily challenge, where industrial clients are running on diesel generators and paying three to five times what they should for power, where the economics of solar and biomass aren’t just “green,” they’re survival.

AI might be collapsing headcount at Block’s offices in San Francisco, but across Africa, we don’t have enough energy professionals to begin with. The International Renewable Energy Agency reported that global RE employment hit 16.2 million in 2023 and is projected to more than double to over 30 million by 2030. And a massive chunk of that growth needs to happen in emerging markets where the talent pipeline is already running on fumes.

The IEA’s own data shows Africa has less than 1 kWh of data centre electricity consumption per capita. The US? Hundreds. Even by 2030, that gap barely narrows. The energy access challenge in Africa is not being solved by Silicon Valley’s language models. It’s being solved by people on the ground doing hard, complex, deeply human work. People who navigate regulatory chaos, build relationships with communities and governments and industrial clients simultaneously, and get creative when the standard playbook doesn’t apply.

In Africa, the standard playbook never applies. That’s not a bug. That’s the whole market.

The Talent Gap Nobody’s Talking About (Or: The Irony That Might Save Us)

While AI is handing out pink slips in tech, the energy industry literally cannot find enough people.

BCG found that 46% of energy companies cite talent skill gaps as the primary barrier to AI adoption. Not budget constraints. Not technology limitations. Talent. Roles requiring both energy domain expertise and AI/data science capabilities are staying open for 90 days or longer. Some go unfilled for six months. Companies are paying 30 to 40% premiums above standard market rates for these hybrid skill sets. And even at those levels, they’re struggling to compete.

The WEF’s Future of Jobs Report 2025 lists renewable energy engineers and environmental engineers among the 15 fastest-growing job categories globally. The clean energy workforce is expected to expand faster than nearly any other sector.

Let me say that again because I need you to actually absorb it: the sector is expanding faster than nearly any other. While Block cuts 40% of its workforce because AI made them redundant, the energy industry is desperately trying to hire people it cannot find.

The bottleneck isn’t demand. The bottleneck is supply. The energy sector is hiring for people who can operate at the intersection of physical energy systems and digital intelligence. That intersection is tiny. And it’s where the next generation of careers lives.

If you’re reading this at 2am like I’m writing it, this is the part you should screenshot.

So Where Does This Leave Us? (Somewhere Between Terror and Defiance, Apparently)

I keep going back and forth between two feelings:

Terror. Because the speed of AI improvement is genuinely exponential, not linear. What can’t be automated today might be trivially automated in 18 months. The junior analyst roles, the entry-level modelling positions, the support functions that are the pipeline for future energy leaders: those are getting compressed right now. Where do the next generation of energy professionals come from if the entry-level jobs disappear? This question haunts me. I don’t have a good answer.

Cautious defiance. Because energy is one of the few sectors where atoms matter as much as bits. You can’t AI your way out of needing a physical solar panel on a physical roof connected to a physical grid. The world needs more energy, not less. Dramatically more. And the harder, messier, more human work of deploying that energy, especially in frontier markets, is not something a language model can do from a data centre in Virginia.