<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Kaykl.uz on Medium]]></title>
        <description><![CDATA[Stories by Kaykl.uz on Medium]]></description>
        <link>https://medium.com/@kaykluz?source=rss-2dfae1fe39fa------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*P6_0WIHrrVn-U9xlc-jT_w.jpeg</url>
            <title>Stories by Kaykl.uz on Medium</title>
            <link>https://medium.com/@kaykluz?source=rss-2dfae1fe39fa------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 01:00:40 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@kaykluz/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Jensen Huang, Nvidia and the Price of One Man]]></title>
            <link>https://kaykluz.medium.com/jensen-huang-nvidia-and-the-price-of-one-man-c1630b5aab53?source=rss-2dfae1fe39fa------2</link>
            <guid isPermaLink="false">https://medium.com/p/c1630b5aab53</guid>
            <category><![CDATA[nvidia]]></category>
            <category><![CDATA[risk-management]]></category>
            <category><![CDATA[geopolitics]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[africa]]></category>
            <dc:creator><![CDATA[Kaykl.uz]]></dc:creator>
            <pubDate>Thu, 12 Nov 2026 19:48:52 GMT</pubDate>
            <atom:updated>2026-11-12T19:48:52.799Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bwbGDu9jRjsJsytF.jpeg" /></figure><p>On Thursday morning Nvidia told the market that Jensen Huang had resigned as president, chief executive and director, with immediate effect. Colette Kress, the chief financial officer, is interim chief executive. Harvey Jones becomes the first chair the board has ever had (<a href="https://www.sec.gov/Archives/edgar/data/0001045810/000104581026000082/nvidiaannouncesceotransiti.htm">Nvidia’s filing</a>). Huang co-founded the company in 1993 and ran it for 33 years.</p><p>The newspapers give several reasons, concerning his private life and an alleged link to Chinese intelligence. I will leave those to the newspapers and to Congress, where he has <a href="https://www.cnbc.com/2026/11/12/nvidia-jensen-huang-ceo-live-updates.html">reportedly agreed to testify</a>. I have no information and a strong suspicion that my opinion would not improve matters.</p><p>What I can do is arithmetic. The shares closed near $160 on Wednesday and opened 13% lower (<a href="https://www.ms.now/news/nvidia-jensen-huang-resigns-affair-china-spying">MS NOW</a>). For a company valued at about $4 trillion, that is roughly $500 billion gone between breakfast and the first coffee. My math might be wrong, so check it. Boston University counted every loan Chinese lenders committed to African governments from 2000 to 2024, the railways, the dams, the airports, the lot, and the total comes to about <a href="https://www.bu.edu/gdp/files/2026/01/GCI-PB-28-CLA-2026-EN.pdf">$181 billion</a>.</p><p>One man left one building in California, and a sum more than twice the size of a quarter-century of Chinese lending to a continent left with him. It may come back next week. That is hardly reassuring. A number that large should not be able to move that fast because of one person, and the fact that it can is the subject of this essay.</p><h3>Key-man risk</h3><p>In my trade the term is key-man risk. Fund documents carry a key person clause: if the named partners leave, the fund stops investing until the investors decide what to do. Lenders ask ten-person developers for key-man insurance, and mark a small sponsor down for being, in the usual phrase, really just one guy.</p><p>Nvidia was really just one guy. Nobody marked it down. In January a reporter asked Huang how long he would stay, and he said the secret of a long tenure was not getting fired and not getting bored. The company had <a href="https://finance.yahoo.com/news/long-jensen-huang-nvidias-ceo-132923787.html">no announced succession plan</a>, and shareholders were said to be uneasy about it, in the way that people are uneasy about a noise in the car they have decided not to investigate.</p><p>In July I wrote <a href="https://kaykluz.com/p/how-to-build-a-billion-dollar-energy-d95">the NVIDIA Thesis</a>, which held the company up as a template for building a billion-dollar energy company in Africa. I listed five things Nvidia did, including funding its own customers and standing behind their debts, and then five ways the playbook can kill you. “Reportedly dating a suspected intelligence asset” was not among them. Nor was the chief executive in any form, nor Taiwan, nor the export licence. I assessed one of the most valuable companies on earth and never asked what happens if the founder leaves, which is the first question I would put to a developer with three staff and a pickup truck. I will be more thorough in future templates.</p><p>The man was only the top layer. Nvidia holds something like <a href="https://www.nasdaq.com/articles/nvidia-earns-69-its-revenue-us-heres-how-tariffs-could-affect-2026">90% of the market for AI accelerator chips</a>. Most of those chips are made by one foundry, TSMC, on one island, Taiwan. Nvidia also <a href="https://kaykluz.com/p/how-to-build-a-billion-dollar-energy-d95">finances a fair number of its own customers</a>, which is why a Yahoo Finance writer could file a piece on Thursday headed <a href="https://finance.yahoo.com/technology/live/nvidia-live-updates-jensen-huang-resigns-cfo-colette-kress-to-serve-as-interim-ceo-140341430.html">“The Bank of Nvidia just lost its banker”</a>, and why the smaller cloud companies that live on its chips fell with it.</p><p>The sociologist Charles Perrow described this shape in 1984, in a book called <em>Normal Accidents</em>. When a system is complex and its parts are tightly coupled, he argued, failures are a property of the design. Nobody has to be stupid. The accident is waiting in the wiring diagram, and the only open questions are which component and which Thursday.</p><p>Everybody has a Jensen: a dependency so reliable for so long that it stopped appearing on the risk register. A founder, a supplier, a lender, a currency, a strait.</p><h3>Trade wars</h3><p>A graphics card used to be something teenagers wanted for Christmas. It is now a controlled export. Washington restricts which Nvidia chips may go to China, and when it relented on one model, the H200, the price was that Nvidia hands <a href="https://www.nasdaq.com/articles/nvidia-earns-69-its-revenue-us-heres-how-tariffs-could-affect-2026">25% of the revenue from those sales</a> to the United States government. A sales commission, payable to a state.</p><p>In 1945 Albert Hirschman published <em>National Power and the Structure of Foreign Trade</em>, a study of how Germany in the 1930s made the small economies of south-eastern Europe dependent on its market and then used the dependence. His point was that trade is never only trade. Whoever can interrupt it holds power over whoever cannot afford the interruption.</p><p>In 2019 Henry Farrell and Abraham Newman updated the idea in a paper called “Weaponized Interdependence”. Global networks, they wrote, are built like hubs and spokes. States that control the hubs can watch what flows through them and can choke it off. Payment systems were their example. Advanced chips have since joined the list.</p><p>Read Thursday announcement in that light. The chief executive of the hub company became the subject of a national security question, and half a trillion dollars moved. I make no judgement on the man. The lesson I take is structural: once a product becomes strategic, the company that makes it becomes an instrument of foreign policy whether it likes it or not, and so do its suppliers and its customers.</p><h3>Neither good nor bad</h3><p>Melvin Kranzberg was a historian of technology at Georgia Tech, and in 1986 he set out six laws. The first is the one people remember: “Technology is neither good nor bad; nor is it neutral.”</p><p>The second half is the part worth keeping. People who say technology is neutral usually mean it is harmless until someone misuses it. Kranzberg meant the opposite. A technology rearranges who depends on whom the moment it arrives, before anybody has decided to use it well or badly. Asking whether AI is good for Africa is like asking whether weather is good for farming.</p><p>Huang’s career is the cleanest illustration I can think of. The same wave that made a graphics-chip company one of the most valuable businesses on earth is what made its chips a matter of state. Had Nvidia still been selling cards to gamers, the private life of its chief executive would have been a story for the business pages, briefly. The technology made him enormously rich and it made him strategically important, and the second of those is not a comfortable thing to be. He rose on it and he fell in the weather it created.</p><p>Daron Acemoglu and Simon Johnson spent a long book, <em>Power and Progress</em>, arguing from a thousand years of evidence that the gains from a new technology go to whoever has the power to direct it, and that broad benefit has always had to be organised and never simply arrived. That is uncomfortable reading for a continent with nearly 18% of the world’s people and <a href="https://www.bakermckenzie.com/en/insight/publications/2026/06/south-africa-the-megawatt-and-the-token-africas-compute-equation">0.6% of its data centre capacity</a>.</p><p>It need not be fatalistic reading. Epictetus, who began life as a slave and so had some standing on the subject, divided the world into what is up to us and what is not. Export controls on chips are not up to us. Neither is the price of oil, the mood of the American Congress, or a strait 6,000 kilometres from Lagos.</p><p>What we borrow in, what we build with, who we sell to, and how many of our eggs sit in each basket: those are up to us. Nassim Taleb’s whole body of work reduces to a sentence a grandmother could have supplied, which is that you cannot predict the shock and you can decide how much of you it is allowed to break.</p><h3>Back to Thursday</h3><p>Nvidia will be fine, probably. It reports results on Tuesday, it has a capable interim chief, and its order book did not resign. The more durable fact is that the most important company in the most important industry of the decade turned out to have a single point of failure that everyone could see and nobody priced.</p><p>Now look at home. Chinese lending to Africa has fallen <a href="https://semafor.com/article/01/23/2026/china-pulls-back-on-funding-african-projects">more than 90% from its 2016 peak</a>, yet <a href="https://ember-energy.org/latest-insights/the-take-off-in-african-solar-that-official-statistics-cant-yet-see/">94% of the solar panels</a> we install are Chinese. Rated African governments owe about <a href="https://newscentraltv.com/?p=383039">$90 billion of external principal</a> this year, in currencies we do not print. And when the Strait of Hormuz closed in February, a continent of oil producers found itself on the wrong side of <a href="https://www.ogj.com/general-interest/economics-markets/news/55363530/iea-middle-east-war-triggers-the-largest-supply-disruption-in-history">the largest oil supply shock on record</a>, because we import most of the fuel we burn.</p><p>We are not better than Nvidia’s shareholders. We have a continent’s worth of arrangements that work because one counterparty keeps turning up: one lender, one supplier, one currency, one fuel, one buyer. AI is tightening every one of those links, mostly without asking us. It will make some Africans rich and it will cost others their business, and in both cases the cause will be something that happened a long way away, on a Thursday.</p><p><strong>P.S.</strong> This is written in a personal capacity. It is not investment advice, and if you were planning to take investment advice from someone who calls himself the Impostor, we should talk about your risk register first.</p><p><strong>P.P.S.</strong> The Impostor has never run a four trillion dollar company and has had mixed results with a four megawatt one. If you think I have this wrong, and I suspect several of you do, say so in the comments. I would rather be corrected in public than be confidently wrong in private.</p><p><em>If this was useful, forward it to the person in your company who is the whole company. They will know why.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c1630b5aab53" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[My AI Would Like to Speak to Your AI]]></title>
            <link>https://kaykluz.medium.com/my-ai-would-like-to-speak-to-your-ai-7431517b61c8?source=rss-2dfae1fe39fa------2</link>
            <guid isPermaLink="false">https://medium.com/p/7431517b61c8</guid>
            <category><![CDATA[humor]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Kaykl.uz]]></dc:creator>
            <pubDate>Wed, 28 Oct 2026 14:16:01 GMT</pubDate>
            <atom:updated>2026-10-28T14:16:01.580Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*yRp7ZT8r5yCmbkKR.png" /></figure><p>Everyone has an AI assistant now.</p><p>Meta has Muse. Anthropic has Claude Duet. OpenAI has Dots. xAI has Grok Bots. There is Instinct, which you can text or call like a suspiciously capable friend who never asks to borrow money. They can monitor inboxes, schedule meetings, book travel, fill forms, prepare forecasts, build software and continue working after you have closed the app and returned to the more important business of watching somebody renovate a kitchen on Instagram.</p><p>This all sounds wonderful.</p><p>The problem is that the person you are trying to impress also has one.</p><p>I have been thinking about this because we appear to have entered a new phase of artificial intelligence. The first phase was about answering questions. You asked ChatGPT to explain quantum mechanics, rewrite an email or produce a meal plan from one egg, half an onion and the terrible consequences of not going grocery shopping.</p><p>The new phase is different. The AI no longer wants to help you think. It wants your passwords.</p><p>Not in an openly criminal way. It simply wants access to your email, calendar, documents, browser, contacts and perhaps your bank account so that it can “act on your behalf.” This phrase appears throughout the marketing because it sounds much nicer than “we have given the computer power of attorney.”</p><p>Muse runs on its own cloud computer and can browse, send emails, fill forms and book things. Claude Duet checks connected accounts, prepares a morning briefing and continues working while you are away. Instinct lives inside messaging and phone calls. Grok Bots can coordinate multiple agents and turn repeated processes into routines.</p><p>These are not chatbots anymore.</p><p>They are employees who live inside your laptop and have never attended safeguarding training.</p><p>A lot of this energy came from OpenClaw, the open-source personal agent that went viral because it offered something closer to a do-it-yourself Jarvis. It could live in chat applications, use tools, follow schedules and carry out tasks on your machine. It also required enough technical confidence that ordinary people could look at the installation instructions and decide they were actually quite happy booking their own dentist appointment.</p><p>The initial noise around OpenClaw eventually became less interesting than what it unleashed. Meta later acknowledged that Muse was heavily inspired by OpenClaw, although it was built separately. The ambition was essentially to take the same idea, remove the part involving configuration files and make it usable by billions of people.</p><p>This is the normal life cycle of a technology idea now. Somebody builds a strange open-source tool. It becomes briefly unavoidable on X. Forty-seven people announce that software has changed forever. A large technology company recreates it with rounded corners. Then six other companies release versions with names suggesting friendship, intuition or mild artistic sensitivity.</p><p>Muse.</p><p>Duet.</p><p>Instinct.</p><p>Dots.</p><p>These are not software names. This is the lineup for a wellness retreat where your calendar will finally learn to forgive you.</p><p>And, because this is artificial intelligence, one assistant is no longer sufficient. Grok lets people create teams of bots. Muse can launch subagents. Your main assistant may soon have an assistant, who may have a research assistant, who will presumably complain that the main assistant keeps scheduling meetings without an agenda.</p><p>We have successfully automated management.</p><p>The same thing is happening with software.</p><p>AI has made it dramatically easier to build an application that looks like an application. You can describe the product, generate the interface, connect a database, add authentication and launch before the original thought has had time to become embarrassed.</p><p>This is extraordinary. It is also why everybody is building the same tool.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*u95VWjMDC9TTpKTb.jpeg" /><figcaption>Each founder assures me his is completely different. Photo: D J Shin / My Toy Museum, <a href="https://commons.wikimedia.org/wiki/File:Gang_of_Six.jpg">Wikimedia Commons</a> (<a href="https://creativecommons.org/licenses/by-sa/3.0/">CC BY-SA 3.0</a>).</figcaption></figure><p>Someone sees a successful note-taking app and produces a “reimagined AI-native workspace.” Somebody else builds an AI voice recorder. Then an AI voice recorder for lawyers. Then an AI voice recorder for lawyers who enjoy running. By Thursday, somebody has built an AI agent that listens to all the recordings and tells everyone that the meeting should have been an email.</p><p>The cost of producing software has fallen faster than the cost of producing a reason for the software to exist.</p><p>That is the awkward part.</p><p>A large number of these products are reverse-engineered versions of something already popular. The interface may look professional — sometimes almost pixel-perfect — but the underlying performance, support and differentiation can still be thin. Many are launched, shown to friends, posted on Product Hunt and abandoned after the downloads fail to arrive.</p><p>This does not mean AI-assisted software development is useless. Far from it. It means building has become less conclusive.</p><p>We used to treat the existence of a working product as evidence that somebody had discovered something important. Building software required money, time, engineers, project managers and at least one person whose only apparent responsibility was asking when the API would be ready. If a product existed, somebody had suffered for it.</p><p>Now a product can exist because somebody had a free weekend and access to Claude Code.</p><p>The software may be real. The company is still awaiting confirmation.</p><p>When production becomes cheap, production stops being the impressive part. The scarce things become distribution, trust, judgment, proprietary insight, access to customers and the patience to maintain the product after the founder has discovered another exciting idea.</p><p>Unfortunately, those are precisely the parts AI cannot complete before lunch.</p><h3>The proxy problem</h3><p>But the stranger problem is not that everybody can build software.</p><p>It is that everybody can now build a proxy for themselves.</p><p>The original promise of these tools was to help us interact more effectively with the real world. AI would improve the website that customers visit. It would polish the proposal that investors read. It would rewrite the email before it reached a colleague. It would turn rough thoughts into a presentation suitable for humans with budgets.</p><p>The destination was always another person.</p><p>We were using machines to make ourselves more understandable, attractive and efficient to humans. The website looked better so that a human would trust us. The email sounded clearer so that a human would reply. The LinkedIn post had a dramatic opening, three lessons and an unnecessary reference to leadership because a human was expected to encounter it while avoiding actual work.</p><p>Then the other humans acquired AI.</p><p>Now I ask my assistant to turn my untidy thoughts into a professional email. I send it to you. Your assistant summarises the professional email into three bullet points because you are busy. You dictate a rough response. Your assistant expands it into four polished paragraphs. My assistant summarises those four paragraphs for me.</p><p>Two humans have exchanged eleven words using 900 words of artificial diplomacy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/918/0*hK96TPaKv53X5iQN.jpeg" /><figcaption>The infrastructure required to tell someone I’m free on Thursday. Illustration: <a href="https://commons.wikimedia.org/wiki/File:Storm_P-maskine.jpg">Storm P, 1920</a> (public domain).</figcaption></figure><p>The computers are having a wonderful correspondence. We are barely involved.</p><p>Imagine the full process. My AI examines your company, studies your website and writes a personalised sales message. Your AI recognises that it is a sales message, summarises it and assigns it a low priority. My AI sends a follow-up. Your AI files it. My AI detects no response and sends something called a “gentle nudge.” Your AI identifies the nudge as neither gentle nor urgent.</p><p>For six weeks, our assistants maintain a professional relationship that neither of us knows exists.</p><p>Eventually, they schedule a meeting.</p><p>We both decline.</p><h3>Better for whom?</h3><p>This creates what might be called the proxy paradox.</p><p>We are building increasingly sophisticated AI systems to improve how we present ourselves to people who are using increasingly sophisticated AI systems to avoid examining what we presented.</p><p>If everyone is polishing, summarising, personalising and optimising at once, the advantage begins to disappear. The baseline simply moves.</p><p>A beautifully designed website once signalled resources, taste and seriousness. Now AI can generate a respectable one before the founder has decided what the business does. Elegant writing once suggested that someone had spent time organising a thought. Now elegance may merely indicate that the prompt included “make this more concise and professional.”</p><p>The output has improved.</p><p>The signal has weakened.</p><p>This is not redundancy in the strict technical sense. The assistants may genuinely save time. They may complete errands, reconcile information, build dashboards and prevent important tasks from disappearing into the swamp beneath your inbox.</p><p>The redundancy appears when we use AI to manufacture human-facing artefacts that another AI immediately removes.</p><p>My agent creates the presentation. Your agent extracts the numbers. My agent writes the background section. Your agent skips it. My agent adjusts the tone to sound warm. Your agent reports that “the sender is expressing warmth.”</p><p>At some point, the obvious question becomes: why are we still generating the PowerPoint?</p><p>If the machines are doing both sides of the exchange, perhaps they should communicate the actual intention, constraints and evidence directly. My assistant tells yours what I want. Yours checks your preferences, authority and calendar. They negotiate the available options. We receive the decision.</p><p>No introductory paragraph.</p><p>No “I hope this email finds you well.”</p><p>Nobody needs to be found.</p><h3>AI slop is social slop</h3><p>I suspect this is partly why people dislike AI slop so intensely.</p><p>The problem is not simply that a machine participated. People already accept machine assistance everywhere: spelling correction, photo editing, search, recommendations, navigation and the small miracle by which Microsoft Teams attempts to remove the sound of your neighbour drilling through concrete.</p><p>The insult is counterfeit effort.</p><p>A message arrives looking as though someone carefully considered you, but you can feel that nobody did. A post imitates conviction without anyone having risked an opinion. An article has headings, confidence and seventeen uses of “moreover,” but no identifiable reason for being alive.</p><p>The content is trying to trigger the social response normally earned by human effort.</p><p>Trust me.</p><p>Notice me.</p><p>Hire me.</p><p>Believe that I sat down and thought about this.</p><p>That last one matters. AI slop does not merely look artificial. It often feels like someone has outsourced the act of caring while retaining the expectation that you should care in return.</p><p>And audiences are becoming sensitive to it. Platforms have started responding to complaints about mass-produced synthetic material, while readers increasingly associate generic AI output with weak originality and emotional connection.[<a href="https://economictimes.indiatimes.com/tech/artificial-intelligence/ai-slop-the-rise-of-ai-generated-content-and-its-impact-on-social-media/articleshow/131314860.cms">economictimes.indiatimes</a>][<a href="https://nytimes.com/2026/08/17/technology/ai-slop.html?eafs_enabled">nytimes</a>][<a href="https://vogue.com/article/the-anti-ai-slop-playbook">vogue</a>]</p><p>Again, this does not make AI the enemy. The distinction is not between “AI” and “human.” It is between amplification and substitution.</p><p>AI can help a person express an actual observation more clearly. That is amplification.</p><p>AI can also generate the observation, manufacture the emotion, polish the prose and publish it under the name of a person who encountered none of it.</p><p>That is administrative identity theft.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*1Al1m-gEhleJo2Fj.jpeg" /><figcaption>An early version. Unfortunately, you still had to attend your own conversations. <a href="https://commons.wikimedia.org/wiki/File:Telephonoscope.jpg">George du Maurier, Punch, 1878</a> (public domain).</figcaption></figure><h3>The world after polish</h3><p>For the last two years, much of the AI conversation has assumed that the winning person will be the person with the most capable tools.</p><p>Perhaps.</p><p>But when everybody has capable tools, the advantage changes. It moves away from production and towards selection.</p><p>What did you notice that the model could not have noticed without you?</p><p>What do you know because you were in the room?</p><p>What are you willing to believe before there is enough public data to make the belief obvious?</p><p>Which customer trusts you enough to describe the real problem instead of the problem they put in the request for proposals?</p><p>What will you continue building after the viral excitement disappears and everyone else moves on to an AI app for interpreting their AI app analytics?</p><p>The answer may be that human advantage does not disappear. It relocates.</p><p>Writing becomes cheaper, but having something to say becomes more valuable. Software becomes easier to build, but choosing a durable problem becomes harder. Websites become more beautiful, but beauty becomes weaker evidence of substance. Assistants become more capable, but permission, trust and responsibility become more consequential.</p><p>The future may contain enormous amounts of machine-to-machine activity. Agents will schedule other agents, negotiate with agents, verify agents and perhaps complain privately about their human operators. That could remove a huge amount of administrative nonsense from everyday life.</p><p>Good.</p><p>I would happily delegate the part of adulthood involving broadband providers.</p><p>But the human-facing layer cannot consist entirely of machines producing increasingly polished material for other machines to compress. Eventually somebody must supply an intention that was not generated, a judgment that carries consequences or a relationship worth preserving.</p><p>Otherwise we have not automated communication.</p><p>We have created two very intelligent receptionists passing elaborate messages between two people sitting silently in the same room.</p><p>Muse is working. Duet is monitoring. Instinct is standing by. Grok has assembled a team of bots. Somewhere, OpenClaw is watching all its children become corporate.</p><p>And then there is me.</p><p>I still have to decide what I actually want.</p><p>As you can see, my assistant has not solved the difficult part.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7431517b61c8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Build a Billion-Dollar Energy Company in Africa: The Starcloud Thesis]]></title>
            <link>https://kaykluz.medium.com/how-to-build-a-billion-dollar-energy-company-in-africa-the-starcloud-thesis-a95681527bc4?source=rss-2dfae1fe39fa------2</link>
            <guid isPermaLink="false">https://medium.com/p/a95681527bc4</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[space]]></category>
            <category><![CDATA[africa]]></category>
            <category><![CDATA[energy]]></category>
            <dc:creator><![CDATA[Kaykl.uz]]></dc:creator>
            <pubDate>Mon, 19 Oct 2026 09:15:43 GMT</pubDate>
            <atom:updated>2026-10-20T16:25:41.446Z</atom:updated>
            <content:encoded><![CDATA[<p>One of my favourite people right now is Philip Johnston.</p><p>Philip is the co-founder and CEO of Starcloud, a company building data centres in space.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/250/0*J8fesuNYGmDenhEl" /></figure><p>Not <em>using space technology to improve data centres</em>. Not <em>software for satellite operators</em>. Actual data centres. In actual space.</p><p>I feel this is a perfectly reasonable basis for affection.</p><p>You have to admire the clarity. Most companies spend six months with consultants looking for a tasteful way to describe what they do. Starcloud appears to have walked into the room, put a GPU on the table and said, “We are sending this upstairs.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*KNtmLS1o3j4OTQ7k.jpeg" /></figure><p>And then it did. In November 2025, Starcloud launched a 60 kg satellite, roughly the size of a small fridge, carrying an Nvidia H100: the sort of chip that normally sweats inside an expensive terrestrial data centre. A month later it trained a small language model in orbit. In March 2026 it raised $170 million at a $1.1 billion valuation, becoming the fastest unicorn in Y Combinator’s history, 17 months after demo day. In August it added another $250 million, and the valuation went to $2.3 billion. The company has about 25 employees.</p><p>Can somebody tell me this is normal?</p><p>Please do not tell me it is impressive. I know it is impressive. I want to know whether it is normal for a group of adults to decide that the answer to congested electricity grids is to fire the computers into orbit, and then be handed several hundred million dollars for their trouble.</p><p>Because this is precisely the behaviour that reminds me never to underestimate an organised group of reasonably intelligent people.</p><p>There is an important difference between ordinary stupidity and organised stupidity. Ordinary stupidity is impulsive. Somebody leans too far back on a plastic chair, physics remembers its responsibilities, and a lesson is delivered.</p><p>Organised stupidity has a WhatsApp group.</p><p>It has workstreams. Someone has built a budget. Someone else has prepared a risk register that rates “death” as medium probability, high impact. There is a procurement schedule, three extension cables and a man called Mike who says, “Don’t worry, I know somebody.”</p><p>Intelligent people do not fall into stupid ideas. They develop them. They run the numbers, commission a feasibility study and collect quotations from three approved suppliers.</p><p>Eventually somebody tries to restore order: “Are we sure this is a good idea?”</p><p>But nobody said it was a good idea. They said it was technically possible.</p><p>Once enough intelligent people discover that something ridiculous is technically possible, good judgement takes annual leave. The budget is approved. Mike is on his way.</p><p>Random stupidity can usually be interrupted by taking someone’s keys. <strong>Organised stupidity has already booked the launch window.</strong></p><p>In Starcloud’s case, that is not a metaphor. Johnston told TechCrunch that securing launch capacity is now one of the company’s biggest costs. They have, quite literally, booked the launch window.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/522/0*4iHCF5keQzDuFsVQ.jpeg" /></figure><p>So I went to the company’s white paper, intending to be clever about the physics. Cooling in a vacuum, where there is no air to carry heat away. Radiation. The nearest technician being several hundred kilometres straight up in a location he cant solve the problem by asking “Have you tried turning it off and on again?”.</p><p>Instead, I got stuck on a table.</p><p>The table compares running a 40 MW computing cluster for ten years on Earth against running it in orbit. On Earth: $140 million of electricity, $20 million of backup power, $7 million of cooling, 1.7 million tonnes of water. In orbit: $2 million for the solar array, $5 million for one launch, $1.2 million for radiation shielding. Total: $167 million against $8.2 million.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/597/0*5eQuQf7HK2OW_Cuu.png" /></figure><p>I have built that type of table before. Diesel against solar for a flour mill in Nigeria. Generator against power purchase agreement for way too many factories in the region. The rows are the same: energy, backup, cooling, and an unnamed line for whatever the grid decided to do this week.</p><p>That was when it clicked. <strong>Starcloud is not really a space company. It is an energy company that found a site with no grid queue, no permits, no night and no community meeting.</strong> Its whole pitch is a cost-of-electricity argument. And its long-term plan for selling to the biggest cloud companies, according to the research firm Sacra, is capacity reservations structured like energy offtake agreements.</p><p>It is an IPP. In orbit. (With better fonts).</p><p>Which brings me back to the question this series keeps asking. How do you build a billion-dollar energy company in a market like Africa, a market that seems determined to take back every margin you create?</p><p>Here is how Starcloud is doing it on a much harder site. And here is what transfers to ours.</p><h3>Part 1: What Starcloud actually did</h3><p>In about two years, Starcloud went from a white paper to a $2.3 billion valuation on roughly $450 million of capital and no disclosed revenue. The sequence matters more than the headline, so here it is, newest first.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/641/0*0eMLFL8eZc4rmZb4.png" /></figure><p>Two things stand out.</p><p>First, revenue is not in the table. Sacra notes the company had <a href="https://sacra.com/c/starcloud/">not disclosed revenue</a> as of May 2026 and describes it as pre-commercial. Every dollar of that $2.3 billion is a price on a future that has been partly demonstrated and mostly not.</p><p>Second, the timeline slipped and the valuation still doubled. Investors were not paying for a delivery date. They were paying for the demonstration, the team and the size of the problem. I have known lenders who would reprice a project for a two-week slip in a land title. Different capital, different rules. We will come back to that, because it is the part that does not transfer cleanly.</p><h3>Part 2: The strategy under the rocket</h3><p>Take the rocket away and Starcloud’s strategy is eight moves. None of them needs orbit. All of them need nerve.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ceBTmofmWTj3Pnkl.jpeg" /></figure><ol><li><strong>Attack the binding constraint, not the visible problem.</strong> Everyone could see the AI chip shortage. Starcloud bet that the harder limit was electricity: grid connections, permits and power plants that take years to build. Its <a href="https://starcloudinc.github.io/wp.pdf">white paper</a> opens with energy, not chips.</li><li><strong>Go where the constraint does not exist.</strong> In the orbit Starcloud chose, the solar panels see the sun more than 95% of the time. The white paper puts the median US solar farm at about 24%. No night, no weather, no grid queue, no planning committee.</li><li><strong>Ride a cost curve someone else is paying for.</strong> Starcloud does not build rockets or chips. It assumes SpaceX will make launch cheap and Nvidia will keep making chips better, and builds the bit in the middle. One rival, Cowboy Space, went the other way and <a href="https://techcrunch.com/2026/05/11/there-arent-enough-rockets-for-space-data-centers-cowboy-space-raised-275-million-to-build-them/">raised $275 million to build its own rockets</a>. We will find out which is braver.</li><li><strong>Prove it small, early and cheap.</strong> Three million dollars, 21 months, one 60 kg satellite, one chip. That single demonstration did more for the valuation than any slide deck could.</li><li><strong>Sell the wedge that pays before the big thesis does.</strong> The grand claim, that orbit beats Earth on cost, needs launch prices that do not exist yet. So the first customers are satellite operators whose data is already in space, such as Capella’s radar imagery, plus government buyers who pay for resilience rather than cheapness. Sacra describes the value as <a href="https://sacra.com/c/starcloud/">architectural rather than cost-based</a>. Customers pay for capability while everyone waits for Starship.</li><li><strong>Design for Lego, not cathedrals.</strong> Design principle five in the white paper reads like a developer’s prayer: grow from one container to many, profitable from the very first, with no large jumps in capital spending.</li><li><strong>Rent the front end, own the hard part.</strong> Crusoe handles the customer interface, billing and distribution. Starcloud keeps the spacecraft, the power and the cooling. The long-term plan is capacity reservations with the big cloud companies, structured like energy offtake agreements. In other words, contracted cash flows that can be financed far more cheaply than venture equity.</li><li><strong>Lock up the scarce input before anyone else notices.</strong> Johnston told TechCrunch that securing launch capacity is now one of the company’s biggest costs, with SpaceX’s Falcon 9 due to retire in 2028. Hence the 88,000-spacecraft filing and the pile of cash. When everyone needs the same rocket, the person holding the booking wins.</li></ol><p>If you have developed an energy project in Africa, every one of those moves should feel familiar. Most of us manage one or two, usually by accident. Starcloud does all eight on purpose, which is why it is worth $2.3 billion and most of us are worth a carefully worded LinkedIn update.</p><h3>Part 3: The translation, move by move</h3><h4>Lesson 1: Find the bottleneck that actually binds</h4><p><strong>The theory.</strong> Eli Goldratt’s Theory of Constraints, set out in his 1984 business novel <em>The Goal</em>, makes one point over and over: a system can only produce as much as its tightest bottleneck allows. Improve anything else and you have improved nothing. You have just made the queue more comfortable.</p><p><strong>The African evidence.</strong> Run through the usual suspects and most of them are not binding.</p><ul><li><strong>Sunshine is not the constraint.</strong> Africa has 60% of the world’s best solar resources and, when the IEA counted, <a href="https://knowledge.energyinst.org/new-energy-world/article?id=127558">1% of installed solar capacity</a>.</li><li><strong>Hardware is not the constraint.</strong> In Nigeria, a 420 W panel retails for about $60, and Ember reckons the diesel it saves can <a href="https://african.business/2025/08/energy-resources/nigeria-leapfrogs-egypt-as-african-solar-panel-imports-spike-by-60">repay it in about six months</a>. Stationary battery packs fell to <a href="https://about.bnef.com/insights/clean-energy/lithium-ion-battery-pack-prices/">$70/kWh in 2025</a>, 45% lower than a year earlier.</li><li><strong>Demand is not the constraint.</strong> Nigerians spend an estimated <a href="https://21stcenturychronicle.com/nigeria-spends-14bn-on-generators-fuel-yearly-akinwumi/">$14 billion a year on generators and fuel</a>, an IMF figure quoted by the African Development Bank’s president.</li></ul><p>What binds is money and trust. The IEA found that the cost of capital for utility-scale clean energy in Africa is <a href="https://www.iea.org/reports/financing-clean-energy-in-africa/executive-summary">at least two to three times higher</a> than in advanced economies and China, even though the technology risk is limited. The trust problem is the counterparty: whether the buyer, the utility or the government will still be paying in year eight.</p><p><strong>The Test:</strong> Write down your company’s binding constraint in one sentence. If the sentence contains “solar panels” or “more sales”, you have probably written down a symptom.</p><h4>Lesson 2: Leave the planet (the grid will do)</h4><p>Starcloud’s boldest move was not technical. It was geographical. It looked at a site where every hard thing lived, permits, interconnection, night, and simply left.</p><p>African energy has its own version, and it is happening without anyone’s permission. Ember tracked <a href="https://www.pv-magazine.com/2025/08/27/african-pv-imports-from-china-up-60-in-one-year-says-ember/">15 GW of Chinese solar panel imports into Africa</a> in the 12 months to June 2025, up 60% in a year. Outside South Africa, imports nearly tripled in two years, to 11.2 GW. Nigeria was the second-largest importer. Ember’s lead analyst says the volumes go <a href="https://www.latitudemedia.com/news/in-africa-the-first-signs-of-solar-adoption-at-scale/">well beyond what listed utility projects need</a>, which means most panels are going behind the meter: onto factories, shops, farms and homes that have quietly stopped waiting.</p><p>That bottom-up transition is the closest thing Africa has to orbit. The billion-dollar company does not fight the grid queue. It builds where the queue does not reach: captive plants, embedded generation, industrial parks, mini-grids, private wires.</p><p>One warning. Orbit has no landlord. African sites do. Leaving the grid is not the same as leaving the regulator, and a model that only works until the rules notice is not a strategy. It is a head start.</p><h4>Lesson 3: Ride a cost curve you are not paying for</h4><p><strong>The theory.</strong> In 1936, an aeronautical engineer called Theodore Wright noticed that every time cumulative aircraft production doubled, the cost per plane fell by a steady percentage. That pattern, now called Wright’s Law, describes solar panels and batteries rather well. Rupert Way and colleagues at Oxford <a href="https://www.oxfordmartin.ox.ac.uk/publications/empirically-grounded-technology-forecasts-and-the-energy-transition-2">tested forecasting methods on more than 50 technologies</a> and found that most energy models had underestimated how fast renewables would spread and overestimated what they would cost.</p><p><strong>The African version.</strong> Starcloud does not build rockets. It waits for SpaceX to make them cheap. The African equivalent is obvious and widely ignored: let manufacturers fight their price war on panels and cells, and build the business that turns cheap hardware into reliable, financed, maintained power. Battery pack prices have fallen <a href="https://www.energy-storage.news/?p=91801">93% since 2010</a>, according to BloombergNEF. Somebody else paid for that. You get to use it.</p><p>The trap is the one from <em>A Tale of Two Fifteen Billions</em>. Northvolt raised about $15 billion, tried to own the factory, and died. Building factories in Africa may be good industrial policy. It is a different company from the one that becomes worth a billion dollars selling energy, and it needs a very different kind of capital and patience.</p><h4>Lesson 4: Sell the wedge that pays today</h4><p><strong>The theory.</strong> In <em>The Prosperity Paradox</em>, Clayton Christensen, Efosa Ojomo and Karen Dillon argue that the innovations that build economies are market-creating. They turn non-consumers, people who could not afford or reach a thing at all, into customers. Their best example is Nigerian. Tolaram arrived in 1988 to sell Indomie noodles, invested <a href="https://ssir.org/articles/entry/making_space_for_prosperity">more than $350 million</a>, and ended up building its own logistics company, electricity, and water treatment, because the infrastructure it needed did not exist. The market pulled the infrastructure in behind it.</p><p><strong>The Starcloud parallel.</strong> Starcloud’s big cost thesis waits on Starship. Its first revenue does not. Satellites already in orbit, with their data already up there, will pay now to process it in place. That is a wedge: a customer who values what you do for a reason that has nothing to do with the grand thesis.</p><p><strong>The African version.</strong> My favourite example is not even an energy company. Gridless is a bitcoin miner, backed by Jack Dorsey’s Block, running in Kenya, Malawi and Zambia. Rural hydro mini-grids often generate more than the village can use, and the surplus is wasted. Gridless sits beside the plant as the buyer of that spare power. At one Kenyan pilot, the tariff to the community reportedly <a href="https://www.nbcnewyork.com/news/business/money-report/how-a-jack-dorsey-backed-bitcoin-miner-uses-a-volcano-in-kenya-to-turn-on-the-lights-in-rural-homes/5338896/">fell from 35 to 25 cents per kWh</a>. That is Starcloud’s logic in reverse. Instead of taking the computer to the sunshine, it takes a computer to the stranded electrons. You do not have to love bitcoin to admire the structure.</p><p>For most developers, the wedge is duller and better: the diesel bill. A factory spending a fortune on generators does not need to believe in your ten-year vision. It needs a cheaper kilowatt-hour next month.</p><h4>Lesson 5: Prove it small, then stage the bet</h4><p><strong>The theory.</strong> Real options thinking, developed in finance by Avinash Dixit and Robert Pindyck, treats each stage of an uncertain investment as buying the right, but not the obligation, to do the next one. Rita McGrath and Ian MacMillan turned this into management practice as <a href="https://hbr.org/1995/07/discovery-driven-planning">discovery-driven planning</a>: list your assumptions, then spend the least money possible testing the most dangerous one first.</p><p>Starcloud’s $3 million satellite was exactly that. It tested whether a data-centre chip could survive orbit before anyone spent a billion finding out.</p><p><strong>The cautionary tale.</strong> In May 2024, Microsoft and G42 announced a $1 billion geothermal-powered data centre at Olkaria in Kenya: 100 MW to start, 1 GW eventually. Kenya’s entire installed capacity is roughly 3 GW, and President Ruto later remarked that switching it on at full size would mean <a href="https://www.datacenterdynamics.com/en/news/microsoft-and-g42-data-center-in-kenya-stalled-due-to-lack-of-power-capacity/">switching off half the country</a>. By May 2026 it had stalled, reportedly after the companies asked the government to <a href="https://www.investmentmonitor.ai/news/microsoft-g42-kenya-data-centre-project-stalls/">commit to buying a fixed amount of cloud capacity every year</a> and the government declined.</p><p>Compare the two first steps. Starcloud’s cost $3 million and could fail quietly. Olkaria’s required a sovereign to underwrite a cloud contract. One of those is an option. The other is a dare.</p><p><strong>The Test.</strong> Before raising for the 50 MW plant, ask what the 500 kW version would prove. Usually it proves the thing that kills most African projects: whether the customer pays, on time, in a currency you can use.</p><h4>Lesson 6: Build in Lego</h4><p><strong>The theory.</strong> Bent Flyvbjerg at Oxford has spent decades building a database of large projects. In <em>How Big Things Get Done</em>, written with Dan Gardner, solar comes out as the best-behaved project type of all, with <a href="https://insideclimatenews.org/news/02022023/inside-clean-energy-lego-building/">average cost overruns of about 1%</a>. Transmission averaged 8%, wind 13%, fossil plants 16%, hydro dams 75% and nuclear 120%. The reason is modularity. A solar farm is one small thing repeated many times, so every unit teaches you how to build the next one better.</p><p>In a <a href="https://arxiv.org/pdf/2606.15757">2026 paper</a>, Flyvbjerg, Alexander Budzier and Maria Christodoulou go further. Across 11,011 projects, what they call “modular natives” (solar, transmission, wind and pipelines) overran by 12% on average, against 54% for everything else. The more important finding is about the bad outcomes. For modular natives, the risk is finite and predictable. For bespoke projects, the tail of disasters is so fat that the average stops meaning anything. In plain English: build in Lego and your worst day is survivable. Build a cathedral and your worst day has no floor.</p><p>Starcloud took this to its logical end. Its white paper wants the data centre to grow container by container, profitable from the first, with no big jumps in capital.</p><p><strong>The older theory.</strong> In 1967, Albert Hirschman published <em>Development Projects Observed</em>, based on about a dozen World Bank projects, <a href="https://www.brookings.edu/book/development-projects-observed/">including railway expansion in Nigeria</a>. He proposed the Hiding Hand: planners underestimate the difficulties, but they also underestimate their own creativity, so ambitious projects are often rescued by ingenuity nobody planned for. It is the most comforting idea in development economics. Hirschman himself used the Nigerian railway as his warning that trouble alone does not guarantee a rescue, and Flyvbjerg and Cass Sunstein later tested the idea against data and found the hand is <a href="https://arxiv.org/pdf/1509.01526">usually far less kind</a>.</p><p>My reconciliation is that: modular projects let creativity rescue you many times, in small pieces. Monolithic projects give you one chance to be rescued and many chances to be ruined.</p><h4>Lesson 7: Rent the front end, own the hard part, then sell the cash flows</h4><p><strong>The theory.</strong> David Teece’s <a href="https://doi.org/10.1016/0048-7333(86)90027-2">1986 paper on profiting from innovation</a> asked why inventors so often lose to imitators. His answer: the value goes to whoever controls the complementary assets, the distribution, service, financing and relationships a product needs to reach a customer. Tarun Khanna and Krishna Palepu add the emerging-markets twist. In markets like ours, many of those supporting pieces are missing altogether: no useful credit bureau, no long-term local-currency debt, no reliable collections. They call these <a href="https://hbr.org/2005/06/strategies-that-fit-emerging-markets">institutional voids</a>. The company that fills a void can end up owning it.</p><p><strong>The Starcloud move.</strong> Starcloud lets Crusoe own the customer interface and billing and keeps the hard engineering. It then plans to sell capacity on contracts that look like offtake agreements, so that cheaper capital can fund the hardware.</p><p><strong>The African version.</strong> Sun King is the best illustration I know. In 2025 it closed a <a href="https://www.citigroup.com/global/news/press-release/2025/citi-sun-king-securitization-deliver-solar-million-kenyans">$156 million Kenyan shilling securitisation</a>, described as the largest ever completed in Sub-Saharan Africa outside South Africa. Securitisation sounds intimidating. It simply means turning thousands of small future customer repayments into one asset that banks and development financiers can buy. Sun King says it has now raised $450 million in local currency across Kenya, Nigeria and Tanzania.</p><p>Read that as strategy. The void was long-term local-currency capital. Sun King filled it for itself, and the filling became the advantage. That is how you attack the binding constraint from Lesson 1. You do not complain about the cost of capital. You manufacture a cheaper version of it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*nCdjPMEGVuC7zBwy.jpeg" /></figure><h4>Lesson 8: Lock up what is scarce, and pick fights incumbents cannot join</h4><p><strong>The theory.</strong> Hamilton Helmer’s <em>7 Powers</em> lists the few ways a business earns lasting returns. Two matter here. A cornered resource is something valuable you control on preferential terms. Counter-positioning is a business model that incumbents cannot copy without damaging the business they already have.</p><p><strong>The Starcloud move.</strong> Launch capacity is getting scarce, so Starcloud is booking early and has filed for 88,000 spacecraft. That is cornering a resource. Its model also counter-positions against terrestrial data-centre landlords, whose whole value sits in land, grid connections and buildings.</p><p><strong>The African version.</strong> Ask what is genuinely scarce in your market. It is almost never panels. It is grid connection points with spare capacity, sites with clean title, licences, relationships with creditworthy buyers, good technicians and lines of local-currency debt. Lock those up early. Then look for the fight incumbents cannot join. A distribution company cannot happily sell solar to its best-paying industrial customers without eating its own tariff revenue. A diesel distributor cannot sell solar without cannibalising its fuel sales. Their hesitation is your opening.</p><h3>Part 4: The Starcloud test</h3><p>Here is the whole thesis as a test you can run on your own company, or on the next pitch deck that lands in your inbox. Give yourself one point for every row where you have a strong answer. Starcloud scores eight. I ran it on my own career and would prefer not to discuss the result.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/638/0*rdshYhRZroVnWzju.png" /></figure><p>Six or more and you are building something investors can price as a platform. Three or fewer and you are building projects, which is honourable, profitable and very rarely worth a billion dollars. As I argued in the Moove piece, own only the electricity and the market pays you a yield. Own the system around it and the market pays you a multiple.</p><h4>Coda: Okay, hear me out</h4><p>AI has an electricity habit. The IEA’s <em>Energy and AI</em> report projected that global data-centre electricity use would <a href="https://www.iea.org/reports/energy-and-ai/executive-summary">more than double to around 945 TWh by 2030</a>, and warned that grid constraints could delay roughly a fifth of planned projects. The machine can write your strategy in six seconds. Getting it connected remains a separate department.</p><p>This is also why I watch the Apple Series <em>Foundation</em> incorrectly. The visuals are extraordinary, and Lee Pace has perfected the art of standing in a room as though standing were an expensive form of government. The Galactic Empire is collapsing. Millions of people are in danger. The series gives us a weapon that destroys planets, and its visual-effects team has <a href="https://www.awn.com/vfxworld/chris-maclean-talks-mule-bokehs-and-visualizing-math-foundation-season-3">talked about the challenge</a> of making that much energy look convincing.</p><p>Everyone else is processing the collapse of civilisation. I am wondering whether the planet killer has a dedicated substation. What is the generation mix? Is there spinning reserve? Did the project reach financial close, or did the Emperor simply issue a sovereign guarantee? Morally appalling, certainly. But may I see the single-line diagram?</p><p>Yes, I know it is fiction. I would prefer that planets remain available for housing, agriculture and the occasional holiday. What captures me is the abundance underneath the story: a civilisation that is not rationing electricity between a steel mill and a housing estate. Magic is often just infrastructure whose development history was cut from the television adaptation.</p><p>AI needs enormous amounts of electricity. That need is pushing companies like Starcloud to rethink where computing should happen. Better AI may help us design better grids, storage and materials. Those energy systems will power more AI. At some point, however intelligent the machines become, somebody will still need to ask whether the transformer has been ordered.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*jOE6oV4vP-ecUZ6K.jpeg" /></figure><p>That is the part of the future I care about, and it is why this series exists. Physics may make a technology possible. The boring machinery of origination, development, finance, construction and operations decides whether anybody gets to use it. Africa has the sunshine, the demand and, increasingly, the hardware. What it has been short of is enough organised stupidity pointed at the right constraint.</p><p>Somewhere in Africa, a group of reasonably intelligent adults is looking at a problem.</p><p>One of them clears his throat.</p><p>“Okay, hear me out.”</p><p>Everybody should leave.</p><p>Instead, somebody opens Excel.</p><p>That is when the billion-dollar company begins.</p><p><strong><em>- S</em></strong></p><h4>Disclaimers</h4><p><strong><em>P.S:</em> </strong>I work in African energy for a living, so I have a commercial interest in the argument that African energy can be built as a platform rather than a pile of projects. Please read Lesson 7 the way you would read a barber’s essay on the benefits of haircuts. I hold no position in Starcloud. I have never worked with Starcloud. I have no non-public information about Starcloud, and on current evidence, Starcloud has no information about me at all. Every number above is public and linked, including the ones I wish were bigger.</p><p><strong><em>P.P.S:</em> The confession.</strong> I wrote “prove it small, early and cheap” with a completely straight face. Philip Johnston designed, built and launched a satellite in 21 months. In my line of work, 21 months is roughly how long it takes to agree which version of the term sheet is the latest version. I have watched a land title search outlive two phones and a change of government.</p><p>If you are Philip Johnston: hello. No pressure. The friendship paperwork can wait, but I would like the single-line diagram for Starcloud-3.</p><p><strong><em>P.P.P.S.</em></strong> If your project has been “shovel-ready” since 2017, please check on the shovel. It may have retired. It may now be teaching younger shovels about the importance of managing expectations.</p><p><strong><em>P.P.P.P.S.</em></strong> If the financial model only works when the currency strengthens, the customer pays early and nothing breaks for twenty years, congratulations. You have developed infrastructure for the Garden of Eden.</p><p><strong><em>P.P.P.P.P.S:</em></strong> “We are raising $20 million to prove the concept” is an extraordinary sentence. Imagine saying that at home. “Darling, I need twenty million dollars.” “For what?” “To find out whether I’m talking nonsense.”</p><p>If this was useful, forward it to the one developer you know who is still waiting for the grid. They have time to read it. If it was not, the unsubscribe button works perfectly. I test it monthly, usually after rereading my own drafts.</p><p><strong>References</strong></p><p><strong>Starcloud</strong></p><ul><li>Feilden, Oltean and Johnston (2024), <a href="https://starcloudinc.github.io/wp.pdf">Why we should train AI in space</a>, Starcloud white paper v1.03</li><li>HPCwire (30 March 2026), <a href="https://www.hpcwire.com/off-the-wire/starcloud-raises-170m-series-a-at-1-1b-valuation-led-by-benchmark-and-eqt-ventures/">Starcloud raises $170M Series A at $1.1B valuation</a></li><li>TechCrunch (21 August 2026), <a href="https://techcrunch.com/2026/08/21/starcloud-raises-200-million-for-orbital-data-centers-as-launch-options-dry-up/">Starcloud raises $250 million for orbital data centers as launch options dry up</a></li><li>TechCrunch (11 May 2026), <a href="https://techcrunch.com/2026/05/11/there-arent-enough-rockets-for-space-data-centers-cowboy-space-raised-275-million-to-build-them/">Cowboy Space raised $275 million to build rockets</a></li><li>Sacra, <a href="https://sacra.com/c/starcloud/">Starcloud company profile</a></li><li>Crusoe (22 October 2025), <a href="https://www.crusoe.ai/resources/newsroom/crusoe-to-become-first-cloud-operator-in-space-through-partnership-with-starcloud">Crusoe to become first cloud operator in space</a></li><li>CNBC (10 December 2025), <a href="https://www.cnbc.com/2025/12/10/nvidia-backed-starcloud-trains-first-ai-model-in-space-orbital-data-centers.html">Starcloud trains first AI model in space</a></li></ul><p><strong>African energy evidence</strong></p><ul><li>IEA (2023), <a href="https://www.iea.org/reports/financing-clean-energy-in-africa/executive-summary">Financing Clean Energy in Africa</a></li><li>Energy Institute, <a href="https://knowledge.energyinst.org/new-energy-world/article?id=127558">Building Africa’s solar power</a>, citing the IEA Africa Energy Outlook 2022</li><li>pv magazine (27 August 2025), <a href="https://www.pv-magazine.com/2025/08/27/african-pv-imports-from-china-up-60-in-one-year-says-ember/">African PV imports from China up 60% in one year</a></li><li>African Business (August 2025), <a href="https://african.business/2025/08/energy-resources/nigeria-leapfrogs-egypt-as-african-solar-panel-imports-spike-by-60">Nigeria leapfrogs Egypt as African solar panel imports spike</a></li><li>Latitude Media, <a href="https://www.latitudemedia.com/news/in-africa-the-first-signs-of-solar-adoption-at-scale/">In Africa, the first signs of solar adoption at scale</a></li><li>BloombergNEF (9 December 2025), <a href="https://about.bnef.com/insights/clean-energy/lithium-ion-battery-pack-prices/">Lithium-ion battery pack prices fall to $108/kWh</a></li><li>Energy-Storage.news (10 December 2025), <a href="https://www.energy-storage.news/?p=91801">Battery pack prices fell 8% since last year</a></li><li>21st Century Chronicle, <a href="https://21stcenturychronicle.com/nigeria-spends-14bn-on-generators-fuel-yearly-akinwumi/">Nigeria spends $14bn on generators, fuel yearly</a></li><li>BusinessDay, <a href="https://businessday.ng/energy/power/article/running-on-empty-inside-nigerias-multibillion-dollar-generator-economy/">Inside Nigeria’s multibillion-dollar generator economy</a></li><li>DatacenterDynamics, <a href="https://www.datacenterdynamics.com/en/news/microsoft-and-g42-data-center-in-kenya-stalled-due-to-lack-of-power-capacity/">Microsoft and G42 data centre in Kenya stalled</a></li><li>Investment Monitor (11 May 2026), <a href="https://www.investmentmonitor.ai/news/microsoft-g42-kenya-data-centre-project-stalls/">Microsoft and G42 Kenya data centre project stalls</a></li><li>Citi (2025), <a href="https://www.citigroup.com/global/news/press-release/2025/citi-sun-king-securitization-deliver-solar-million-kenyans">Sun King securitisation to deliver solar to over a million Kenyans</a></li><li>CNBC via NBC New York, <a href="https://www.nbcnewyork.com/news/business/money-report/how-a-jack-dorsey-backed-bitcoin-miner-uses-a-volcano-in-kenya-to-turn-on-the-lights-in-rural-homes/5338896/">How a Block-backed bitcoin miner uses a volcano in Kenya</a></li></ul><p><strong>Theory</strong></p><ul><li>Goldratt, E. (1984), <em>The Goal</em>, North River Press</li><li>Wright, T. P. (1936), “Factors affecting the cost of airplanes”, <em>Journal of the Aeronautical Sciences</em></li><li>Way, R., Ives, M., Mealy, P. and Farmer, J. D. (2022), <a href="https://www.oxfordmartin.ox.ac.uk/publications/empirically-grounded-technology-forecasts-and-the-energy-transition-2">Empirically grounded technology forecasts and the energy transition</a>, <em>Joule</em></li><li>Christensen, C., Ojomo, E. and Dillon, K. (2019), <em>The Prosperity Paradox</em>, Harper Business; <a href="https://ssir.org/articles/entry/making_space_for_prosperity">excerpt on Tolaram</a></li><li>Dixit, A. and Pindyck, R. (1994), <em>Investment under Uncertainty</em>, Princeton University Press</li><li>McGrath, R. and MacMillan, I. (1995), <a href="https://hbr.org/1995/07/discovery-driven-planning">Discovery-driven planning</a>, <em>Harvard Business Review</em></li><li>Flyvbjerg, B. and Gardner, D. (2023), <em>How Big Things Get Done</em>; overrun figures via <a href="https://insideclimatenews.org/news/02022023/inside-clean-energy-lego-building/">Inside Climate News</a></li><li>Flyvbjerg, B., Budzier, A. and Christodoulou, M. (2026), <a href="https://arxiv.org/pdf/2606.15757">Towards a theory of modular natives</a></li><li>Hirschman, A. (1967), <a href="https://www.brookings.edu/book/development-projects-observed/">Development Projects Observed</a>, Brookings</li><li>Flyvbjerg, B. and Sunstein, C. (2016), <a href="https://arxiv.org/pdf/1509.01526">The principle of the malevolent hiding hand</a></li><li>Teece, D. (1986), <a href="https://doi.org/10.1016/0048-7333(86)90027-2">Profiting from technological innovation</a>, <em>Research Policy</em></li><li>Khanna, T., Palepu, K. and Sinha, J. (2005), <a href="https://hbr.org/2005/06/strategies-that-fit-emerging-markets">Strategies that fit emerging markets</a>, <em>Harvard Business Review</em></li><li>Helmer, H. (2016), <em>7 Powers: The Foundations of Business Strategy</em></li></ul><p><strong>Coda</strong></p><ul><li>IEA (2025), <a href="https://www.iea.org/reports/energy-and-ai/executive-summary">Energy and AI: executive summary</a></li><li>AWN (2025), <a href="https://www.awn.com/vfxworld/chris-maclean-talks-mule-bokehs-and-visualizing-math-foundation-season-3">Chris MacLean on visualising the maths of <em>Foundation</em> season 3</a></li><li>OpenAI (8 September 2026), <a href="https://openai.com/index/navier-stokes-solution/">On the Navier–Stokes Millennium Prize Problem</a></li><li>The Neuron (9 September 2026), <a href="https://theneuron.ai/news/inside-openais-navierstokes-claim-the-proof-the-ai-effort-and-the-credit-fight/">Inside OpenAI’s Navier–Stokes claim</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a95681527bc4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The HUAWEI Thesis: How to Build a Billion-Dollar Energy Company in Africa]]></title>
            <link>https://kaykluz.medium.com/the-huawei-thesis-how-to-build-a-billion-dollar-energy-company-in-africa-8ee2174f3d60?source=rss-2dfae1fe39fa------2</link>
            <guid isPermaLink="false">https://medium.com/p/8ee2174f3d60</guid>
            <category><![CDATA[china]]></category>
            <category><![CDATA[business]]></category>
            <category><![CDATA[energy]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[business-strategy]]></category>
            <dc:creator><![CDATA[Kaykl.uz]]></dc:creator>
            <pubDate>Wed, 16 Sep 2026 11:15:13 GMT</pubDate>
            <atom:updated>2026-09-16T11:15:13.510Z</atom:updated>
            <content:encoded><![CDATA[<p>Somewhere on my laptop there is a bill of quantities for a solar and storage plant in Ghana. A few weeks ago I was going through it line by line, doing the thing I do where I pretend to check the arithmetic but am really checking the brands.</p><ul><li>Inverters: Huawei.</li><li>Batteries: Huawei.</li><li>Monitoring platform: Huawei.</li></ul><p>Nothing unusual in that. Open a commercial and industrial solar tender in Lagos, Accra or Abidjan and the odds are decent that at least one of the three expensive boxes has the little red flower on it. In my corner of the market Huawei has become what I privately call the cheap premium: no longer the cheapest kit in the tender, because there are now six Chinese brands underneath it on price, but the kit the lender’s technical adviser stops asking questions about.</p><p>What made me put the laptop down was what happened next. I picked up my phone to make tea (I do not need my phone to make tea; this is a separate problem) and YouTube, which has correctly identified me as a man of simple pleasures, served me a forty-minute documentary on how the United States killed Huawei in 2019.</p><p>Killed. Past tense.</p><p>I sat in my kitchen holding a document that said otherwise.</p><p>Then I did the maths on my own week. The base stations along the road to every site I have ever visited in Ghana or Nigeria: Huawei, in all likelihood, because Huawei built <a href="https://www.atlanticcouncil.org/blogs/africasource/the-digital-infrastructure-imperative-in-african-markets/">about seven in ten of the continent’s 4G networks</a>. The colleague in Johannesburg who sends me voice notes from a phone with no Google on it: Huawei. The inverter on page four of that bill of quantities: Huawei, <a href="https://www.woodmac.com/press-releases/global-pv-inverter-shipments-grew-by-10-in-2024-to-589-gwac/">first in the world by shipments for the tenth year running</a>.</p><p>So either the documentary was wrong, or I was dealing in ghosts.</p><p>It turns out the documentary was right, and that is the interesting part. The United States really did take almost everything from Huawei between December 2018 and September 2020.</p><p>BUT….</p><p>The company is bigger now. Not in spite of the sequence; partly because of it. The mechanics of how are, I think, the most useful set of lessons this series has produced so far for anyone trying to build a billion-dollar energy company in a market that keeps trying to take things away from them.</p><p>I did what I do when my inbox and the internet disagree. I went and read everything.</p><p>This Substack is reader-supported. To receive new posts and support my work, consider becoming a free or paid subscriber.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Hie38srF9DsyS9X6.jpeg" /><figcaption>Commissioning day at a Justrite Superstore in Lagos State, 2024. Behind the gate: solar-plus-storage plant, Huawei Digital Power kit, financed and owned as an IPP by Empower New Energy, built by Powercell. Phase I covered ten stores — 3.05 MWp of PV and 6 MWh of batteries, $6.5 million. Phase II takes the portfolio to twenty-one sites and $11 million, the largest retail solar programme in West Africa. <strong>I worked this project at Empower</strong></figcaption></figure><h3>Part 1: The company that keeps turning up</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*efSKEELUDgR0xZfw.jpeg" /><figcaption>“Huawei Logo.” 1000 Logos. <a href="https://1000logos.net/huawei-logo/">https://1000logos.net/huawei-logo/</a></figcaption></figure><p>Huawei built roughly 70 percent of Africa’s 4G networks and about half its 3G, per the <a href="https://www.atlanticcouncil.org/blogs/africasource/the-digital-infrastructure-imperative-in-african-markets/">Atlantic Council</a> and <a href="https://merics.org/en/comment/african-networks-smartphones-and-surveillance">Merics</a>. In Nigeria alone the count is reported at <a href="https://weetracker.com/2025/10/17/the-10-chinese-companies-building-the-africas-connectivity-infrastructure/">around 27,500 sites and 10,000 kilometres of fibre</a>.</p><p>The rectifier in the cabinet, the battery under it, the solar array on the pole, the controller that decides when the diesel runs: Huawei sells all of that too, and has done since before anyone in Lagos said the words “energy transition” out loud.</p><p>The inverter on the roof: Huawei shipped <a href="https://www.woodmac.com/press-releases/global-pv-inverter-shipments-grew-by-10-in-2024-to-589-gwac/">176 gigawatts of PV inverters in 2024</a>, first in the world for the tenth consecutive year, with Wood Mackenzie singling out Africa as one of its strong regions. Huawei and Sungrow between them now hold 55 percent of the planet’s inverter market and nobody else has more than five.</p><p>The visible part of Huawei (Phones) got amputated. The invisible parts compounded. And the invisible parts, it turns out, had been prepared for the amputation since 2003.</p><h3>Part 2: The amputation, in dates</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*5ASMO6pOFwhUX5qM.jpeg" /><figcaption>Huawei’s Bantian campus, Shenzhen. Official Huawei photo. Credit: Huawei</figcaption></figure><p>I want to lay the sequence out plainly, because the strategy only makes sense against the exact timing of what was taken and when.</p><p><strong>1 December 2018.</strong> Huawei’s chief financial officer, Meng Wanzhou, is arrested in Vancouver on a US extradition request during a stopover. She will spend almost three years under house arrest in Canada.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/705/0*1GQpyiJJk_mTXKct.jpeg" /><figcaption>Meng Wanzhou leaving court in Vancouver after her December 2018 arrest on a US extradition request. She spent almost three years under house arrest.</figcaption></figure><p><strong>16 May 2019.</strong> The US Commerce Department adds Huawei and dozens of affiliates to the Entity List. American companies now need a licence to sell it anything, and the default answer is no.</p><p><strong>19 May 2019.</strong> Google <a href="https://www.aljazeera.com/economy/2019/5/20/google-suspends-some-business-with-huawei-after-us-blacklist">suspends the transfer of hardware, software and technical services</a> to Huawei. Future Huawei phones outside China will ship without the Play Store, Gmail or YouTube. Within the week Intel and Qualcomm <a href="https://m.gsmarena.com/newscomm-37114p2.php">begin cutting chip supply</a>, and ARM, whose architecture underlies almost every smartphone processor, follows.</p><p><strong>15 May 2020.</strong> Commerce amends the foreign direct product rule specifically to catch Huawei’s workaround. The press release title says the quiet part loudly: <a href="https://www.commerce.gov/news/press-releases/2020/05/commerce-addresses-huaweis-efforts-undermine-entity-list-restricts">“Commerce Addresses Huawei’s Efforts to Undermine Entity List”</a>. Any chip made anywhere in the world with American tools now needs a licence if it is Huawei’s design.</p><p><strong>17 August 2020.</strong> The rule is <a href="https://www.wiley.law/alert-Commerce-Tightens-Huawei-Restrictions-Aims-to-Close-Loopholes">widened again</a> to close the remaining gaps and 38 more affiliates are listed. TSMC, which had been manufacturing Huawei’s most advanced processors, stops taking orders in September.</p><p><strong>November 2020.</strong> Huawei sells its Honor sub-brand to a consortium for <a href="https://www.phonearena.com/news/smartphone-shipments-by-huawei-decline-sharply-in-2021_id138129">a reported fifteen billion dollars plus</a>, so that Honor can go back to buying Qualcomm chips and Google services.</p><p><strong>Now the damage.</strong></p><p>Huawei shipped <a href="https://www.phonearena.com/news/smartphone-shipments-by-huawei-decline-sharply-in-2021_id138129">240.6 million phones in 2019</a>, second in the world behind Samsung, ahead of Apple. In the second quarter of 2020, running on stockpiled chips, it was <a href="https://www.cnbc.com/2020/10/30/huawei-q3-smartphone-shipments-plunge-as-u.s.-sanctions-continue-to-bite.html">briefly the largest smartphone maker on earth</a>. By the fourth quarter it had <a href="https://www.cnbc.com/2021/01/28/huawei-q4-smartphone-shipments-plunge-41percent-as-us-sanctions-bite.html">fallen to sixth</a>. In 2021 it shipped 35 million, a fall of 81.6 percent, and its global share went from 15 percent to 3. In 2022 it was <a href="https://www.scmp.com/tech/big-tech/article/3270682/tech-war-huawei-faced-difficult-days-global-smartphone-sales-plunged-executive-says">tenth, on 28.1 million</a>. Group revenue <a href="https://www.globaltimes.cn/page/202508/1342077.shtml">fell 28.6 percent in 2021</a>. Richard Yu, who runs the consumer business, later admitted the humiliation in one line: the company that led the world in 5G networks <a href="https://www.scmp.com/tech/big-tech/article/3270682/tech-war-huawei-faced-difficult-days-global-smartphone-sales-plunged-executive-says">could not sell a 5G phone</a>.</p><p>The industry consensus at that point was that Huawei’s handset business would wind down like Nokia’s had. Analysts wrote the obituaries. I would have signed one.</p><p>Here is the same sequence as a ledger, because I think in ledgers and because the right-hand column is this whole essay.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/613/0*vTiuthUUZMVy30E_.png" /></figure><p>Read the last row again. Huawei was locked out of the markets that pay the most per site, and left alone in the markets that have the most sites.</p><h3>Part 3: The letter</h3><p>On the morning of 17 May 2019, the day after the Entity List announcement, an engineer called <strong>He Tingbo</strong> sent an email to the staff of HiSilicon, Huawei’s chip subsidiary. The internet had never heard of her. Within a day, most of Chinese tech had.</p><p>The letter said, in effect: we have spent years designing chips we hoped never to use. Tonight they become the main product. She described the unit’s job as <a href="https://www.usnews.com/news/top-news/articles/2026-05-25/huaweis-chip-queen-etches-her-name-in-chinas-tech-folklore">“building a backup lifeline for Huawei and for the whole country”</a>. Inside Huawei those chips had a nickname. <strong>Spare tyres.</strong></p><p>I went looking for where the spare tyres came from, and this is where I uncovered the first strategic note.</p><p>In 2003, <strong>Ren Zhengfei</strong>, the founder, put <strong>He Tingbo</strong> in charge of chip development with <a href="https://en.wikipedia.org/wiki/He_Tingbo">an annual budget of 400 million dollars</a> and an instruction that reads, from here, like paranoia.</p><p><em>Design the parts we currently buy from Americans. Not to sell them. To have them, in case the day comes when the Americans will not sell.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*1qSSnH1BBsQFU5pA.jpeg" /><figcaption>Huawei Founder — Ren Zhengfei</figcaption></figure><p>HiSilicon was formally created the following year. The first generation chips flopped; the K3V1 and K3V2 were widely mocked. Ren’s response, according to <a href="https://eu.36kr.com/en/p/3825779103290243">accounts of the period</a>, was not to cut the budget. It was to complain that she was spending it too slowly.</p><p>Do the arithmetic at the 2003 run rate alone and you get more than six billion dollars spent on a component nobody was allowed to ship, over sixteen years, before a single sanctioned quarter. The real figure is higher because the budget grew.</p><p>This bit reorganised my thinking about the whole company. The discovery is not that Huawei survived the ban. Lots of companies survive things. The discovery is that survival had a line item. <em>There was a budget for the flat tyre in the year the tyres were new.</em></p><p>I have sat through a lot of investment committees on African energy companies. I have built the models for some of them, which is worse, because I know where the bodies are. Every one of them had a single point of failure it had collectively decided not to think about. One inverter supplier. One currency. One offtaker. One bank. One import licence. One fuel contract. The spare tyre was never in the budget because the budget was already too tight to include a tyre we might never use, and every board I have ever presented to would have said so.</p><p>Ren’s answer to that objection, as far as I can reconstruct it, was three words: <a href="https://eu.36kr.com/en/p/3825779103290243">it must be spent</a>.</p><p>If you have read the Foundry Problem series on this blog you will recognise the shape. Africa’s energy sector imports the foundational layer and builds on top of it, and calls the dependence “efficiency” right up until the supplier changes its mind. Huawei looked at the same dependence in 2003, with a much bigger balance sheet and a much nastier counterparty, and decided the efficiency was fake.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*w03cCg-PV8hrWWE3.png" /><figcaption>Huawei: ‘Chip Queen’ He Tingbo’s Legacy in Tech. Source: <a href="https://www.deccanherald.com/business/companies/meet-he-tingbo-the-huaweis-chip-queen-who-has-her-name-etched-in-chinas-tech-folklore-4015212">https://www.deccanherald.com/business/companies/meet-he-tingbo-the-huaweis-chip-queen-who-has-her-name-etched-in-chinas-tech-folklore-4015212</a></figcaption></figure><h3>Part 4: Country and city</h3><p>To understand why the spare tyres worked when they were finally fitted, you have to go back further, to the first time Huawei lost a supplier.</p><p>Huawei was founded in Shenzhen in 1987 with <a href="https://www.companieshistory.com/huawei/">21,000 yuan of registered capital</a>, which was less than three thousand dollars then, as a sales agent for a Hong Kong company that made small telephone switchboards for hotels. In 1990 that supplier sold itself, and Huawei’s business <a href="https://www.scribd.com/document/882000330/Huawei-History">ended overnight</a>. Ren had to build a switch or close the company. Three years later Huawei shipped the C&amp;C08, a 2,000-line digital exchange and later a 10,000-line one, the most capable switch made in China at the time.</p><p>I did not know until this month what the name means. The C&amp;C stands for <a href="https://en.eeworld.com.cn/news/qrs/eic544385.html">Country and City</a>: the countryside surrounds the city. That was the marketing plan, written into the product code. Ericsson, Alcatel and Lucent owned Beijing and Shanghai. Huawei went to county exchanges the foreigners could not be bothered with, sold cheaper kit that broke more often, and left engineers on site to fix it until it didn’t. By 1998 it was the largest supplier of digital switches in the country.</p><p>Then it did the same thing to the world.</p><p>In November 2000 Ren travelled to Africa with China’s vice premier, and the contracts followed the trip: <a href="https://csis-website-prod.s3.amazonaws.com/s3fs-public/legacy_files/files/publication/130215_competitiveness_Huawei_casestudy_Web.pdf">twenty million dollars in Ethiopia in 2003, a 200 million dollar CDMA network in Nigeria in 2005</a>, then Ghana, Kenya, Morocco, Mauritius and Congo in 2006. Wharton was writing case studies about <a href="https://knowledge.wharton.upenn.edu/article/huawei-technologies-a-chinese-trail-blazer-in-africa/">Huawei as a trailblazer in Africa</a> in 2009, four years before it launched a phone anyone in Europe wanted. Africa and Russia were the training ground. Europe came after.</p><p>The purest expression of the doctrine is a product most of my readers have driven past without noticing. In 2017 Huawei launched <a href="https://www.huawei.com/en/news/2020/2/huawei-ruralstar-pro-solution">RuralStar</a>: a base station for villages of a thousand people, built by throwing out everything a base station is supposed to be. The tower became a pole. The microwave backhaul became a wireless relay hop off a neighbouring site, so no line of sight and no dish. The diesel generator became six solar panels and a battery. Deployment cost fell by more than half and the operator’s payback <a href="https://www.huawei.com/en/huaweitech/publication/winwin/31/how-three-stars-are-making-life-better">dropped from ten years to three</a>. By the end of its first year there were <a href="https://developingtelecoms.com/telecom-technology/wireless-networks/7593-huawei-ruralstar-self-contained-cell-site.html">two thousand of them across a dozen countries</a>, including Ghana, Nigeria, Kenya and Côte d’Ivoire.</p><p>Notice what RuralStar is. It is a telecom product where the energy engineering is the innovation. The radio was fine. The site was the problem, and the site was mostly a power problem.</p><p>In the Citadel piece I argued that Citadel built its financial layer first and bought physical assets later, and that Africa forces you to build backwards.</p><p>Huawei is the company that built backwards on purpose. It started in the ugliest markets with the ugliest problems, learned to keep a network alive where the diesel does not arrive, and only then walked into Vodafone’s office. When the ban came, the ugly layer was the one that survived, because nobody in Washington had thought to sanction a base station in Kaduna.</p><h3>Part 5: Huawei came with a bank attached</h3><p>There is a second reason Huawei won the countryside, and it is the reason most relevant to the readers of this blog who spend their lives trying to close a project finance.</p><p>In 2004 Huawei signed a <a href="https://phys.org/news/2019-05-huawei-key-beneficiary-china-subsidies.html">ten billion dollar credit line with the China Development Bank</a> whose purpose was to lend money to Huawei’s customers so they could buy Huawei equipment. In 2009 it was <a href="https://www.lightreading.com/network-technology/huawei-s-lucky-number-30b">tripled to thirty billion</a>. The Export-Import Bank of China added its own facilities. The Center for American Progress went through the public record and found that between 1997 and 2019 Chinese policy banks had lent <a href="https://www.washingtonpost.com/politics/2020/11/19/huawei-has-hidden-advantage-its-push-global-dominance-now-us-is-trying-counter-it/">14.8 billion dollars across 99 Huawei-related projects</a> around the world, and that Africa was the largest borrower, with 57 loans worth 4.7 billion. The Wall Street Journal’s wider estimate of state support, in grants, credit, tax breaks and cheap land, was <a href="https://techblog.comsoc.org/2019/12/26/wsj-chinas-financial-support-aided-huaweis-rise-to-1-telecom-vendor-in-the-world/">as much as 75 billion dollars</a>. Huawei disputes the framing and says the CDB line has rarely been more than ten percent drawn, which may be true and is also beside the point.</p><p>The point was made by Fred Hochberg, who used to run the US Export-Import Bank, and who is not a man inclined to compliment Huawei. What Huawei did, he said, was make sure that <a href="https://techblog.comsoc.org/2019/12/26/wsj-chinas-financial-support-aided-huaweis-rise-to-1-telecom-vendor-in-the-world/">“when they made a bid, it came with financing terms”</a> the competition could not match.</p><p>Ask the Nigerian operators and you get the same answer from the buyer’s side. The chairman of the licensed telecoms operators’ association told DW that Huawei’s kit <a href="https://www.homelandsecuritynewswire.com/dr20220210-africa-embraces-huawei-technology-despite-security-concerns">“comes with terms that are very attractive to operators”</a>, and that it is simply easy to work with them.</p><p>Now stand this next to two earlier pieces in this series.</p><p>In the MOOVE Thesis I argued the machine was an IPP with wheels: buy the asset with debt, place it with an operator who cannot fund the capex, collect at source. In the Kobo360 anti-template the lesson was harsher: offtaker credit quality only counts once it is structured into the funding, and Kobo360 had better credit than Moove and died anyway because it never was.</p><p>Huawei did both, in 2004, to telecom operators, and the operator never had to structure anything. The quotation and the term sheet were the same document. The operator’s credit was underwritten by a policy bank before the operator had finished reading the bill of quantities.</p><p>I have lost tenders on tariff to competitors whose plant was worse than mine, because their bid had a bank attached and mine had a bank “in advanced discussions”. Every developer reading this has.</p><p>The African energy company that wins the next decade will not be the one with the lowest cents per kilowatt-hour. It will be the one that walks into the factory with the Afreximbank, AFC or local-bank term sheet already stapled to page one.</p><h3>Part 6: The energy business was standing next to the base station</h3><p>This is where my day job enters the story, and where I had initially the wrong mental model.</p><p>I had assumed Huawei “diversified” into energy the way conglomerates diversify: a strategy offsite, a new division, a logo.</p><p>It is not what happened.</p><p>The energy business was standing next to every base station Huawei had ever sold, and it had the largest line in the customer’s operating budget.</p><p>Africa has <a href="https://mymotherlode.com/news/world/10791470/africas-cellphone-towers-turn-to-solar-as-diesel-costs-surge.html/amp">roughly 500,000 telecom towers</a>. The majority run on diesel. CrossBoundary Energy puts it at <a href="https://crossboundaryenergy.com/how-energy-can-be-digital-infrastructures-new-strategic-advantage/">about 70 percent</a>, with energy accounting for 30 to 60 percent of tower opex in remote areas. Nigerian operators now spend <a href="https://mymotherlode.com/news/world/10791470/africas-cellphone-towers-turn-to-solar-as-diesel-costs-surge.html/amp">around 400 million dollars a year</a> keeping their sites powered after subsidy removal sent diesel up as much as 200 percent. Vodacom’s energy bill across its African operations was <a href="https://www.rochesterfirst.com/news/tech-news/ap-africas-cellphone-towers-turn-to-solar-as-diesel-costs-surge/">300 million dollars in 2025</a>.</p><p>Huawei sold the radio. Then it sold the rectifier the radio needed. Then the battery the rectifier charged. Then the solar array that charged the battery, the controller that decided when the diesel ran, and eventually the AI that predicts the weather and schedules the whole site’s power like a scarce resource: Orange Côte d’Ivoire reported <a href="https://technologymagazine.com/news/mwc-2026-orange-huawei-drive-africa-digital-growth">a 45 percent reduction in rural site downtime</a> on that system at MWC this year. RuralStar bundled all of it into one pole.</p><p>Then Huawei took that catalogue, the power electronics, the thermal management, the battery management, the cloud that watches it all, out of the telecom cabinet and put it on a factory roof. That is FusionSolar. Huawei Digital Power was carved out as a standalone company <a href="https://baike.baidu.com/en/item/Huawei%20Digital%20Power%20Technology%20Co.,%20Ltd./1421549">in June 2021</a>, which is to say a year after the foreign direct product rule and while the smartphone business was in freefall. Its revenue went from <a href="https://electroiq.com/stats/huawei-statistics/">55.2 billion yuan in 2023 to 68.7 billion in 2024</a> to <a href="https://eu.36kr.com/en/p/3748132636295944">77.3 billion in 2025</a>, which at today’s rates is around 10.8 billion dollars. For scale, Tesla’s energy generation and storage division, the one Wall Street calls the good half of Tesla, <a href="https://thenextweb.com/news/huawei-digital-power-revenue-tesla-energy-division">turned over 12.8 billion in 2025</a>.</p><p>So when I see the little red flower on page four of a Ghanaian bill of quantities, what I am actually looking at is a telecom site power product, twenty years old, that grew up and moved out.</p><p>Two things follow for anyone building in African energy, and they point in opposite directions.</p><p>The first is the lesson. Huawei did not find a new market. It noticed that the energy bill was the biggest line in its existing customer’s opex and that it already owned the cabinet the bill lived in. If you run a solar plant next to an industrial load, the base station, the cold room, the borehole pump and the data centre are all standing there with the same bill, and you already own the cabinet.</p><p>The second is the warning. The company that taught the world to start in the countryside is, in African C&amp;I energy, now the city. It is the incumbent on the bill of quantities, the safe answer, the brand the lender does not question. The countryside is open. Somebody is going to do to Huawei’s inverter business in Africa what Huawei did to Ericsson in Sichuan, and the six cheaper brands underneath it in every tender I see are the early signal.</p><h3>Part 7: The thief who pays three times more than he collects</h3><p>I need to deal with the intellectual property story, because it is the part of the Huawei narrative that most of my readers have absorbed without checking, and because the checking produced the single most surprising number in this piece.</p><p>In February 2020 the US Justice Department expanded its case against Huawei into a racketeering indictment alleging decades of trade secret theft. Huawei denies it. I am not a court and I am not going to adjudicate it here. What I can read is the public record of who pays whom.</p><p>In June 2019, one month after the Entity List, Huawei told Verizon it was using <a href="https://www.insurancejournal.com/news/national/2021/07/14/622576.htm">more than 230 Huawei patents and owed more than a billion dollars</a> in licensing fees. In February 2020 it sued. In July 2021, days into a federal jury trial in Texas, <a href="https://www.bloomberg.com/news/articles/2021-07-12/huawei-says-it-has-settled-patent-lawsuits-with-verizon-in-texas">Verizon settled on confidential terms</a>. Verizon, which is barred from putting a single Huawei box in its network, pays Huawei.</p><p>The portfolio behind that is not something you assemble in a hurry. Huawei filed its first Chinese patent in 1995. It has been the <a href="https://www.malaymail.com/news/money/mediaoutreach/2025/11/11/advancing-innovation-with-openness-huawei-holds-the-6th-innovation-and-ip-forum/426323">largest user of the international PCT filing system since 2014</a>, according to WIPO’s own assistant director general, with 6,600 international applications in 2024 alone. At the end of 2025 it held <a href="https://www.huawei.com/en/annual-report/2025">165,000 active granted patents</a>. Its 5G patents are licensed on <a href="https://www.huawei.com/en/news/2025/11/2025ipr-advancing-innovation">more than 2.7 billion devices</a>, its Wi-Fi patents on 1.2 billion, its video codec patents on 3.2 billion. About fifty Fortune Global 500 companies are licensees. It has signed <a href="https://www.huawei.com/en/ipr">more than 260 licence and cross-licence agreements</a> with American, European, Japanese and Korean companies, and it charges <a href="https://www.mobileworldlive.com/verizon/verizon-huawei-settle-us-patent-dispute/">a maximum of 2.50 dollars per 5G phone</a>. Licensing brought in about 630 million dollars in 2024.</p><p>Huawei’s chief legal officer said at its IP forum in November 2025 that over the years the company has <a href="https://www.huawei.com/en/news/2025/11/2025ipr-advancing-innovation">paid nearly three times more in royalties than it has received</a>. Qualcomm’s own 10-K records <a href="https://www.sec.gov/Archives/edgar/data/804328/000172894921000076/Financial_Report.xlsx">1.8 billion dollars</a> booked from Huawei in a single quarter of 2020 under a settlement and a new licence, signed while the Entity List was already in force.</p><p>Whatever you believe about the indictment, that is not the balance sheet of a company that gets its technology for free. It is the balance sheet of the industry’s most reliable payer, who also happens to write a lot of the rules: Huawei submitted <a href="https://www.insidenova.com/online_features/press_releases/advancing-innovation-with-openness----huawei-holds-the-6th-innovation-and-ip-forum/article_d9d58b9d-fcd5-5d25-a8bd-3a954daaf93e.html">more than 10,000 technical contributions to standards bodies</a> in 2024.</p><p>And the rules are where the toll booth is. A company that spends thirty years helping to write the 5G standard collects on phones it will never sell, in countries that have banned it, from carriers that cannot buy its equipment.</p><p>The energy translation is one I have been circling for two years without naming. In African energy nobody owns the standard. Not the PPA template, not the metering data format, not the settlement protocol, not the interconnection spec.</p><p>When I built the Africa Energy Software Map last month, the finding that stung was that 474 software products claim to serve the market and only 20 could evidence a deployment. The reason is that there is no standard for them to plug into, so every one of them is a bespoke integration and every integration dies with its champion.</p><p>Whoever writes the spec that the lenders adopt, the way Huawei’s polar-coding patents became part of the 5G spec that Verizon’s own suppliers had to follow, collects on plants they never built. In the NVIDIA piece I called CUDA a standard with a toll booth. Huawei’s patent portfolio is the same thing from the other side of the ban.</p><h3>Part 8: If you cannot build the best chip, build the biggest room</h3><p>I want to spend a moment on the chips, because the Ascend story is where the doctrine stops being defensive and becomes the most interesting infrastructure play in the world, and because it is, underneath, a telecom trick.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7qjRcL8QRzhpyyrmblqqcA.png" /><figcaption>HiSilicon die lifted from a Mate 60 Pro. TechInsights found a Kirin 9000s made by SMIC on a 7nm-class process the sanctions were supposed to make impossible. <a href="https://www.bloomberg.com/news/features/2023-09-04/look-inside-huawei-mate-60-pro-phone-powered-by-made-in-china-chip">https://www.bloomberg.com/news/features/2023-09-04/look-inside-huawei-mate-60-pro-phone-powered-by-made-in-china-chip</a></figcaption></figure><p>When TSMC stopped in September 2020, Huawei was left with chip designs it could not manufacture. For three years its flagship phones ran on Qualcomm chips <a href="https://www.phonearena.com/news/kirin-9000s-is-really-kirin-9000-says-tipster_id151514">deliberately restricted to 4G</a> under limited licences. Then, in the last week of August 2023, while the US Commerce Secretary was on a visit to Beijing, Huawei put the Mate 60 Pro on sale without announcing what was inside it. TechInsights bought one, took it apart, and found a <a href="https://www.techinsights.com/blog/techinsights-finds-smic-7nm-n2-huawei-mate-60-pro">Kirin 9000s made by SMIC on a 7-nanometre process</a> that the sanctions were supposed to make impossible. One TechInsights analyst called it a <a href="https://en.prothomalo.com/science-technology/qtoviiozas">“slap in the face”</a> for Washington. In 2025 Huawei was the <a href="https://eu.36kr.com/en/p/3748132636295944">largest smartphone seller in China</a>, on 16.4 percent of the market.</p><p>The AI chips are the same story with more zeros. The Ascend 910C <a href="https://www.trendforce.com/news/2025/09/18/news-huawei-unveils-ascend-950-with-in-house-hbm-in-2026-touts-superpod-to-rival-nvidia/">went into mass production in the first quarter of 2025</a>. It is two older 910B dies packaged together, and on paper it delivers about 80 percent of an Nvidia H100, a chip from 2022. On its own it would not worry anyone in Santa Clara.</p><p>So Huawei does not sell it on its own. In March 2025 it launched the <a href="https://www.huawei.com/en/news/2025/9/hc-xu-keynote-speech">Atlas 900 A3 SuperPoD</a>: 384 of those chips wired together across sixteen cabinets so tightly that, from the software’s point of view, they behave like one computer. The interconnect that makes that possible, UnifiedBus, is the thing Huawei spent thirty years learning to do in telecom: move enormous amounts of data between boxes that are individually unremarkable.</p><p>Then, at Huawei Connect in September 2025, the rotating chairman Eric Xu did something Western executives rarely do. He conceded that China would lag in semiconductor manufacturing <a href="https://scalebytech.com/huawei-reveals-ascend-950-960-and-970-chips-to-power-top-ai-supercomputing-clusters">“for a relatively long time”</a>, and then laid out a roadmap built entirely on that concession: <a href="https://www.rcrwireless.com/20250922/ai-infrastructure/huawei-ai-chips">Ascend 950PR in the first quarter of 2026 with Huawei’s own high-bandwidth memory, 950DT in the fourth, 960 in 2027, 970 in 2028</a>, each roughly doubling the last. The Atlas 950 SuperPoD, due in the fourth quarter of this year, links 8,192 chips. The Atlas 950 SuperCluster links 64 of those into <a href="https://convergedigest.com/huawei-unveils-next-gen-ascend-chips-superpods-and-unifiedbus-interconnect/">more than 520,000 chips across 10,000 cabinets</a>, and the 960 version scheduled for 2027 goes to a million. Huawei also said it would open-source the software stack, the CANN compiler and the Mind toolchains, by the end of 2025, which is the CUDA playbook run by a company that could not buy a GPU.</p><p>Whether the room is really faster than the chip is an argument for people who read benchmarks for pleasure. What matters for this series is what happened next on the demand side, because a supplier’s strategy is only a strategy if a customer turns up.</p><p>On 24 April 2026 DeepSeek released V4 with <a href="https://tech-insider.org/deepseek-v4-huawei-ascend-1-6-trillion-parameter-moe-2026/">day-zero support on Ascend</a>. DeepSeek <a href="https://aiproem.substack.com/p/part-2-what-deepseek-v4-means-for">gave Huawei early access to the model and did not give it to Nvidia</a>. Huawei says part of the smaller V4-Flash was trained on its chips; the conservative reading, which I share, is that frontier training still happens on Nvidia and what moved to Huawei is inference, which is a different problem and a much bigger market.</p><p>On 17 July Moonshot released Kimi K3, a 2.8 trillion parameter model. On 19 July it <a href="https://www.pymnts.com/news/artificial-intelligence/2026/moonshot-halts-new-kimi-k3-subscriptions-demand-overwhelms-compute/">paused new subscriptions</a> because, in its own words, <a href="https://www.pymnts.com/news/artificial-intelligence/2026/moonshot-halts-new-kimi-k3-subscriptions-demand-overwhelms-compute/">“our GPUs are feeling it”</a>. A Chinese lab with a model competitive with the best American ones could not serve the customers who wanted to pay for it.</p><p>And two weeks ago Bloomberg reported that DeepSeek plans to install <a href="https://winbuzzer.com/2026/09/09/deepseek-reportedly-plans-160000-plus-huawei-chips-model-serving-xcxwbn/">at least 160,000 Ascend 950DT chips at a gigawatt-scale data centre it is building in Ulanqab, Inner Mongolia</a>, a hardware order reported at <a href="https://www.bighatgroup.com/blog/china-ai-weekly-2026-09-13/">around 18 billion yuan</a>. Neither company has confirmed it. Bloomberg’s sources also say Huawei can only make 950DTs in the low hundreds of thousands this year, and that the memory supply is a bottleneck, so filling the order could take more than a year. That is the honest state of the spare tyre in September 2026: the demand has arrived and the factory is the constraint.</p><p>Two observations from the energy side, since a gigawatt-scale campus is an energy project first and a compute project second.</p><p>The first is that the same catalogue that keeps a base station alive in Ulanqab’s -30 degree winters, the rectifiers, the batteries, the liquid cooling, the site controller, sits under the SuperCluster, and Huawei sells that too. If you read my Zuckerberg piece you know I think superintelligence is priced in kilowatt-hours. Huawei appears to agree, and it is the only company in the world that sells the chip, the interconnect, the cloud and the substation from one catalogue.</p><p>The second is the lesson. In the NVIDIA piece, move two was “sell the stack, not the component.” Huawei ran that move from the opposite direction. It could not get the component. So it sold the room.</p><p>The African energy builder is in Huawei’s 2020 position, permanently. You will never get the cheapest turbine, the cheapest capital, or the largest plant. Every asset you build will be small and individually unimpressive: a one megawatt rooftop in Ota, a two megawatt biomass boiler in Tema, a four hundred kilowatt mini-grid in Plateau.</p><p>The 910C is not the product. The SuperPoD is.</p><p>The portfolio, wired together with one metering platform, one O&amp;M fleet, one SPV structure and one lender, is the thing a bank can underwrite as a single asset. Moove got to 2.1 billion dollars with 42,000 individually unremarkable cars and one bus.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*yHpq0e5pMz-7de1P.jpeg" /></figure><h3>Part 9: Did it work?</h3><p>I promised myself I would not write “the ban backfired” because it is the lazy version, and because it is only half true. So here are both columns.</p><p><strong>The ban worked.</strong> Huawei’s global smartphone business is gone outside China: of roughly 48 million phones shipped in 2024, <a href="https://chinamade.tech/blog/huawei-comeback-story">about 46 million stayed home</a>. It is out of American, British, Swedish and Australian 5G networks and has been <a href="https://www.rusi.org/explore-our-research/publications/commentary/huaweis-spare-tyre-20-and-limits-us-sanctions">designated on the Entity List with around 160 affiliates</a>. Its CFO lost three years of her life. It sold Honor. Its revenue fell by more than a quarter in 2021. Its cloud business, the one that is supposed to monetise the chips, saw <a href="https://www.cnbc.com/2026/03/31/huawei-annual-report-2025-cloud-computing-revenue.html">external revenue fall 3.5 percent in 2025</a> while Alibaba’s grew 36. And the company still cannot make enough advanced chips to fill a single customer’s order.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*3Olu-Wik8xABRoTH.jpeg" /><figcaption>Canalys worldwide smartphone shipments, Q1 2018–Q4 2020. Huawei (including Honor) falls out of the top five for the first time in six years after the Entity List and Google cutoff. Canalys chart reproduced in Ars Technica, “From first to sixth: Huawei’s phone business tanks thanks to US sanctions,” January 2021. <a href="https://arstechnica.com/gadgets/2021/01/huaweis-phone-sales-plummet-42-percent-in-q4-2020-its-now-in-sixth-place/">https://arstechnica.com/gadgets/2021/01/huaweis-phone-sales-plummet-42-percent-in-q4-2020-its-now-in-sixth-place/</a></figcaption></figure><p><strong>The company won.</strong> Huawei’s 2025 revenue was <a href="https://www.huawei.com/en/news/2026/3/annual-report-2025">880.9 billion yuan</a>, about 127.5 billion dollars, its second-highest year ever and <a href="https://eu.36kr.com/en/p/3748132636295944">1.05 billion yuan short of the 2020 record</a>. Net profit was 68 billion yuan, up 8.6 percent. R&amp;D was <a href="https://www.huawei.com/en/annual-report/2025">192.3 billion yuan, 21.8 percent of revenue</a>, which is roughly 27.5 billion dollars and <a href="https://axis-intelligence.com/huawei-statistics/">about two and a half times what Ericsson and Nokia spent combined</a>. Over the decade it has spent 1.382 trillion yuan on research. 114,000 of its people, 53.7 percent of the workforce, are in R&amp;D. ICT infrastructure did 375 billion yuan, the consumer business 344.5 billion, digital power 77.3 billion, and the car business <a href="https://www.trendforce.com/news/2026/04/01/news-huawei-2025-r-cloud-revenue-slips/">45 billion, up 72 percent</a>. The Ascend developer ecosystem passed <a href="https://whbl.com/2026/03/31/huawei-posts-2-2-growth-in-annual-revenue/">four million</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/754/0*z0srfu8Wp4Gxk2rL.png" /><figcaption>Huawei group revenue, billion yuan. Peak 2020, collapse 2021, second-highest year in 2025 at 880.9 billion yuan (~$127.5bn). <a href="https://www.statista.com/statistics/233071/revenue-of-huawei/">https://www.statista.com/statistics/233071/revenue-of-huawei/</a></figcaption></figure><p>And the annual report that carries those numbers was signed by Huawei’s rotating chairwoman, Sabrina Meng. Meng Wanzhou. The CFO who was arrested in Vancouver in December 2018.</p><p>The ban did what bans do. It removed the visible part of the company from the places that could see it. Then the invisible parts kept compounding, because the invisible parts had a spare tyre, a bank attached to every bid, a training ground in the hardest markets on earth, and a set of patents that made the people banning it into paying customers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*_CRVtmnuzES3zu0J.jpeg" /><figcaption>Sabrina Meng (Meng Wanzhou), Fortune Most Powerful Women Asia 2025. Source: Fortune, “Meng Wanzhou (Sabrina Meng).” <a href="https://fortune.com/ranking/most-powerful-women-asia/2025/meng-wanzhou-sabrina-meng/">https://fortune.com/ranking/most-powerful-women-asia/2025/meng-wanzhou-sabrina-meng/</a></figcaption></figure><h3>Part 10: The Spare Tyre Doctrine</h3><p>Here is what I am taking from all of this for the African energy builder. Seven moves, in the order Huawei ran them, each with the test I would now put to a management team before I put their numbers in a model.</p><h3>Move 1. Start in the country, not the city.</h3><p>Huawei’s first breakthrough was a switch named after the strategy, sold to county exchanges Ericsson could not be bothered with. Its first foreign markets were Russia and Africa. Europe was the graduation, not the entry.</p><p>In African energy the city is the utility-scale tender, the DFI-backed IPP, the multinational’s headquarters plant with a global procurement team. The country is the tier-two industrial cluster, the agro-processor, the tower company, the cold chain, the loads that are too small for the big developers and too complex for the banks.</p><p><strong><em>The test:</em></strong><em> would MASDAR bid for this customer? If yes, you are in the city. Go find the county. Benchmark: Huawei was the largest switch supplier in China within five years of the C&amp;C08. Give yourself five years to become the default supplier in one neglected segment before you touch a capital-city tender.</em></p><h3>Move 2. Sell the site, not the box.</h3><p>RuralStar cut deployment cost by more than half not by making a cheaper radio but by redesigning everything around the radio: the tower, the backhaul, the power. The energy engineering was the product.</p><p>Do not sell a solar plant. Sell a powered site: generation, storage, cooling, monitoring, maintenance and financing as one contract with one counterparty. This is the argument I made in the “Decarbonize Your Factory” piece and Huawei is the proof it scales to a hundred billion dollars.</p><p><strong><em>The test:</em></strong><em> can your customer buy any one piece of your offer without you? If yes, you are a box, and boxes are commoditised by the second Chinese brand to arrive.</em></p><h3>Move 3. Bring the bank with you.</h3><p>Ten billion dollars from CDB in 2004, thirty billion by 2009, 57 loans into Africa alone. The bid came with financing terms. The customer’s credit was underwritten before the customer asked.</p><p>Arrange the DFI, export credit or local bank line before the tender, not after the award. The quotation and the term sheet should be the same document. This is the Kobo360 lesson (credit only counts once it is structured into the funding) executed by the vendor rather than the borrower.</p><p><strong><em>The test:</em></strong><em> does your bid contain financing terms your competitor cannot match? If not, you are competing on tariff, and tariff is a race to zero that Huawei’s cheaper cousins will win.</em></p><h3>Move 4. Pack the spare tyre before the flat.</h3><p>Four hundred million dollars a year from 2003 on chips nobody was allowed to ship. Sixteen years. The founder’s only complaint was the pace of spending.</p><p>Name the one input a counterparty can weaponise against you.</p><p>An import licence. A currency. A single inverter OEM. A single fuel supplier. A single offtaker. A single bank. Then fund the alternative before you need it, and protect that line in the budget from the board. Huawei’s own stated rule is that R&amp;D never falls below ten percent of revenue; it is at 21.8 percent now. An energy company does not need that number. It needs a fixed slice of EBITDA, decided in advance, that pays for the second supplier’s qualification, the local-currency facility, the fuel contract you are not using yet.</p><p><strong><em>The test:</em></strong><em> name the supplier who could end you with one email. What did you spend on their replacement last year? If the answer is zero, you do not have a tyre, you have a hope.</em></p><h3>Move 5. Own the standard, then rent it.</h3><p>Ten thousand standards contributions a year. 165,000 patents. 2.50 dollars per 5G phone from carriers that cannot buy a Huawei box. And, crucially, three times more paid out than taken in, because the point of a standard is that everyone uses it, including you.</p><p>Write the PPA template the lenders adopt. Write the metering data specification the regulator references. Write the settlement protocol the aggregators plug into. Contribute to the grid codes and the net billing rules (I have spent more hours on NERC’s than I would like to admit). Once your document is the standard, you collect on plants you did not build, from developers who compete with you.</p><p><strong><em>The test:</em></strong><em> if you shut down tomorrow, does anyone still have to use your document? If not, you own a business. Not a standard.</em></p><h3>Move 6. When you cannot build the best unit, build the biggest room.</h3><p>The 910C is 80 percent of a three-year-old Nvidia chip. The SuperPoD is 384 of them behaving as one. The SuperCluster is 520,000. The interconnect is thirty years of telecom engineering. The concession that the unit will never be the best was made out loud, on stage, by the chairman.</p><p>Accept that every asset you build in Africa will be small and sub-scale. Make the portfolio the product. The bus is your data layer, your contract layer, your single SPV, your single O&amp;M fleet, your single lender relationship. RAMCO’s instinct in Nigeria, whatever becomes of it, is exactly this instinct. Moove’s 42,000 cars are exactly this instinct.</p><p><strong><em>The test:</em></strong><em> can a lender underwrite your portfolio as one asset? If each plant needs its own diligence, its own lawyer and its own model, you have 384 chips and no bus.</em></p><h3>Move 7. Let the ban do the marketing.</h3><p>DeepSeek gave Huawei early access and not Nvidia because Nvidia was not allowed in the room. Kimi K3 ran out of GPUs and the only domestic pool large enough to help was Huawei’s. Every restriction Washington added became a customer acquisition channel for the Ascend ecosystem. Four million developers did not arrive because the chip was best. They arrived because the alternative was banned.</p><p>In African energy the ban is the grid. Every collapse, every subsidy removal, every diesel spike, every FX crunch is a demand signal and a channel. Design the business so that when the crisis hits, customers are routed to you by the crisis itself. This is the argument of “The Demand Was Never Small” and Huawei is the proof at national scale.</p><p><strong><em>The test: </em></strong><em>when the grid collapsed last quarter, did your pipeline grow or shrink? If it shrank, you are selling into good times, and Africa does not sell those.</em></p><p>Here it is as one table, because I know at least a third of you scrolled here first.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/542/0*x4TcFFdtDL2Gdjni.png" /></figure><p>A note on sequence, because Huawei’s order matters more than its list.</p><ul><li>Moves 1 to 3 were built between 1993 and 2009.</li><li>Move 4 ran in parallel from 2003 and was invisible until 2019.</li><li>Move 5 started with a patent in 1995 and only became a business after the ban made licensing the one revenue line nobody could sanction.</li><li>Moves 6 and 7 were only possible after 2020, and only because 1 to 5 were already there.</li></ul><p>If you are building in African energy, 1 to 3 are your first three years, 4 and 5 start on day one at whatever scale you can afford, 6 becomes possible once move 3 has given you a repeatable capital stack, and 7 is a design principle rather than a phase.</p><h3>Part 11: What does not cross the water</h3><p>I owe you the other side, because a blueprint that pretends the balance sheet is copyable is a fantasy, and I have written enough of those in pitch decks to recognise one.</p><p><strong>The state.</strong> Whatever the true number between Huawei’s “ten percent drawn” and the Journal’s 75 billion, no African energy company is going to get a thirty billion dollar policy bank line, and nobody is going to instruct MTN to buy seventy percent of its equipment from you, which is what Beijing <a href="https://www.washingtonpost.com/politics/2020/11/19/huawei-has-hidden-advantage-its-push-global-dominance-now-us-is-trying-counter-it/">reportedly did</a> for Huawei and ZTE during China’s 4G build. Move 3 transfers. The scale of move 3 does not.</p><p><strong>The patience.</strong> Huawei is privately held, with no listed shareholders asking why 400 million dollars a year is going into a product that does not ship. Your investors will ask. Move 4 transfers only if you decide the size of the tyre before the first board meeting and write it into the shareholders’ agreement.</p><p><strong>The scale.</strong> 27.5 billion dollars a year of R&amp;D and 114,000 engineers is not a number, it is a country. The doctrine scales down, though, and this is the part I find hopeful: your spare tyre is not a fab. It is a second qualified inverter OEM, a naira facility, a fuel supply contract you are not drawing on, a second offtaker in the same cluster. Those cost hundreds of thousands, not billions, and they are still almost never in the budget.</p><p><strong>The controversy.</strong> Huawei is a company that half the world’s governments do not trust with their networks, on grounds that range from documented to asserted, and the racketeering indictment is unresolved. I have not tried to settle any of that here and you should not read this piece as a verdict on it. Being unwelcome in half the world is itself a constraint on the blueprint. Read it for the sequencing, not the geopolitics.</p><p><strong>The home market.</strong> A billion domestic customers who cannot buy Google either is a floor no African company has. The nearest equivalent is a regional trade bloc that actually functions, and I will believe in that one when a Ghanaian inverter clears Seme border without a phone call.</p><p><em>-S</em></p><p><strong>P.S. — Disclosures first.</strong></p><ul><li>In my day job I write Huawei equipment into bills of quantities, I have negotiated against Huawei’s pricing, and I have lost tenders to bids that came with a bank attached. <strong>Huawei has never paid me a kobo, and neither, for the record, has anyone else for writing this.</strong></li><li>I worked on <a href="https://www.empowernewenergy.com/post/empower-new-energy-delivers-first-of-its-kind-solar-pv-battery-energy-storage-system-for-justrite">Empower New Energy’s Justrite solar-plus-storage programme</a> in Nigeria. Empower financed it. Powercell built it. Huawei Digital Power supplied the kit. I have used a commissioning photograph from that project in Part 6. <strong>Huawei did not pay for the photograph or the essay.</strong></li><li>If you think I have the ban wrong, that it worked and I am reading a survival story into a subsidy story, say so in the comments, publicly, so we can both be embarrassed later with a timestamp. I have changed my mind once already writing this and would prefer to do it in company.</li><li>If it was useful, send it to the one person you know who is currently carrying a single-supplier risk and calling it a partnership. If it was not, the unsubscribe link is at the bottom and it works even without Google services.</li></ul><h3>References</h3><p><strong>Huawei filings and statements</strong></p><ul><li>Huawei, <a href="https://www.huawei.com/en/annual-report/2025">2025 Annual Report</a> and <a href="https://www.huawei.com/en/news/2026/3/annual-report-2025">press release, 31 March 2026</a></li><li>Huawei, <a href="https://www.huawei.com/en/ipr">Innovation and Intellectual Property</a></li><li>Huawei, <a href="https://www.huawei.com/en/news/2025/11/2025ipr-advancing-innovation">6th Innovation and IP Forum, 11 November 2025</a>; <a href="https://www.insidenova.com/online_features/press_releases/advancing-innovation-with-openness----huawei-holds-the-6th-innovation-and-ip-forum/article_d9d58b9d-fcd5-5d25-a8bd-3a954daaf93e.html">PR Newswire version</a>; <a href="https://www.malaymail.com/news/money/mediaoutreach/2025/11/11/advancing-innovation-with-openness-huawei-holds-the-6th-innovation-and-ip-forum/426323">Malay Mail version with WIPO remarks</a></li><li>Eric Xu, <a href="https://www.huawei.com/en/news/2025/9/hc-xu-keynote-speech">Huawei Connect 2025 keynote</a></li><li>Ren Zhengfei, <a href="https://www.huawei.com/au/news/2019/huawei-founder-ren-zhengfei-interview-with-chinese-media">interview with Chinese media, May 2019</a></li><li>Huawei, <a href="https://www.huawei.com/en/news/2020/2/huawei-ruralstar-pro-solution">RuralStar Pro launch, February 2020</a>; <a href="https://www.huawei.com/en/huaweitech/publication/winwin/31/how-three-stars-are-making-life-better">How three stars are making life better</a></li></ul><p><strong>2025 results coverage</strong></p><ul><li>CNBC, <a href="https://www.cnbc.com/2026/03/31/huawei-annual-report-2025-cloud-computing-revenue.html">Huawei’s cloud computing revenue dropped in 2025</a></li><li>Reuters via WHBL, <a href="https://whbl.com/2026/03/31/huawei-posts-2-2-growth-in-annual-revenue/">Huawei posts slower annual revenue growth of 2.2%</a></li><li>TrendForce, <a href="https://www.trendforce.com/news/2026/04/01/news-huawei-2025-r-cloud-revenue-slips/">Huawei 2025 R&amp;D spending reaches record 192.3B yuan</a></li><li>36Kr, <a href="https://eu.36kr.com/en/p/3748132636295944">Huawei’s 2025 revenue hits 880.9 billion yuan</a></li><li>Axis Intelligence, <a href="https://axis-intelligence.com/huawei-statistics/">Huawei statistics 2026</a></li><li>Global Times, <a href="https://www.globaltimes.cn/page/202508/1342077.shtml">Huawei H1 2025 results</a></li></ul><p><strong>The bans</strong></p><ul><li>US Commerce Department, <a href="https://www.commerce.gov/news/press-releases/2020/05/commerce-addresses-huaweis-efforts-undermine-entity-list-restricts">press release, 15 May 2020</a></li><li>Wiley, <a href="https://www.wiley.law/alert-Commerce-Tightens-Huawei-Restrictions-Aims-to-Close-Loopholes">Commerce tightens Huawei restrictions, August 2020</a></li><li>Al Jazeera, <a href="https://www.aljazeera.com/economy/2019/5/20/google-suspends-some-business-with-huawei-after-us-blacklist">Google suspends some business with Huawei</a></li><li>RUSI, <a href="https://www.rusi.org/explore-our-research/publications/commentary/huaweis-spare-tyre-20-and-limits-us-sanctions">Huawei’s ‘Spare Tyre 2.0’ and the limits of US sanctions</a></li></ul><p><strong>The collapse and recovery</strong></p><ul><li>CNBC, <a href="https://www.cnbc.com/2021/01/28/huawei-q4-smartphone-shipments-plunge-41percent-as-us-sanctions-bite.html">Q4 2020 shipments plunge 41%</a>; <a href="https://www.cnbc.com/2020/10/30/huawei-q3-smartphone-shipments-plunge-as-u.s.-sanctions-continue-to-bite.html">Q3 2020</a></li><li>PhoneArena, <a href="https://www.phonearena.com/news/smartphone-shipments-by-huawei-decline-sharply-in-2021_id138129">2021 shipments down 81%</a></li><li>SCMP, <a href="https://www.scmp.com/tech/big-tech/article/3270682/tech-war-huawei-faced-difficult-days-global-smartphone-sales-plunged-executive-says">Huawei faced difficult days as global sales plunged</a></li><li>Light Reading, <a href="https://www.lightreading.com/5g/huawei-2021-sales-decimated-by-us-sanctions">Huawei 2021 sales decimated by US sanctions</a></li><li>TechInsights, <a href="https://www.techinsights.com/blog/techinsights-finds-smic-7nm-n2-huawei-mate-60-pro">SMIC 7nm in the Mate 60 Pro</a>; Reuters via Prothom Alo, <a href="https://en.prothomalo.com/science-technology/qtoviiozas">teardown report</a></li><li>China Made &amp; Tech, <a href="https://chinamade.tech/blog/huawei-comeback-story">Huawei’s comeback</a></li></ul><p><strong>HiSilicon and the spare tyre</strong></p><ul><li>Reuters via US News, <a href="https://www.usnews.com/news/top-news/articles/2026-05-25/huaweis-chip-queen-etches-her-name-in-chinas-tech-folklore">Huawei’s ‘chip queen’</a></li><li>36Kr, <a href="https://eu.36kr.com/en/p/3825779103290243">A female Huawei executive has gone viral</a></li><li>Wikipedia, <a href="https://en.wikipedia.org/wiki/He_Tingbo">He Tingbo</a></li></ul><p><strong>Ascend, SuperPoDs and the AI demand</strong></p><ul><li>Converge Digest, <a href="https://convergedigest.com/huawei-unveils-next-gen-ascend-chips-superpods-and-unifiedbus-interconnect/">Huawei unveils next-gen Ascend chips, SuperPoDs and UnifiedBus</a></li><li>TrendForce, <a href="https://www.trendforce.com/news/2025/09/18/news-huawei-unveils-ascend-950-with-in-house-hbm-in-2026-touts-superpod-to-rival-nvidia/">Huawei unveils Ascend 950</a></li><li>RCR Wireless, <a href="https://www.rcrwireless.com/20250922/ai-infrastructure/huawei-ai-chips">Huawei outlines roadmap for Ascend AI chips</a></li><li>ScaleByTech, <a href="https://scalebytech.com/huawei-reveals-ascend-950-960-and-970-chips-to-power-top-ai-supercomputing-clusters">Huawei reveals Ascend 950, 960 and 970</a></li><li>Tech Insider, <a href="https://tech-insider.org/deepseek-v4-huawei-ascend-1-6-trillion-parameter-moe-2026/">DeepSeek V4 on Huawei Ascend</a>; AI Proem, <a href="https://aiproem.substack.com/p/part-2-what-deepseek-v4-means-for">What DeepSeek V4 means for Huawei and Nvidia</a></li><li>PYMNTS, <a href="https://www.pymnts.com/news/artificial-intelligence/2026/moonshot-halts-new-kimi-k3-subscriptions-demand-overwhelms-compute/">Moonshot halts new Kimi K3 subscriptions</a>; SCMP, <a href="https://www.scmp.com/tech/article/3361172/kimi-k3-developer-suspends-new-subscriptions-amid-compute-constraints">Kimi K3 developer suspends new subscriptions</a></li><li>WinBuzzer on Bloomberg, <a href="https://winbuzzer.com/2026/09/09/deepseek-reportedly-plans-160000-plus-huawei-chips-model-serving-xcxwbn/">DeepSeek reportedly bets on 160,000-plus Huawei chips</a>; Big Hat Group, <a href="https://www.bighatgroup.com/blog/china-ai-weekly-2026-09-13/">China AI Weekly, 13 September 2026</a></li></ul><p><strong>Patents and licensing</strong></p><ul><li>Insurance Journal (Reuters), <a href="https://www.insurancejournal.com/news/national/2021/07/14/622576.htm">Verizon, Huawei reach settlement</a>; Bloomberg, <a href="https://www.bloomberg.com/news/articles/2021-07-12/huawei-says-it-has-settled-patent-lawsuits-with-verizon-in-texas">settlement mid-trial</a></li><li>Mobile World Live, <a href="https://www.mobileworldlive.com/verizon/verizon-huawei-settle-us-patent-dispute/">Verizon, Huawei settle US patent dispute</a></li><li>Qualcomm, <a href="https://www.sec.gov/Archives/edgar/data/804328/000172894921000076/Financial_Report.xlsx">Form 10-K FY2021</a></li></ul><p><strong>Africa: telecom and towers</strong></p><ul><li>Atlantic Council, <a href="https://www.atlanticcouncil.org/blogs/africasource/the-digital-infrastructure-imperative-in-african-markets/">The digital infrastructure imperative in African markets</a></li><li>Merics, <a href="https://merics.org/en/comment/african-networks-smartphones-and-surveillance">African networks, smartphones and surveillance</a></li><li>WeeTracker, <a href="https://weetracker.com/2025/10/17/the-10-chinese-companies-building-the-africas-connectivity-infrastructure/">Ten Chinese companies building Africa’s connectivity</a></li><li>DW via Homeland Security Newswire, <a href="https://www.homelandsecuritynewswire.com/dr20220210-africa-embraces-huawei-technology-despite-security-concerns">Africa embraces Huawei technology</a></li><li>Developing Telecoms, <a href="https://developingtelecoms.com/telecom-technology/wireless-networks/7593-huawei-ruralstar-self-contained-cell-site.html">RuralStar self-contained cell site</a>; GSMA, <a href="https://www.gsma.com/solutions-and-impact/technologies/networks/gsma_resources/case-study-ruralstar-huawei-safaricom/">RuralStar Safaricom case study</a></li><li>Technology Magazine, <a href="https://technologymagazine.com/news/mwc-2026-orange-huawei-drive-africa-digital-growth">Orange and Huawei at MWC 2026</a></li><li>AP, <a href="https://mymotherlode.com/news/world/10791470/africas-cellphone-towers-turn-to-solar-as-diesel-costs-surge.html/amp">Africa’s cellphone towers turn to solar</a>; CrossBoundary Energy, <a href="https://crossboundaryenergy.com/how-energy-can-be-digital-infrastructures-new-strategic-advantage/">Energy as digital infrastructure’s advantage</a>; GSMA, <a href="https://www.gsma.com/solutions-and-impact/connectivity-for-good/mobile-for-development/uncategorized/tower-power-africa-energy-challenges-and-opportunities/">Tower power Africa</a></li></ul><p><strong>Africa: phones</strong></p><ul><li>Canalys via CNBC Africa, <a href="https://www.cnbcafrica.com/media/7756131450891/canalys-africas-smartphone-market-grew-7-to-192mn-units-in-q225">Africa’s smartphone market Q2 2025</a></li><li>Omdia, <a href="https://omdia.tech.informa.com/pr/2025/may/africa-extends-its-smartphone-growth-streak-but-2025-projected-to-see-moderate-3percent-growth">Africa Q1 2025</a></li><li>Intelpoint on StatCounter, <a href="https://intelpoint.co/insights/in-february-2025-samsung-29-80-achieved-the-largest-mobile-market-share-to-usage-across-africa/">Africa usage share</a>; <a href="https://intelpoint.co/insight-category/smartphone/">South Africa and Ghana</a></li></ul><p><strong>Energy and inverters</strong></p><ul><li>Wood Mackenzie, <a href="https://www.woodmac.com/press-releases/global-pv-inverter-shipments-grew-by-10-in-2024-to-589-gwac/">Global PV inverter shipments 2024</a>; <a href="https://www.woodmac.com/press-releases/solar-inverter-ranking-h1-2025/">H1 2025 rankings</a></li><li>TNW, <a href="https://thenextweb.com/news/huawei-digital-power-revenue-tesla-energy-division">Huawei Digital Power rivals Tesla’s energy division</a></li><li>ElectroIQ, <a href="https://electroiq.com/stats/huawei-statistics/">Huawei statistics 2024 segments</a>; Baidu Baike, <a href="https://baike.baidu.com/en/item/Huawei%20Digital%20Power%20Technology%20Co.,%20Ltd./1421549">Huawei Digital Power Technology</a></li></ul><p><strong>History and financing</strong></p><ul><li>Knowledge at Wharton, <a href="https://knowledge.wharton.upenn.edu/article/huawei-technologies-a-chinese-trail-blazer-in-africa/">Huawei: a Chinese trailblazer in Africa (2009)</a></li><li>CSIS, <a href="https://csis-website-prod.s3.amazonaws.com/s3fs-public/legacy_files/files/publication/130215_competitiveness_Huawei_casestudy_Web.pdf">China’s competitiveness: Huawei case study</a></li><li>EEWorld, <a href="https://en.eeworld.com.cn/news/qrs/eic544385.html">Huawei’s ups and downs over 30 years</a></li><li>Washington Post / CAP, <a href="https://www.washingtonpost.com/politics/2020/11/19/huawei-has-hidden-advantage-its-push-global-dominance-now-us-is-trying-counter-it/">Huawei’s hidden advantage</a></li><li>IEEE ComSoc on WSJ, <a href="https://techblog.comsoc.org/2019/12/26/wsj-chinas-financial-support-aided-huaweis-rise-to-1-telecom-vendor-in-the-world/">China’s financial support and Huawei’s rise</a></li><li>Light Reading, <a href="https://www.lightreading.com/network-technology/huawei-s-lucky-number-30b">Huawei’s lucky number: 30B</a>; phys.org/AFP, <a href="https://phys.org/news/2019-05-huawei-key-beneficiary-china-subsidies.html">Huawei a key beneficiary of subsidies</a></li></ul><p><strong>Earlier in this series</strong></p><ul><li>Billion-Dollar Blueprints: Citadel, NVIDIA, the MOOVE Thesis, and Kobo360: The Anti-Moove</li><li>The Foundry Problem series; The Open-Source Map of African Energy Software and Companies; Decarbonize Your Factory. Don’t Spend Your Money; The Demand Was Never Small; The Future Is for Everyone With a Substation</li></ul><h3>Data Caveats</h3><p>Things I could not nail down, or that I want to flag before somebody else does.</p><ul><li><strong>Huawei’s share of African inverters and batteries.</strong> There is no audited public number. I am relying on Wood Mackenzie’s global ranking, which names Africa as a strong Huawei region, plus what I see across the tenders I work on. Treat it as an informed observation, not a statistic.</li><li><strong>27,500 sites in Nigeria</strong> comes from a secondary source (WeeTracker) and I could not trace it to a Huawei filing.</li><li><strong>About two million Huawei phones sold outside China in 2024</strong> is from a secondary source (chinamade.tech) summarising analyst data. The 48 million global figure is consistent with IDC and Counterpoint reporting; the domestic split is less firmly sourced.</li><li><strong>Engineers sleeping in customers’ equipment rooms</strong> is from personal experience, I spent 3 years powering Huawei Radios in Nigeria with Solar+Batteries.</li><li><strong>The DeepSeek 160,000 chip order</strong> is Bloomberg reporting as of 9 September 2026. Neither DeepSeek nor Huawei has confirmed it.</li><li><strong>Whether V4 was trained on Ascend.</strong> Huawei says part of V4-Flash was. DeepSeek has not said. Most analysts read it as inference on Huawei, training still on Nvidia. I have written it that way.</li><li><strong>Fortune Global 500 licensees.</strong> Huawei said 48 at its November 2025 forum and 55 on its IP page. I wrote “about fifty.”</li><li><strong>Huawei’s numbers are Huawei’s numbers.</strong> The annual report is KPMG-audited but the company is unlisted, and segment definitions changed in 2023, so the long series is not perfectly comparable.</li><li><strong>The six billion dollars</strong> is my arithmetic (400 million times sixteen years) and understates the real spend, because the budget grew and I could not find a disclosed cumulative figure.</li><li><strong>Yuan to dollar conversions</strong> are at roughly 7.15 unless the source gave its own figure.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8ee2174f3d60" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Get Your Energy Project Taken Seriously by an Investor]]></title>
            <link>https://kaykluz.medium.com/how-to-get-your-energy-project-taken-seriously-by-an-investor-6d4e1e6833eb?source=rss-2dfae1fe39fa------2</link>
            <guid isPermaLink="false">https://medium.com/p/6d4e1e6833eb</guid>
            <category><![CDATA[energy]]></category>
            <category><![CDATA[development]]></category>
            <category><![CDATA[project-finance]]></category>
            <category><![CDATA[infrastructure]]></category>
            <dc:creator><![CDATA[Kaykl.uz]]></dc:creator>
            <pubDate>Fri, 04 Sep 2026 12:36:01 GMT</pubDate>
            <atom:updated>2026-09-04T12:36:01.071Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xXdMuBwuomm1q0-g.jpeg" /><figcaption>The finished product gets the photograph. Most of this article is about everything that had to happen before it existed. <strong>Kamoa-Kakula solar plant, DRC.</strong> Photo: <strong>CrossBoundary Energy.</strong></figcaption></figure><p>I have sat through enough first meetings between project developers and investors to notice a slightly uncomfortable pattern.</p><p>The developer will often leave thinking the meeting went rather well. The investor asked questions, seemed engaged, requested the model and perhaps even said something encouraging about the opportunity. Then comes the familiar line: <em>“This is interesting. Keep us updated as the project develops.”</em></p><p>A few weeks pass. Nothing happens.</p><p>The developer follows up, sends another version of the model, perhaps adds a few slides to the deck and eventually concludes that the investor wasn’t serious, isn’t deploying capital, doesn’t understand the market or has suddenly decided that Africa is too risky.</p><p>Any of those things can be true. Investors waste developers’ time too.</p><p>But quite often the simpler explanation is that the project was not nearly as developed as the developer thought it was.</p><p>This distinction matters because the first serious conversation with an infrastructure investor is not really a pitch competition. You are not trying to persuade someone that renewable energy is a large market or that African businesses spend too much money on power. Most investors looking at the sector already believe some version of that.</p><p>They are trying to work out whether <strong>this particular opportunity is real enough to justify spending more time on it.</strong></p><h3>Understand what the investor is actually trying to establish</h3><p>An investor may ask you twenty questions about the customer, tariff, EPC, land, permits, technology, financing structure and financial model. It can feel as though you are being pulled in twenty different directions.</p><p>Usually you aren’t.</p><p>Most of those questions are attempts to understand a handful of things: whether the project can actually be built, whether someone will pay for what it produces, whether that counterparty can keep paying, whether the economics compensate for the risks and, increasingly importantly, whether the team sitting across the table can get all of this from PowerPoint to commercial operation.</p><p>That last question is easy to underestimate.</p><p>Infrastructure development is full of perfectly attractive projects that never become assets.</p><p>The technology works. The demand exists. The economics can look excellent. Yet somewhere between the first site visit and financial close, something refuses to cooperate. The customer will not accept the required contract term. The roof cannot carry the proposed system. The EPC number moves. The lender wants security the sponsor cannot provide. The tariff works in dollars but the customer earns in local currency. A permit everyone assumed would take six weeks takes six months.</p><p>An investor who has done enough projects knows this, so they are evaluating the project and the developer at the same time.</p><h3>Don’t lead with the IRR</h3><p>One of the easiest ways to make an early-stage project sound more developed than it is is to lead with the financial model.</p><p><em>“The project generates a 22% IRR.”</em></p><p>That is useful information, but it doesn’t tell me very much until I understand what went into the 22%.</p><p>Where did the load data come from? Has the customer confirmed it? Is the EPC price based on a detailed scope or a budget quotation? What happens if construction is delayed by six months? How is the tariff indexed? Who carries the currency risk? What assumptions have been made about degradation, replacement capex and availability? Is the land secured? Has anyone actually looked at the customer’s ability to honour a fifteen-year contract?</p><p>You can produce an extremely precise answer from extremely uncertain assumptions.</p><p>That isn’t a criticism of financial models. We need them. It is simply important to remember what a model does: it converts a set of assumptions into numbers. It doesn’t make the assumptions true.</p><p>The more experienced the investor, the more likely they are to spend the first meeting interrogating the inputs rather than admiring the output.</p><h3>Learn the difference between interest and commitment</h3><p>This is probably one of the most useful distinctions in project development.</p><p>A customer telling you they are interested in solar is not an offtake agreement. A landowner telling you a site is available is not site control. An EPC contractor sending a preliminary quotation is not a fixed construction price. A bank saying it would be interested in financing the project is not a term sheet.</p><p>And an investor asking you to send the project is definitely not financing.</p><p>Early-stage projects are full of statements that sound like facts because nobody has yet forced them to become commitments.</p><p>A large part of development is doing exactly that.</p><p>You take <em>“the customer wants this”</em> and turn it into a signed agreement. You take <em>“there should be enough land”</em> and establish control over a suitable site. You take a preliminary technical concept and turn it into something an EPC contractor is prepared to price and stand behind. You take an attractive financial model and gradually replace its assumptions with evidence.</p><p>This is why two apparently identical 10 MW projects can be worth completely different amounts.</p><p>One is 10 MW in Excel.</p><p>The other has been developed.</p><h3>Tell me what’s wrong with the project</h3><p>I become slightly suspicious when someone presents an infrastructure project with no obvious problem.</p><p>There is almost always a problem.</p><p>Maybe the customer has weak credit. Perhaps the tariff needs indexation that hasn’t been agreed yet. The roof requires reinforcement. The grid study isn’t finished. The feedstock assumptions are based on information nobody has independently verified. The EPC price is still preliminary. A permit is outstanding.</p><p>None of these automatically makes a project bad. In many cases they are simply development work.</p><p>What worries investors more is discovering a material issue halfway through diligence that the developer clearly knew about but decided not to mention.</p><p>There is a much better way to handle it:</p><p><em>“The biggest unresolved issue today is the customer’s credit profile. We have three years of financials, but we don’t yet have enough comfort for the proposed tenor. We’re looking at a parent guarantee and a shorter initial contract structure, and we expect to know whether either works within the next month.”</em></p><p>I would rather hear that than another five minutes about the size of the African renewable-energy market.</p><p>You have identified the problem, you understand why it matters and you have a plan for resolving it. That is what development looks like.</p><h3>Make your project easy to diligence</h3><p>This one sounds administrative until you have been on the other side of it.</p><p>Suppose the annual consumption in the financial model is different from the number in the proposal. The EPC quotation assumes a different capacity from the technical design. The tariff in the presentation is 8.2 cents but the model says 8.7. Someone asks where the customer’s load profile came from and nobody on the call is quite sure.</p><p>None of these things individually kills a project.</p><p>But each inconsistency creates another question, and every question creates work for somebody.</p><p>That matters because investors are not looking at your project in isolation. There may be another ten opportunities competing for the same investment team’s time. If one project has a clean data room, consistent numbers, clearly labelled documents and a developer who can explain where every important assumption came from, while another requires three emails to establish which financial model is current, the difference becomes surprisingly important.</p><p>A good data room will never turn a bad project into a good one.</p><p>A bad data room can certainly make a good project look worse than it is.</p><h3>Don’t ask whether they’re interested</h3><p>By the end of a good first meeting, I don’t think the most useful question is:</p><p><em>“So, are you interested?”</em></p><p>It is too easy to say yes. Nobody wants to spend thirty minutes discussing your project and then end the call with <em>“No, not particularly.”</em></p><p>I’d rather know what prevents the investor from doing something.</p><p>Ask what would need to be true for the project to progress internally. If the economics remain where they are, what are the two or three issues they would need resolved before taking it to investment committee?</p><p>Now the answer becomes useful.</p><p>They may tell you that the customer’s credit isn’t strong enough. Perhaps they need site control, a firmer EPC proposal or evidence of a particular permit. They may tell you the project is below their minimum ticket size.</p><p>Occasionally you discover something even more useful: this investor was never a realistic source of capital for the project in the first place.</p><p>That is not a failed meeting. You have just saved yourself three months of follow-ups.</p><h3>Treat fundraising as another source of project data</h3><p>After enough investor conversations, patterns begin to emerge.</p><p>One investor questioning your customer credit assessment may simply have a different risk appetite. If six credible investors independently arrive at the same concern, I would stop treating it as an investor problem.</p><p>The same applies when everyone struggles to understand the commercial structure, asks why the EPC price looks low or keeps challenging the same assumption in the model.</p><p>Write these questions down.</p><p>Not because investors are always right. They aren’t. But repeated confusion is information.</p><p>Sometimes the project needs work. Sometimes the explanation needs work. You need to know which.</p><p>This is one reason fundraising and project development shouldn’t be treated as completely separate activities. The market is constantly telling you where it thinks the weak points are.</p><p>Use that information.</p><h3>Project development is the product</h3><p>It took me a while to appreciate this properly.</p><p>We often talk as though a developer starts with an energy project and then goes looking for someone to finance it. But at the earliest stage, what you really have is an opportunity surrounded by uncertainty.</p><p>Development is the process of removing enough of that uncertainty for somebody to put serious capital behind it.</p><p>You validate the demand, secure the site, establish the technical solution, negotiate the commercial structure, investigate the customer, obtain the permits, firm up the construction cost, allocate the risks and work out whether the economics still make sense after reality has had its turn with the spreadsheet.</p><p>By the end, if you have done the job properly, you have produced something quite different from what you started with.</p><p>You have produced an investable project.</p><p>There is no great shortage of solar panels, inverters, batteries or EPC contractors. There isn’t necessarily a shortage of capital looking for good infrastructure opportunities either.</p><p>The scarce thing is the messy bit in between: projects that have been developed far enough, and well enough, for that capital to actually be deployed.</p><p>That is the thing you are selling to the investor.</p><p>And if the response after the meeting is <em>“keep us updated as it develops,”</em> sometimes the best thing to do is exactly that.</p><p>Go develop it.</p><p><strong>P.S.</strong> I have been unfair to Excel in this article. Excel has never claimed that the land was secured, the customer was creditworthy or the EPC price was firm. We put those assumptions in the cells ourselves. The spreadsheet was just following orders.</p><p><strong>P.P.S.</strong> None of this means investors are always right. I have seen investors pass on good projects, misunderstand markets and spend months diligencing opportunities they were never really capable of financing. That deserves its own article. Bankability is a two-sided problem.</p><p><strong>P.P.P.S.</strong> The inverse problem is knowing when to stop. Developers are very good at finding one more structure, investor, tariff or study that might save a project. Sometimes that is persistence. Sometimes the project is dead and nobody wants to say it. I think that deserves the next article.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6d4e1e6833eb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Open-Source Map of African Energy Software and Companies]]></title>
            <link>https://kaykluz.medium.com/the-open-source-map-of-african-energy-software-and-companies-a4dc2caed083?source=rss-2dfae1fe39fa------2</link>
            <guid isPermaLink="false">https://medium.com/p/a4dc2caed083</guid>
            <category><![CDATA[africa]]></category>
            <category><![CDATA[energy]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Kaykl.uz]]></dc:creator>
            <pubDate>Thu, 27 Aug 2026 10:56:01 GMT</pubDate>
            <atom:updated>2026-08-27T10:56:01.125Z</atom:updated>
            <content:encoded><![CDATA[<p>One year ago, I wrote an article.</p><p>The title was: <a href="https://kaykluz.com/p/why-is-no-one-building-the-software"><strong>Why Is No One Building the Software Stack for Energy in Africa?</strong></a></p><p>I had watched a brilliant Nigerian developer demonstrate an AI coding assistant.</p><p>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.</p><p>So I did what any reasonable energy professional would do.</p><p>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.</p><p>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.</p><p>I decided this was absurd. That part, I still believe.</p><p>The title was where I got into trouble. Because it turned out people <strong>were</strong> building. A lot of people.</p><p>I simply could not see them.</p><h3>The article escaped</h3><p>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.</p><p>Instead, founders sent it to founders. Investors forwarded it to portfolio companies. People I had never met began introducing themselves with some variation of:</p><blockquote>“Hello. We are actually building one of the things you said nobody was building.”</blockquote><p>Which was slightly embarrassing. It was also wonderful.</p><p>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.</p><p>That article genuinely changed my life. Not because it gave me the answer. Because it found the people.</p><h3>An annoying correction</h3><p>After enough of these conversations, I had to admit something. My original question was wrong.</p><p>Africa did not have a zero-builder problem. It had a <strong>visibility problem</strong>.</p><p>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.</p><p>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?</p><p>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.</p><p>A visible market compounds. An invisible market keeps introducing itself from the beginning.</p><h3>So I opened a spreadsheet</h3><p>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.</p><p>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.</p><p>Two problems forced it to grow up.</p><p>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.</p><p>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.</p><p>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.</p><p>So the spreadsheet gradually became a connected database. That database is now the first version of the <strong>Africa Energy Atlas</strong>.</p><h3>The first layer is live</h3><p>Today, I am launching Phase 1:</p><h3><a href="https://map.kaykluz.com/">The Africa Energy Software Map</a></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/824/0*GVQXGXA46sJ5bncZ.png" /></figure><p>The first release contains <strong>2,265 organisations</strong>, <strong>528 products</strong>, activity across <strong>54 African countries</strong>, and <strong>41 software categories</strong>: 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.</p><p>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.</p><p>Every record should eventually answer three different questions:</p><blockquote><strong><em>Who are they?</em></strong></blockquote><blockquote><strong><em>What have they built?</em></strong></blockquote><blockquote><strong><em>Where is there evidence that it is being used?</em></strong></blockquote><p>Those are different questions. The database treats them that way.</p><p>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.</p><p>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.</p><h3>Why it is open source</h3><p>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.</p><p>So the whole project is open:</p><h3><a href="https://github.com/kaykluz/africa-energy-atlas">Africa Energy Atlas on GitHub</a></h3><p>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.</p><p>It is also why the Atlas cannot become a commercial directory where companies pay for favourable placement. Visibility can be sponsored. Facts cannot.</p><p>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.</p><h3>The map will be wrong</h3><p>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.</p><p>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.</p><p>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.</p><h3>The invitation</h3><p>Go to <a href="https://map.kaykluz.com/">map.kaykluz.com</a>. Search for your company. Search for a problem you are trying to solve. Search your country, or the category you think nobody is building.</p><p>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.</p><p>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.</p><p>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.</p><p>The builders were here all along. They just could not always find one another.</p><p>Now, at least, we have a map. Let us make it accurate together.</p><p><strong>Explore the map:</strong> <a href="https://map.kaykluz.com/">map.kaykluz.com</a></p><p><strong>Inspect, contribute or fork the project:</strong> <a href="https://github.com/kaykluz/africa-energy-atlas">github.com/kaykluz/africa-energy-atlas</a></p><p><strong>Read the article that started this:</strong> <a href="https://kaykluz.com/p/why-is-no-one-building-the-software">Why Is No One Building the Software Stack for Energy in Africa?</a></p><p>— S<br><em>Chief Impostor, provisional cartographer, recovering spreadsheet architect</em></p><p><strong>P.S.</strong> 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><strong>P.P.S.</strong> 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><strong>P.P.P.S.</strong> The repository currently has fewer circular references than the spreadsheet it replaced. This is the highest technical standard I am prepared to guarantee.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a4dc2caed083" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Watermarking May Be the Reason World Models Replace LLMs]]></title>
            <link>https://kaykluz.medium.com/watermarking-may-be-the-reason-world-models-replace-llms-6a035515c826?source=rss-2dfae1fe39fa------2</link>
            <guid isPermaLink="false">https://medium.com/p/6a035515c826</guid>
            <category><![CDATA[world-models]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[v-jepa]]></category>
            <dc:creator><![CDATA[Kaykl.uz]]></dc:creator>
            <pubDate>Wed, 12 Aug 2026 15:12:54 GMT</pubDate>
            <atom:updated>2026-08-12T15:12:54.403Z</atom:updated>
            <content:encoded><![CDATA[<p><em>You Can’t Watermark a Thought — Why Yann LeCun’s World Models May Beat LLMs</em></p><p>On Tuesday morning I was doing what I always do before a diligence call, which is re-reading a memo I had already read three times so I could feel prepared instead of being prepared, when someone dropped a TechCrunch link into the group chat:</p><p><strong>“Anthropic says it will watermark text generated by its AI models.”</strong></p><p>My first reaction was not intellectual. My first reaction was: <em>wait, my memos?</em></p><p>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 <em>tighten</em> them. There’s a difference, and I will die on this hill, and apparently the hill is now <a href="https://techcrunch.com/2026/08/11/anthropic-says-it-will-watermark-text-generated-by-its-ai-models/">watermarked</a>.</p><p>Somewhere in Brussels, a machine can now detect that my “per my last email” had a co-author.</p><p>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.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*pYEGTwUG92f11Isv.png" /></figure><p>Here’s what I learned, roughly in the order I learned it.</p><p><strong>First: the watermark isn’t hidden <em>in</em> the text. It IS the text.</strong> 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: <em>ah. Loaded dice. Claude was here.</em></p><p>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 <a href="https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content">support page</a> 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.</p><p><strong>Second: this has been running quietly for two years and nobody noticed.</strong> Google’s DeepMind shipped SynthID-Text inside Gemini back in 2024. They published it in <a href="https://www.nature.com/articles/s41586-024-08025-4"><em>Nature</em></a> (actual <em>Nature</em>, 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.</p><p>I found that genuinely unsettling in a way I couldn’t name yet. Hold that thought.</p><p><strong>Third: the reason this is suddenly everywhere is not a moral awakening.</strong> It’s <a href="https://artificialintelligenceact.eu/transparency-rules-article-50/">Article 50 of the EU AI Act</a>, 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 <a href="https://digital-strategy.ec.europa.eu/en/news/strong-backing-code-practice-transparency-ai-generated-content">190 companies signed</a> the accompanying code of practice: Anthropic, OpenAI, Google, Meta, Microsoft, the whole class photo.</p><p>In deal terms: this isn’t ESG. This is covenant compliance.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*IyXoCera_ftCJd11.png" /></figure><p>Now. Here is where I started enjoying myself, because the <em>attack</em> literature is where academics stop being polite.</p><p>There is a paper (an actual peer-reviewed ICML paper) called <a href="https://arxiv.org/abs/2311.04378"><strong>“Watermarks in the Sand.”</strong></a> Boaz Barak’s group at Harvard, proving that “strong” watermarking is <em>mathematically impossible</em> 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.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/816/0*XInOHjaKGJw2itbW.png" /></figure><p>It gets better. Researchers at ETH Zurich showed you can <a href="https://watermark-stealing.org/"><strong>steal a watermark</strong></a>: query the model enough to reverse-engineer the dice-shaving, then either scrub it off your text <em>or stamp it onto text the model never wrote.</em> Cost of the attack: under fifty dollars. Success rate: over 80%.</p><p>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 <em>asymmetric risk profile</em>, and we would not be smiling when we said it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/661/0*V3qzkzjN3YLBbfIE.png" /></figure><p>And my favourite: watermarks are <a href="https://arxiv.org/abs/2402.14904"><strong>radioactive</strong></a>. 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.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*ynNkxE4ymf_SDzTC.png" /></figure><p>Meanwhile OpenAI (and I had to double-check this) has reportedly had a <a href="https://the-decoder.com/openai-has-a-highly-accurate-chatgpt-text-detector-but-wont-release-it-for-now/">~99.9% accurate detector</a> 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 <em>the thing punishes exactly the wrong people.</em></p><p>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.</p><p>I was about to close the laptop.</p><p>Then I had the small thought that turned into this essay.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*-JrJPkt3xSqcBX3I" /></figure><p>Every single thing I’d read — the dice-shaving, the sand, the radioactivity, the fifty-dollar heists — all of it attaches to one place: <strong>the output.</strong> The words. The pixels. The mouth of the machine.</p><p>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.</p><p>Yann LeCun. The JEPA guy. He left Meta in December, started <a href="https://techcrunch.com/2026/03/09/yann-lecuns-ami-labs-raises-1-03-billion-to-build-world-models/">AMI Labs in Paris, and raised a <strong>$1.03 billion</strong> seed round</a> (reportedly Europe’s largest ever) to build “world models” instead of language models.</p><p>If you haven’t met JEPA, the short version: a language model predicts the next <em>word</em>. A video generator predicts the next <em>pixel</em>. JEPA predicts the next <em>meaning</em>. It watches the world and forecasts what happens next in an abstract representation space, and <a href="https://ai.meta.com/blog/v-jepa-2-world-model-benchmarks/">Meta’s own description</a> of V-JEPA 2 is that it deliberately <em>discards</em> pixel-level detail as unpredictable noise. It doesn’t paint you a picture of the ball landing. It just… knows where the ball lands.</p><p>It is, architecturally, the quiet one in the meeting. Sees everything. Says nothing.</p><p>And here’s what stopped me: <strong>you cannot watermark that.</strong></p><p>Not “it’s hard.” Not “the tech isn’t ready.” There is <em>no attachment point.</em> 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.</p><p>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.</p><p>Even the law agrees, accidentally. Article 50 attaches to synthetic <em>audio, image, video, or text</em>. 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.</p><p>You can’t be quoted in the minutes if you never speak. Every deal lawyer knows this. Apparently LeCun does too.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Dw9XKIRedi33HgjE.jpeg" /></figure><p>And that’s when the actual shape of the thing showed up, five hours after the group chat message:</p><p><strong>This was never about detection. It’s a tax on speaking.</strong></p><p>Everything we call “generative AI” earns its keep by producing <em>content</em>; 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 <em>mouth</em>.</p><p>World models are a bet that the value of AI eventually migrates from the mouth to the <em>mind</em>: 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.</p><p>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.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/811/0*RRpNn4hOUjUGHDx5.png" /></figure><p>Now let me argue with myself, because the research kept me honest on three points.</p><p>One: silent models aren’t <em>invisible</em>. There’s a whole parallel literature (<a href="https://arxiv.org/abs/2201.11692">SSLGuard</a>, <a href="https://arxiv.org/abs/2410.14273">REEF</a>) on fingerprinting models by their <em>internal representations</em>, 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 (<em>who copied whom</em>, not <em>what did it say</em>), but it exists.</p><p>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, <a href="https://nvidianews.nvidia.com/news/nvidia-launches-cosmos-world-foundation-model-platform-to-accelerate-physical-ai-development">those outputs get watermarked</a>. The exemption isn’t for “world models.” It’s for <em>silence.</em></p><p>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.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*dFPfYwLW8b-RRK1s.jpeg" /></figure><p>Still. I keep coming back to the picture of it.</p><p>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.</p><p>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.</p><p>I know which one sounds like the senior partner.</p><p>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.</p><p><em>— Kay</em></p><p><em>Papers, for the fellow sickos: </em><a href="https://arxiv.org/abs/2311.04378"><em>Watermarks in the Sand (ICML </em></a><em>’24) · </em><a href="https://watermark-stealing.org/"><em>Watermark Stealing (ICML </em></a><em>’24) · </em><a href="https://www.nature.com/articles/s41586-024-08025-4"><em>SynthID-Text (Nature, </em></a><em>’24) · </em><a href="https://arxiv.org/abs/2402.14904"><em>Radioactive watermarks (NeurIPS </em></a><em>’24) · </em><a href="https://arxiv.org/abs/2502.12150"><em>LLM Idiosyncrasies (ICML ‘25)</em></a><em>: the one showing models are 97% identifiable by their word habits alone, no watermark needed, which possibly deserves its own essay · </em><a href="https://arxiv.org/abs/2506.09985"><em>V-JEPA 2</em></a><em> · </em><a href="https://arxiv.org/abs/2410.14273"><em>REEF (ICLR ‘25)</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6a035515c826" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Where Does Mini-Grid Revenue Actually Come From?]]></title>
            <link>https://kaykluz.medium.com/where-does-mini-grid-revenue-actually-come-from-ae59f1bb09b9?source=rss-2dfae1fe39fa------2</link>
            <guid isPermaLink="false">https://medium.com/p/ae59f1bb09b9</guid>
            <category><![CDATA[mini-grid]]></category>
            <category><![CDATA[infrastructure]]></category>
            <category><![CDATA[solar-energy]]></category>
            <category><![CDATA[energy]]></category>
            <dc:creator><![CDATA[Kaykl.uz]]></dc:creator>
            <pubDate>Sun, 09 Aug 2026 12:46:01 GMT</pubDate>
            <atom:updated>2026-08-09T12:46:01.253Z</atom:updated>
            <content:encoded><![CDATA[<p>Last Tuesday, around 9pm, I was reviewing two mini-grid proposals with a mug of coffee that had gone cold an hour earlier.</p><p>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.</p><p>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.</p><p>500 × 250 × 30. About ₦3.75 million a month.</p><p>I stared at it. That couldn’t be right. That’s the revenue of a decent Lagos restaurant. For an entire power utility.</p><p>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.</p><p>I checked the tariff. ₦250/kWh. Correct, and honestly fair; try running a diesel gen for less.</p><p>I checked the connections. Five hundred. Correct.</p><p>Which left the quiet little number in the middle. The kilowatt-hour. How much electricity does a newly connected rural household actually buy?</p><p>I did not know. Twelve years in this industry. I did not know.</p><p>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.</p><p>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 <strong>productive use</strong>.</p><p>The maize mill. The welder. The cold room. Demand triples, the project sings, everyone claps.</p><p>My first reaction, and I wrote this in my notebook, was: “so the whole sector is fiction?”</p><p>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 <em>can</em> survive. So what did those sites have that this spreadsheet only had as a curve?</p><p>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.</p><p>None of them got their demand from a growth assumption.</p><p>And that’s when the envelope finally made sense, at roughly 10:15pm.</p><p>The spreadsheet assumes demand <em>exists</em>, 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 <strong>manufactured</strong>. 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.</p><p>Once you see it that way, everything in the model reads differently.</p><p>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.</p><p>The “5% O&amp;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.</p><p>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.</p><p>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.</p><p>It wasn’t there. In either one.</p><p>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: <strong>who is manufacturing the demand, and did anyone budget for them?</strong></p><p>If the answer is in the spreadsheet, you might have a utility.</p><p>If the answer is a curve, you have a donation with an invoice printer.</p><p>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.</p><p>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.</p><h3>The Usual Asterisks</h3><p>*<em>My maths might be wrong. Please check it.</em></p><p>*<em>The numbers are rounded and anonymised; if they resemble your proposal, that is rather the point.</em></p><p>*<em>If you operate mini-grids and your demand curve came true, I am begging you to tell me how.</em></p><p>— S <em>(Chief Impostor, back-of-envelope enthusiast, cold coffee survivor)</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*9qEgPGEV79mrzng8.png" /></figure><p><em>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.</em></p><p><em>P.P.S. Found an error? Perfect. Drop a comment and I’ll fix it in v47_FINAL_FINAL_actually_final.</em></p><p><em>P.P.P.S. Yes, I reheated the coffee at 10:45pm. No, it did not help.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ae59f1bb09b9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Build a Billion-Dollar Energy Company in Africa: The MOOVE Thesis]]></title>
            <link>https://kaykluz.medium.com/how-to-build-a-billion-dollar-energy-company-in-africa-the-moove-thesis-681be93b7382?source=rss-2dfae1fe39fa------2</link>
            <guid isPermaLink="false">https://medium.com/p/681be93b7382</guid>
            <category><![CDATA[energy]]></category>
            <category><![CDATA[valuation]]></category>
            <category><![CDATA[africa]]></category>
            <dc:creator><![CDATA[Kaykl.uz]]></dc:creator>
            <pubDate>Fri, 07 Aug 2026 12:51:01 GMT</pubDate>
            <atom:updated>2026-08-07T12:51:01.330Z</atom:updated>
            <content:encoded><![CDATA[<p>I have a bad habit.</p><p>Whenever an African startup raises an indecent amount of money, I ignore the funding announcement.</p><p>Not completely. I read the number. I am not a monk.</p><p>Then I close the article and go looking for the smallest version of the company I can find.</p><p>The original pitch.</p><p>The first market.</p><p>The awkward early product description written before a brand agency taught the logo how to breathe.</p><p>Because the latest press release tells you what a company wants to become.</p><p>The first hundred customers tell you what it understood before everybody else did.</p><p>On 5 August 2026, Moove announced that it had raised $250 million at a $2.1 billion valuation.</p><p>I had initially heard $2.5 billion.</p><p>I was wrong by $400 million, which is an acceptable rounding error in venture capital and a dismissal offence in project finance.</p><p>The announcement came with the usual collection of numbers designed to make the rest of us reconsider what we have done with our lives.</p><p>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.</p><p>Seventy-six.</p><p>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.</p><p>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.</p><p>Or perhaps it did not.</p><p>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?</p><p>So I did what any normal person would do.</p><p>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.</p><p>I went looking for the machine.</p><p>Because companies do not accidentally become worth two billion dollars.</p><p>Somewhere, years earlier, somebody built a machine that kept producing customers, assets, data, revenue and trust. The latest funding round is not the machine.</p><p>It is the receipt.</p><p>And the more I looked at Moove, the more I realised that its machine has surprisingly little to do with cars.</p><p>It is a machine for turning visible demand into financeable infrastructure.</p><p>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.</p><p>Let me explain.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ogs_k4tZEtwog_fc.jpeg" /></figure><h3>I went looking for the 76-car company</h3><p>Ladi Delano and Jide Odunsi met while studying in London: Jide at the London School of Economics and Ladi at SOAS.</p><p>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.</p><p>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.</p><p>Not a charity. Not a panel discussion. <strong>A business.</strong></p><p>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.</p><p>But many of the people willing to drive could not afford a vehicle, and banks generally would not lend to them.</p><p>The banks saw informal income. No payslip. No acceptable credit history. No conventional collateral.</p><p>No thank you.</p><p>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.</p><p>This is how poverty reproduces itself with excellent paperwork.</p><p>Moove’s first important insight was that a vehicle in this context was not primarily a consumer purchase. It was a productive asset.</p><p>In Ladi’s formulation, a car in Africa was not merely a luxury. It could be the foundation of an income-generating business.</p><p>That distinction changes everything.</p><p>A bank looking at the driver saw a person with insufficient assets.</p><p>Moove looked at the same driver and saw an asset that could generate hundreds of small, observable cash flows every week.</p><p><strong>The question was no longer: What does this person already own?</strong></p><p><strong>It became: What could this person reliably produce if we financed the missing tool?</strong></p><p>That is a very different credit question.</p><p>And it is much closer to infrastructure finance than consumer lending.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/552/0*QaGTMDwMerIpAKEY.jpeg" /></figure><h3>Three parties standing around the same missing car</h3><p>There were effectively three participants in the original Moove system.</p><ol><li>The driver had the labour but not the vehicle.</li><li>Uber had the customer demand but not enough reliable vehicle supply.</li><li>The financial system had capital but could not confidently underwrite the driver.</li></ol><p>Each party possessed something the others needed. Nobody had put the pieces together.</p><p>Moove stepped into the middle. It financed the vehicle.</p><p>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.</p><p>Moove was not simply lending money and hoping the borrower remembered to transfer something at the end of the month.</p><p>It was sitting inside the cash-flow system.</p><p>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.</p><p>That meant it could reduce one of the ugliest risks in lending: the difference between a customer being <strong>able</strong> to pay and being <strong>willing</strong> to pay.</p><p>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.</p><p>Simple.</p><p>Except it is not simple at all.</p><p>It required Moove to combine financing, data integration, vehicle procurement, collections, insurance, maintenance and fleet operations inside one business.</p><p>The elegance was visible to the customer.</p><p>The complexity remained with Moove.</p><p>That is usually where the value is.</p><h3>The car was a small factory</h3><p>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.</p><p>The car was a tiny factory.</p><p>It converted labour and time into trips. Trips became revenue. Revenue repaid the car. The car eventually became the driver’s asset.</p><p>The same object was simultaneously:</p><ul><li>the productive equipment;</li><li>the source of revenue;</li><li>part of the collateral;</li><li>a monitored data point;</li><li>and, through the platform, connected to the collection system.</li></ul><p>That combination is unusually powerful.</p><p>Traditional lenders often try to predict creditworthiness by examining a person’s financial past.</p><p>Moove could examine the productive asset’s present.</p><ul><li>How many trips did it complete?</li><li>How much did it earn?</li><li>How consistently did it operate?</li><li>Was the driver’s activity increasing or declining?</li><li>Was maintenance behaviour creating future risk?</li></ul><p>The underwriting file was not merely a payslip, a bank statement and one uncle’s land document from 1987.</p><p>It was work happening in real time.</p><p>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.</p><p>This is the first major lesson for African energy.</p><p><strong>LESSON 1 — Do not finance the customer’s poverty. Finance the asset’s productivity.</strong></p><p>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.</p><p>It may still have customers, revenue, machinery, employees and a commercially useful reason to consume energy.</p><p>The absence of a pristine balance sheet does not mean the absence of productive demand.</p><p>Sometimes it means the financial system is asking the wrong questions.</p><p>Moove asked a better one.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*UuxqsE5pjQR9dVJ_.jpeg" /></figure><h3>Uber was doing three jobs</h3><p>People often describe Uber as Moove’s partner.</p><p>That is true, but incomplete.</p><p>Inside the original model, Uber performed at least three critical functions.</p><ul><li>It was the <strong>demand engine</strong>: the place where rides were requested.</li><li>It was the <strong>data rail</strong>: the system that recorded trips, earnings and driver behaviour.</li><li>And it was the <strong>payment rail</strong>: the place through which the money moved.</li></ul><p>This is why the relationship mattered so much.</p><p>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.</p><p>Moove had not merely found a distribution partner. It had attached itself to an operating marketplace in which demand, data and money already existed.</p><p>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 <strong>not</strong> work.</p><p>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.</p><p>The vehicles existed.</p><p>The cash-flow machine did not.</p><p>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.</p><p>That may be the most important confession in the entire story.</p><p>Moove’s model is not really vehicle-dependent.</p><p>It is <strong>demand-platform-dependent</strong>.</p><p>A car without trips is debt with seat belts.</p><p>A power plant without contracted demand is the same thing with a fence around it.</p><p>African energy has built plenty of both.</p><h3>The ugly work was the moat</h3><p>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.</p><ul><li>Moove had to buy cars.</li><li>Register them.</li><li>Insure them.</li><li>Track them.</li><li>Repair them.</li><li>Recover them when things went wrong.</li><li>Manage drivers.</li><li>Integrate with platforms.</li><li>Raise debt.</li><li>Manage currency exposure.</li><li>Report to lenders</li><li>Navigate transport regulators.</li><li>Build charging facilities.</li><li>And eventually operate fleets around the clock.</li></ul><p>This is less “fintech” and more “several difficult businesses in a trench coat”.</p><p>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.</p><p>There was no credible version of the business in which Moove built an app and outsourced all the unpleasant parts.</p><p>The unpleasant parts were the product.</p><ul><li>A driver did not merely need a loan approval. The driver needed a functioning car, insurance, maintenance support and enough uptime to continue earning.</li><li>Uber did not merely need another finance provider. It needed vehicles and drivers that remained available.</li><li>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.</li></ul><p>Moove had to become the connective tissue between all of them.</p><p>This is one reason asset-heavy African businesses become vertically integrated. Not always because the founders woke up wanting to own the universe.</p><p>Sometimes because the interfaces between specialist providers do not work.</p><p>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.</p><p>The customer ends up carrying the failure between all five.</p><p>Moove absorbed those interfaces.</p><p>That is expensive.</p><p>It is operationally brutal.</p><p>It is also hard to copy.</p><h3>Even the hiring mistakes were informative</h3><p>Moove did not scale perfectly.</p><p>No company does, although the retrospective case studies usually tidy this part away.</p><p>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.</p><p>The result was what Tingting Peng described as a stage mismatch.</p><p>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.</p><p>These are different sports.</p><p>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.</p><p>That organisational structure is more important than it looks.</p><p>A capital-intensive multinational cannot run every operational decision from headquarters.</p><p>But it also cannot allow each market to invent its own capital strategy, underwriting standards and risk appetite.</p><p>Local execution.</p><p>Central capital discipline.</p><p>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”.</p><h3>Capital was not funding. It was part of the product.</h3><p>Every asset-heavy startup begins with a mildly offensive paradox.</p><ul><li>It needs assets to generate revenue.</li><li>It needs revenue to attract debt.</li><li>It needs debt to buy the assets.</li></ul><p>Welcome to infrastructure.</p><p>Moove initially had to use equity to put vehicles on the road.</p><p>That was expensive capital doing the work of asset finance, but there was no established portfolio performance to show lenders yet.</p><p>The first vehicles produced the first repayments. The repayments produced data. The data made the next pool of vehicles easier to finance.</p><p>Gradually, the capital stack could evolve.</p><p>This is the sequence:</p><ul><li><strong>Equity pays for uncertainty.</strong></li><li><strong>Debt pays for repetition.</strong></li></ul><p>Moove learned this in real time.</p><p>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.</p><p>Moove shifted towards local lenders and local-currency facilities.</p><p>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.</p><p>Meanwhile, the company invested in lender-grade reporting and internal controls so that institutional capital could understand what it was financing.</p><p>This is not glamorous work.</p><p>Nobody has ever gone viral on LinkedIn for reconciling an asset register.</p><p>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.</p><p>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.</p><p>The capital stack was not sitting backstage supporting the business.</p><p>It <strong>was</strong> the business.</p><p>A vehicle-finance company with the wrong financing structure is just a very organised way to go bankrupt.</p><p>Moove was not only learning how to underwrite drivers.</p><p>It was learning how to make its own balance sheet underwritable.</p><p>That distinction is the difference between a small leasing company and a global infrastructure platform.</p><h3>Then the company started buying the operating system</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*eJEWhB3Crg5eRcI5.jpeg" /></figure><p>In January 2025, Moove acquired Kovi, a Brazilian mobility company.</p><p>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.</p><p>This is what a sensible acquisition looks like. Not merely more customers. More learning per customer.</p><p>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.</p><p>The acquisition helped Moove move further away from being a company that merely financed cars.</p><p>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.</p><p>That knowledge becomes valuable far beyond the original loan.</p><p>It improves underwriting. It reduces downtime. It informs procurement. It strengthens residual-value assumptions. It makes lenders more comfortable.</p><p>And every improvement makes the next asset slightly cheaper to finance.</p><p>The loop tightens.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/564/0*qk8Qd3gflojAlR2u.jpeg" /></figure><h3>The Waymo “pivot” that was not really a pivot</h3><p>Then came the part that initially confused me.</p><p>In December 2024, Moove announced a partnership with Waymo.</p><p>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.</p><p>At first glance, this looks like a radical departure.</p><p>The company began by helping African drivers own vehicles.</p><p>Now it is building infrastructure for vehicles that do not need drivers.</p><p>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.</p><p>But strategically, the move is much more continuous than it appears.</p><p>Look at the original system again.</p><ul><li>Uber had demand and software.</li><li>Drivers had labour.</li><li>The missing layer was financed, reliable vehicle supply.</li><li>Moove built that layer.</li></ul><p>Now look at autonomous mobility.</p><p>Waymo has the autonomous-driving system and the passenger platform.</p><p>The missing layer is still physical.</p><p>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.</p><p>The driver disappeared. The operating problem did not.</p><p>In some ways, it became more demanding.</p><p>A human driver can notice a flat tyre, find a mechanic, clean the back seat and complain loudly when the charger is not working.</p><p>An autonomous fleet still needs all those things. It simply cannot do any of them itself.</p><p>Moove followed the bottleneck.</p><p>First, the bottleneck was access to vehicles.</p><p>Then it was fleet uptime, maintenance and charging.</p><p>Now it is the physical operating infrastructure required to deploy autonomous vehicles at scale.</p><p>The company did not abandon its underlying model.</p><p>It extended it.</p><p>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.</p><p>This is the second major lesson.</p><p><strong>LESSON 2 — Do not remain loyal to the first product. Remain loyal to the system-level problem.</strong></p><p>Moove’s first product was vehicle finance.</p><p>Its enduring problem is enabling mobility platforms to deploy productive physical assets at scale.</p><p>The car was never the whole business.</p><p>The operating layer was.</p><h3>The five actual Mooves</h3><p>I tried to turn this into a perfect MOOVE acronym. The second O refused to cooperate. I respect its boundaries.</p><p>So instead, here are the five actual moves hiding beneath the company’s growth.</p><h3>Move one: find the demand before financing the asset</h3><p>Moove did not begin by buying ten thousand cars and then opening PowerPoint to investigate where ride-hailing might be popular.</p><p>It connected vehicle deployment to an existing demand platform.</p><p>This reduced utilisation risk. It provided operating data. It created a path to collection.</p><p>And it gave the company a partner with a commercial reason to help the fleet succeed.</p><p>The experiments that struggled were often the ones without a sufficiently strong demand engine.</p><p>Asset-first businesses ask: What can we build?</p><p>Demand-led infrastructure businesses ask: What productive activity is already constrained by a missing asset?</p><p>That is a much better place to start.</p><h3>Move two: finance productivity, not existing wealth</h3><p>The target customer did not have enough conventional collateral. But the financed asset could create measurable income.</p><p>Moove underwrote that income.</p><p>This does not remove risk.</p><p>Drivers can work less. Demand can fall. Vehicles can break. Regulations can change. Platforms can alter commissions.</p><p>But these are observable operating risks, not vague judgments about whether an informal worker “looks bankable”.</p><p>The data moves the conversation from prejudice to performance.</p><h3>Move three: control the cash-flow loop</h3><p>Demand alone is not enough.</p><p>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”.</p><p>Moove embedded repayment inside the platform’s payment flow. That substantially reduced collection friction.</p><p>The lesson is not that every company must literally deduct money every week.</p><p>The lesson is that the financing structure and the revenue structure cannot live in separate universes.</p><p>The closer repayment sits to the productive cash flow, the more financeable the asset becomes.</p><h3>Move four: turn operations into data and data into cheaper capital</h3><p>Every vehicle generated more than revenue.</p><p>It generated evidence. Utilisation. Maintenance behaviour. Collections. Default patterns. Residual values. City-level economics.</p><p>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.</p><p>This is the flywheel:</p><p>Demand → productive asset → embedded collection → operating data → cheaper capital → more assets.</p><p>The cars move.</p><p>The information compounds.</p><h3>Move five: follow the bottleneck until you become infrastructure</h3><p>Moove did not expand randomly into every financial product that could fit on a strategy slide.</p><p>It tried adjacent ideas, including other vehicle categories and financial services, and dropped some when the market structure was not strong enough.</p><p>But it kept moving deeper into the bottlenecks surrounding its core demand platforms.</p><ul><li>Vehicle access.</li><li>Fleet availability.</li><li>Maintenance.</li><li>Charging.</li><li>Depots.</li><li>City operations.</li><li>Autonomous fleet infrastructure.</li></ul><p>That is how a product company becomes an infrastructure company.</p><p>Not by announcing a platform.</p><p>By repeatedly taking responsibility for the next failure point.</p><h3>Then I had an unpleasant thought</h3><p>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.</p><p>That was mildly irritating.</p><p>In the first essay, the <strong>CITADEL Thesis</strong>, 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”.</p><p>In the second, the <strong>NVIDIA Thesis</strong>, 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.</p><p>Moove supplies the missing operating instruction.</p><p>Citadel tells you where the information advantage sits.</p><p>NVIDIA tells you how to make the ecosystem grow.</p><p>Moove tells you how the money reaches your account on Tuesday.</p><p>That is not a small contribution.</p><p>Because the average African distributed-energy company still scales one bespoke transaction at a time.</p><p>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.</p><p>Repeat the entire process from zero.</p><p>Each project may be good. The company does not necessarily compound. It accumulates.</p><p>There is a difference.</p><p>Moove did not accumulate unrelated vehicle loans.</p><p>It built a repeatable loop around a demand network.</p><p>That is what African energy needs.</p><h3>The closest thing energy has to Uber</h3><p>The obvious question is: Who is Uber in the energy version?</p><p>It is probably not one company across the entire continent. Energy demand is too fragmented. Regulation is too local. The assets vary too much.</p><p>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.</p><p>But within individual sectors, there are organisations that perform some of the same functions Uber performed for Moove.</p><ul><li>A telecom tower company controls hundreds or thousands of sites.</li><li>An industrial-park operator aggregates multiple businesses behind one location or grid connection.</li><li>A retail or franchise network has standardised sites and central procurement.</li><li>A logistics platform sees vehicle activity and can forecast charging demand.</li><li>An agricultural off-taker knows which processors and cold stores are active and commercially productive.</li><li>A payment company sees merchant revenues and settles funds into their accounts.</li><li>A bank, insurer or consumer-goods distributor may have hundreds of branches, warehouses or retail partners whose operations depend on reliable electricity.</li></ul><p>The perfect anchor is not simply an organisation with many sites. It is an organisation that already possesses three things:</p><ol><li><strong>Demand visibility.</strong></li><li><strong>Performance data.</strong></li><li><strong>Some influence over the payment rail.</strong></li></ol><p>That is energy’s Uber.</p><p>The more of those three functions the partner controls, the more powerful the model becomes.</p><h3>Imagine this</h3><p>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.</p><p>Many of those merchants run small generators because unreliable electricity threatens refrigeration, lighting, security, connectivity or basic trading hours.</p><p>A normal solar company approaches the merchants individually. Five thousand sales conversations. Five thousand site visits. Five thousand credit assessments. Five thousand collections problems.</p><p>And, eventually, five thousand reasons the sales team needs another quarter.</p><p>The Moove-style company starts elsewhere.</p><ul><li>It partners with the platform.</li><li>It uses transaction history and energy-metering data to identify a first cohort of commercially active businesses.</li><li>It installs standardised solar-and-storage packages.</li><li>The merchant pays a service fee that is designed to sit below its previous power cost.</li><li>The payment is collected through an agreed settlement mechanism.</li><li>The provider owns and maintains the equipment during the contract.</li><li>The platform benefits because its merchants remain open, connected and able to transact.</li><li>The merchant gets more reliable and potentially cheaper power without finding the upfront capital.</li><li>The energy company gets a visible demand pool, a data partner, a distribution channel and a collection mechanism.</li><li>The first 100 systems generate performance and repayment data.</li><li>That data supports a warehouse facility for the next 1,000.</li><li>The larger portfolio can eventually be refinanced or securitised.</li></ul><p>This is not merely rooftop solar. It is embedded productive-infrastructure finance.</p><p>The same architecture could be built around telecom towers, cold-chain networks, petrol stations, healthcare facilities, bank branches, logistics fleets or industrial clusters.</p><p>Different equipment.</p><p>Same machine.</p><h3>Your first 76 systems</h3><p>Moove’s first fleet was not 42,000 cars. It was 76.</p><p>That matters.</p><p>The first version of an African energy Moove should not begin with a press release announcing a 3 GW pan-African pipeline.</p><p>We already have enough 3 GW pipelines. Some of them are old enough to vote.</p><p>The first version should be narrow enough to learn from.</p><p>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.</p><p>Here is what I would build.</p><h3>1. Choose a productive-demand network</h3><p>Do not begin with “African SMEs”. That is not a customer segment. It is a census category.</p><p>Choose one network whose operations are sufficiently similar: pharmacies, cold stores, telecom sites, fuel stations, small manufacturers, delivery fleets, clinics or retail branches.</p><p>The network should have recurring economic activity and a clear cost associated with unreliable power.</p><h3>2. Meter before financing</h3><p>Install monitoring before committing serious capital.</p><p>Learn the load. Grid availability. Generator runtime. Fuel consumption. Operating hours. Seasonality. Outage impact. Payment behaviour.</p><p>The first product may be measurement. That is fine.</p><p>Moove could underwrite because it could see the work.</p><p>The energy company must learn to see the electrons.</p><h3>3. Standardise the assets</h3><p>Do not custom-design a miniature engineering masterpiece for every customer. Create a limited number of system architectures that cover most of the network.</p><p>Standard components. Standard contracts. Standard monitoring. Standard maintenance procedures. Standard spare parts.</p><p>The goal is not to eliminate engineering judgment.</p><p>It is to stop spending senior engineering time rediscovering that a 50 kW inverter remains a 50 kW inverter in Ibadan.</p><h3>4. Put payment beside performance</h3><p>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.</p><p>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.</p><p>The exact structure will vary.</p><p>The principle will not.</p><p><strong>Collections must be designed, not hoped for.</strong></p><h3>5. Use expensive capital only for the uncertainty</h3><p>The first systems may require equity, grant-backed first-loss capital or an anchor partner’s guarantee.</p><p>That is acceptable. Their job is to prove the model.</p><p>Once utilisation, performance, collections and loss rates become predictable, move the assets into debt structures suited to their cash flows.</p><p>Prefer local-currency financing where revenues are local currency.</p><p>Use revolving or warehouse facilities where repeated deployment requires capital to be recycled.</p><p>Do not raise permanent venture equity to fund every panel for the next fifteen years.</p><p>Equity should build the machine.</p><p>Debt should fill it with assets.</p><h3>6. Own the operational nervous system</h3><p>Whether the physical equipment sits in a separate asset company or special-purpose vehicle, the platform must retain the data and operating intelligence.</p><p>Every alarm. Every payment. Every maintenance visit. Every battery replacement. Every hour of downtime. Every customer complaint. Every EPC failure. Every forecast error.</p><p>That is the underwriting engine for the next portfolio.</p><p>Outsourcing installation is fine.</p><p>Outsourcing organisational learning is not.</p><h3>7. Follow the next real bottleneck</h3><p>Once the company reliably supplies electricity, the next bottleneck may be cooling. Charging. Steam. Power quality. Energy procurement. Equipment financing. Grid balancing.</p><p>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.</p><p>Moove did not become autonomous-mobility infrastructure by randomly adding products.</p><p>It followed the operational constraint.</p><p>The energy company should do the same.</p><h3>This is not “solar finance”</h3><p>There are already many companies offering solar loans, leases and power-purchase agreements.</p><p>That is not, by itself, the Moove thesis. The Moove thesis is the closed loop.</p><ol><li>The demand partner helps identify productive need.</li><li>The data improves underwriting.</li><li>The financed asset creates or protects cash flow.</li><li>The payment mechanism captures part of that cash flow.</li><li>The operator keeps the asset productive.</li><li>The resulting performance data attracts cheaper capital.</li><li>Then the same system is repeated across the network.</li></ol><p>Remove any one of those pieces and you may still have a perfectly good solar company.</p><p>But you probably do not have the machine.</p><p>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.</p><p>The scarce capability is coordinating demand, credit, capital, technology, collections and operations at portfolio scale.</p><p>That is what Moove built in mobility.</p><p>That is where the billion-dollar energy company will be built too.</p><h3>Where the analogy breaks</h3><p>Now for the annoying part.</p><p>Cars and energy systems are not the same.</p><p>I know.</p><p>Twelve years in energy has at least taught me that a Toyota Corolla and a 1 MW rooftop have different technical characteristics.</p><p>The Moove analogy is useful precisely until it becomes too convenient.</p><p>There are several important differences.</p><h3>1 — Energy assets are harder to recover</h3><p>A vehicle can be repossessed and redeployed.</p><p>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.</p><p>A solar asset is often physically and contractually tied to one site.</p><p>That makes customer selection, access rights, termination compensation and step-in provisions more important.</p><h3>2 — Avoided cost is not automatically cash flow</h3><p>A driver completes a trip and receives revenue.</p><p>A solar system may merely reduce what the customer would otherwise have spent.</p><p>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?</p><p>Savings can be economically real and still be difficult to collect.</p><p>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.</p><h3>3 — The contracts are longer</h3><p>A vehicle-finance arrangement may run for a few years.</p><p>An energy-as-a-service agreement can run for ten, fifteen or twenty.</p><p>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.</p><p>The underwriting cannot stop at today’s energy bill.</p><p>It must ask whether the customer, contract and asset remain defensible through several business cycles.</p><h3>4 — There may be no dominant demand platform</h3><p>Uber provides a concentrated marketplace. Most energy sectors are fragmented.</p><p>The energy company may have to assemble its own network through multiple anchor partners.</p><p>That makes the go-to-market slower.</p><p>It also makes the eventual platform more defensible, because the aggregation capability itself becomes valuable.</p><h3>5 — Control creates responsibility</h3><p>Moove’s embedded collection system reduces risk.</p><p>It also creates power.</p><p>And power can be abused.</p><p>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.</p><p>I do not know enough about those individual contracts to adjudicate the dispute.</p><p>But the lesson travels. A closed payment loop can be a moat. It can also become a chokehold.</p><p>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.</p><p>Otherwise “embedded finance” becomes a polite phrase for ensuring the customer experiences the downside first.</p><p>The company must be bankable.</p><p>The customer must also survive.</p><p>These are not competing objectives, no matter what the model says in Cell F97.</p><h3>The autonomous bet is still a bet</h3><p>It is also worth saying plainly that Moove’s $2.1 billion valuation is not proof that every part of this strategy will work.</p><p>A private valuation is a negotiated price in a funding round. It is not a public-market referendum delivered by God.</p><p>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.</p><p>The capital intensity could become a strength or a spectacularly expensive problem.</p><p>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.</p><p>That thesis is credible.</p><p>It is not inevitable.</p><p>The relevant lesson is not “copy everything Moove does”.</p><p>The lesson is to understand the sequence that earned Moove the right to make the bet.</p><p>It did not begin by constructing robotaxi depots and waiting for autonomy to arrive.</p><ol><li>It began with 76 productive assets attached to existing demand.</li><li>It built operating capability.</li><li>It accumulated data.</li><li>It raised increasingly sophisticated capital.</li><li>It expanded across markets.</li><li>It bought complementary technology.</li><li>It became useful to larger platforms.</li><li>Then it moved into the next infrastructure layer.</li></ol><p>The order matters.</p><p>Build the staircase before announcing the penthouse.</p><h3>What the billion-dollar energy company actually looks like</h3><p>It probably does not describe itself as a solar developer.</p><p>It may begin there.</p><p>But eventually it becomes the operating and financing layer beneath productive energy demand.</p><p>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.</p><p>Its flywheel looks like this:</p><p>Productive demand<br>→ financed energy asset<br>→ reliable operations<br>→ controlled payment<br>→ verified performance data<br>→ cheaper capital<br>→ more productive assets.</p><p>Eventually, the company becomes indispensable to the network.</p><p>The retailer does not call it to buy panels. The retailer expects its sites to have power.</p><p>The logistics platform does not negotiate separately for chargers, batteries, grid connections and fleet-energy software. The infrastructure layer handles it.</p><p>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.</p><p>Customers buy uptime.</p><p>Capital providers buy predictable cash flows.</p><p>The platform owns the knowledge between the two.</p><p>That is the company. Not the panel. Not the battery. Not even the electrons.</p><p>The machine around them.</p><h3>Start with the work</h3><p>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.</p><p>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.</p><p>That is the playbook.</p><ul><li>Find the work.</li><li>Finance the tool.</li><li>Control the cash flow.</li><li>Keep the tool productive.</li><li>Learn from every cycle.</li><li>Use the learning to reduce the cost of the next tool.</li><li>Then follow the bottleneck.</li></ul><p>The next billion-dollar African energy company may not begin with a billion-dollar pipeline.</p><p>I hope it does not.</p><p>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.</p><p>Then 760.</p><p>Then a local-currency warehouse facility.</p><p>Then 7,600.</p><p>Then the next infrastructure bottleneck.</p><p>The headlines will eventually call it an energy unicorn.</p><p>They will write about the valuation. The investors. The geographic footprint. The founders smiling in company hoodies.</p><p>And somewhere in the announcement there will be a photograph of a very large facility with a name invented by a branding consultant.</p><p>But the company will have been built years earlier. Inside the first small, boring, functioning loop.</p><p>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.</p><p>The car was never the company.</p><p>The loop was.</p><p>And the panel will not be the company either.</p><p><strong><em>-S</em></strong></p><p><strong>P.S.</strong> 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><strong>P.P.S.</strong> 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><strong>P.P.P.S.</strong> 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><strong>P.P.P.P.S.</strong> 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.</p><p><em>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.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=681be93b7382" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Decarbonize Your Factory. Don’t Spend Your Money]]></title>
            <link>https://kaykluz.medium.com/decarbonize-your-factory-dont-spend-your-money-24e4f6ef9810?source=rss-2dfae1fe39fa------2</link>
            <guid isPermaLink="false">https://medium.com/p/24e4f6ef9810</guid>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[energy]]></category>
            <category><![CDATA[africa]]></category>
            <category><![CDATA[renewable-energy]]></category>
            <dc:creator><![CDATA[Kaykl.uz]]></dc:creator>
            <pubDate>Wed, 05 Aug 2026 13:41:07 GMT</pubDate>
            <atom:updated>2026-08-05T13:41:07.801Z</atom:updated>
            <content:encoded><![CDATA[<p><em>The Impostor’s guide to decarbonising your operations end to end, today, without spending a dime upfront.</em></p><p>I have a strange job.</p><p>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.</p><p>Then I watch them not believe me.</p><p>Fair enough. If a stranger told me the same thing, I would also assume he was selling timeshares.</p><p>So consider this the receipts.</p><p>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.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*rmZfJNx2FRpn-Pai.png" /><figcaption>Concept illustration generated with OpenAI.</figcaption></figure><h3>The Numbers Work</h3><p>Most decarbonisation presentations still begin with a familiar assumption: the green option is environmentally responsible but financially painful.</p><p>That assumption is increasingly out of date.</p><p>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.<a href="https://www.irena.org/Publications/2026/Jul/Renewable-Power-Generation-Costs-in-2025">[8]</a></p><p>The other half of the story is that the incumbent became more expensive.</p><p>The green option did not suddenly become cheap. The incumbent became expensive, twice over: once at the fuel pump and once at the bank.</p><p><strong>At the pump</strong>: 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<a href="https://nerc.gov.ng/media/april-2024-supplementary-order-to-the-multi-year-tariff-ordermyto-2024/">[1]</a>, which is to say: factories. Diesel went from around ₦260 per litre in 2021 to ₦1,600 to ₦1,950 in 2026<a href="https://microdata.nigerianstat.gov.ng/index.php/catalog/158/related-materials">[2]</a>. Domestic gas is priced in dollars ($2.18/MMBtu base from April 2026)<a href="https://www.mondaq.com/nigeria/oil-gas-electricity/1769114/the-nigerian-midstream-and-downstream-petroleum-regulatory-authority-nmdpra-establishes-2026-domestic-base-price-and-wholesale-pricing-framework-for-natural-gas-for-the-strategic-sectors">[3]</a>, so every leg of naira depreciation, over 60% against the dollar between 2022 and 2025<a href="https://www.imf.org/-/media/files/publications/cr/2026/english/1ngaea2026001.pdf">[4]</a>, 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.</p><p><strong>At the bank:</strong> Nigeria’s monetary policy rate sat at a record 27.5% for most of 2025 before a token cut to 27.0% in September<a href="https://www.cbn.gov.ng/MonetaryPolicy/Activities.html">[5]</a>, with commercial banks lending to ordinary corporates at 30% and beyond<a href="https://nairametrics.com/2025/09/08/nigerian-businesses-struggle-to-service-loans-as-interest-rates-hit-36/">[6]</a>. Ghanaian bank lending has historically averaged near 30%<a href="https://www.bog.gov.gh/economic-data/interest-rates/">[7]</a>. 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.</p><p>Which is why the sentence that actually matters in this article is this one: <strong>you no longer have to pay the capex.</strong></p><p>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.</p><p>Let me walk the stack.</p><h3>Layer 1: Electricity. Solar, batteries, and now the grid itself</h3><p>Commercial and industrial solar in West Africa stopped being a pilot programme years ago.</p><p>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.<a href="https://www.daystar-power.com/en/">[23]</a></p><p>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<a href="https://crossboundaryenergy.com/crossboundary-energy-secures-40-million-equity-norfund-solar-portfolio-africa/">[9]</a>. By mid-2025 its awarded portfolio had reached roughly $707 million across 560MW of generation and 695MWh of storage in 20 countries<a href="https://www.miga.org/press-release/miga-support-over-100-energy-projects-20-african-countries">[10]</a>.</p><p>Daystar Power, bought by Shell in 2022<a href="https://www.shell.com/what-we-do/renewable-power/renewable-power-news-releases/shell-completes-acquisition-of-solar-solutions-provider-daystar-power-group.html">[11]</a>, sells the same promise in four words: no upfront capital expenditure<a href="https://www.africa-newsroom.com/press/daystar-power-announces-completion-of-acquisition-by-shell?lang=en">[12]</a>. Empower New Energy is rolling out a portfolio of commercial-scale solar-plus-storage plants for Justrite Superstores across Nigeria<a href="https://www.empowernewenergy.com/post/empower-new-energy-and-afrigreen-enter-into-a-financing-agreement-to-fund-solar-and-storage-systems">[13]</a> under the same structure, a programme I worked on personally in a previous role.</p><p>And the bench keeps getting deeper: ManoCap Energy replaces diesel generators with solar-and-battery hybrids for C&amp;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<a href="https://www.mirova.com/sites/default/files/2024-07/PR_Mirova-Gigaton-provides-USD-20%20million-debt-funding-West-Africa-renewable-energy-specialist-ManoCap-Energy.pdf">[14]</a>, while Germany’s GRIPS Energy runs solar-as-a-service into markets as unfashionable as Senegal’s farming belt<a href="https://grips-energy.com/scl-pv-project/">[15]</a>.</p><p>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.<a href="https://be-cs.co/finlays-project/">[25]</a><a href="https://www.the-star.co.ke/news/2023-10-16-james-finlay-commissions-24mwp-solar-power-projects">[26]</a><a href="https://www.standardmedia.co.ke/business/business/article/2001483576/tea-firm-commissions-24-megawatt-solar-plant">[27]</a> 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.<a href="https://be-cs.co/del-monte-kenya-partners-with-becs-to-commission-major-807kw-solar-power-plant/">[24]</a> Even the shoppers at Nairobi’s Junction Mall are browsing under our rooftop panels.<a href="https://www.linkedin.com/posts/lorraine-hutchison-ceng-mice-75046150_it-was-great-to-attend-this-site-visit-in-ugcPost-7296162086437453827-mkXC/">[28]</a><a href="https://www.linkedin.com/feed/update/urn:li:ugcPost:7296097010401660928/">[29]</a> I should know the layer well: I have done more than fifty C&amp;I clean energy projects across the region in the last five years, and this is where most of them started.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*D_TdXfJyqUly5t7o.jpeg" /><figcaption>James Finlay Kenya’s 2.0MW solar project in Kericho. Image credit: BECS official project release.</figcaption></figure><p>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<a href="https://about.bnef.com/insights/clean-transport/lithium-ion-battery-pack-prices-fall-to-108-per-kilowatt-hour-despite-rising-metal-prices-bloombergnef/">[16]</a>, 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<a href="https://www.woodmac.com/reports/energy-markets-belching-in-the-background-sizing-africas-distributed-diesel-power-landscape-and-displacement-opportunity-150023037/">[17]</a>. Every megawatt-hour a battery shifts is a megawatt-hour a genset does not burn at 30-plus US cents<a href="https://www.woodmac.com/reports/energy-markets-belching-in-the-background-sizing-africas-distributed-diesel-power-landscape-and-displacement-opportunity-150023037/">[17]</a>. 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<a href="https://www.kamoacopper.com/en/kamoa-copper-and-crossboundary-energy-sign-agreement-for-a-groundbreaking-baseload-renewable-energy-system-in-the-drc/">[18]</a>. If a battery can baseload a copper mine, it can carry your factory through the evening shift.</p><p>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<a href="https://www.miga.org/project/konexa-nbp2-and-solar-project-0">[19]</a>. 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<a href="https://www.etana.energy/article/etana-energy-sibanye-stillwater-ppa">[20]</a> at a stated 20 to 30% saving against utility rates<a href="https://thevault.exchange/?get_group_doc=245%2F1771568795-ssw-presentation-H2-YE-2025-20Feb2026.pdf">[21]</a>. 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.</p><p>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.</p><p>That last option matters because not every factory has enough roof or land.</p><p><strong>Upfront cost to the corporate in every case above: nothing.</strong></p><h3>Layer 2: Thermal heat. Steam, and now the whole powerhouse</h3><p>Electricity is maybe a third of a food factory’s energy problem. The rest is steam<a href="https://doi.org/10.1016/j.rser.2021.110856">[30]</a>, 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.</p><p>Steam is invisible, extremely hot and financially consequential.</p><p>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.<a href="https://doi.org/10.1016/j.rser.2021.110856">[30]</a></p><p>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.<a href="https://be-cs.co/sustainable-biomass-maximising-emission-reductions-in-industry/">[40]</a></p><p>Thermal is also the part of this stack where I am not merely studying other people’s case studies.</p><p>It is my day job.</p><p><strong>The proposition:</strong> 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<a href="https://www.iscc-system.org/">[31]</a>.</p><p>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.</p><p>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<a href="https://be-cs.co/ghana-gets-greener-energy-becs-signs-major-deal-with-kasapreko-plc/">[32]</a>, 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<a href="https://be-cs.co/ghana-gets-greener-energy-becs-signs-major-deal-with-kasapreko-plc/">[32]</a>: the first project of its kind in the region, delivering clean industrial steam at lower cost and lower emissions than the fossil alternative.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*DU2SULoter6fTwSW.jpeg" /><figcaption>Kasapreko’s biomass-to-steam plant at Tema, Ghana, under construction. Image credit: BECS official project release.</figcaption></figure><p>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.</p><p>What is already public: over 150 tonnes per hour of bioenergy projects in development across the continent<a href="https://be-cs.co/sustainable-biomass-maximising-emission-reductions-in-industry/">[40]</a>, 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.<a href="https://www.engineeringnews.co.za/article/john-thompson-becs-to-deliver-biomass-based-steam-for-africas-industries-2025-08-12">[41]</a> That is rather the point.</p><p>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<a href="https://crossboundaryenergy.com/project/kasapreko-phase-2/">[22]</a>. One Ghanaian beverage company, quietly stacking zero-capex decarbonisation layers while larger multinationals are still holding workshops about it.</p><p>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<a href="https://be-cs.co/energy-as-a-service/">[39]</a>, 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.<a href="https://be-cs.co/energy-as-a-service/">[39]</a> 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.<a href="https://be-cs.co/sustainable-biomass-maximising-emission-reductions-in-industry/">[40]</a> The boiler is the flagship. The fleet behind it covers the whole energy bill.</p><p>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<a href="https://be-cis.com/becis-1st-biomass-power-plant-project-developed-for-heineken-in-cambodia/">[42]</a><a href="https://www.khmertimeskh.com/501156393/heineken-opens-kingdoms-largest-biomass-plant/">[43]</a>; another Heineken plant in Indonesia has taken 100% of its steam from agricultural waste since 2018.<a href="https://be-cis.com/case-study/heineken-indonesia/">[44]</a> 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.<a href="https://be-cis.com/danone-builds-bioenergy-plant-to-mitigate-climate-change-and-endorse-circularity/">[45]</a></p><p>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.<a href="https://enertech.com.ph/elementor-8513/">[49]</a><a href="https://www.nestle.com.ph/media/news/cabuyao-biomass-boiler">[46]</a><a href="https://www.philstar.com/business/2026/04/28/2523996/nestl-unveils-new-biomass-facility-cabuyao-plant/amp/">[47]</a><a href="https://www.bworldonline.com/corporate/2026/04/28/745844/nestle-phl-launches-biomass-facility-to-support-net-zero-goal/">[48]</a> 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.<a href="https://boilerworldupdate.com/bioenergy-steam-solution-biomass-boilers-and-energy-as-a-service-in-southeast-asia/">[33]</a> Behind them, a solar book past 300MWp, and lenders like Clifford Capital, Pentagreen, BII and FMO writing eight-figure cheques against the portfolio.<a href="https://britchamvn.com/wp-content/uploads/2024/08/BECIS_Slide_EWG.pdf">[79]</a><a href="https://be-cis.com/green-loan/">[80]</a><a href="https://be-cis.com/bii-and-fmo-back-becis-with-joint-us50-million-financing-facility/">[81]</a><a href="https://be-cis.com/new-equity-financing/">[82]</a><a href="https://be-cis.com/us45-million-in-new-equity/">[83]</a> At some point ‘unproven’ stops being an available objection.</p><p>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<a href="https://www.sciencedirect.com/science/article/pii/S0973082609600168">[34]</a>.</p><p>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<a href="https://expogr.com/kenyaenergy/detail_news.php?newsid=4419&amp;pageid=2">[35]</a>. 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.</p><p>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<a href="https://www.revistamundoeletrico.com.br/geracao/biomassa/edp-entrega-caldeira-de-biomassa-a-vapor-para-a-fabricante-de-cacau-e-chocolate-barry-callebaut/">[36]</a>. Nestlé’s Kabirwala plant in Pakistan runs a 20 tonne-per-hour biomass boiler covering 85% of its thermal needs<a href="https://www.nestle.pk/sites/g/files/pydnoa361/files/2025-01/press-release-KBF.pdf">[37]</a>. Diageo’s East African breweries burn macadamia shells and bagasse<a href="https://www.diageo.com/en/news-and-media/stories/2023/eabl-met-with-king-charles-iii-on-his-visit-to-kenya">[38]</a>. The fossil side, meanwhile, is a dollar-priced gas bill in a depreciating currency. That is not a baseline worth defending.</p><h3>Layer 3: Electric Vehicles. Forklifts to 220-tonne mining trucks</h3><p>Africa’s electric-mobility story is too often reduced to motorcycles.</p><p>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.</p><p>You never own the asset. You never carry the battery risk. I can put any of these on a term sheet today.</p><p>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<a href="https://doi.org/10.21203/rs.3.rs-9213519/v1">[50]</a>; 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<a href="https://doi.org/10.21203/rs.3.rs-9213519/v1">[50]</a>.</p><p>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<a href="https://nigeria.endeavor.org/the-lifelong-build-the-next-chapter-for-christian-wessels-and-jasper-graf-von-hardenberg-2/">[51]</a>, 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<a href="https://gcpf.lu/gcpf-invests-in-african-electric-logistics-leader-bisedge/">[52]</a> for clients like Coca-Cola and Heineken<a href="https://www.bisedge.net/?trk=organization_guest_main-feed-card-text">[53]</a>, and a fresh $20 million from Metier Private Equity (May 2026) to scale<a href="https://www.prnewswire.com/news-releases/bisedge-secures-20-million-investment-from-metier-private-equity-to-expand-electric-intralogistics-footprint-302765546.html">[54]</a>. Each forklift converted saves around 55,000 litres of diesel and roughly 124 tonnes of CO2 a year<a href="https://www.energiseafrica.com/news/a-greener-more-prosperous-future-for-nigeria">[55]</a>. Nobody puts forklifts in a sustainability report. They should; the payback is boring and fast, and someone else will finance the whole fleet.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*rL0-f1rQG_XNWnzN.jpeg" /><figcaption>A Bisedge electric Linde E30 forklift and operator in Nigeria. Image credit: responsAbility/Bisedge official press release.</figcaption></figure><p>On the road, Kenya’s BasiGo sells electric buses on a pay-as-you-drive model<a href="https://www.basi-go.com/pay-as-you-drive-">[56]</a> (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<a href="https://www.basi-go.com/in-the-news/4ytd8iccdo21tbjwg812upo8f6ka2n">[57]</a>), with 114 buses in service by November 2025 and operators reporting up to 70% operating cost savings<a href="https://africasignal.com/basigo-driving-africas-electric-bus-revolution/amp/">[58]</a>, helped by Kenya Power’s e-mobility tariff dropping to KES 8/kWh off peak<a href="https://newsroom.kplc.co.ke/articles/kenya-power-kicks-off-customer-transition-to-e-mobility-tariff">[59]</a>. Ampersand and Spiro run tens of thousands of motorcycles on battery-swap subscriptions<a href="https://apnews.com/article/e91ff820907dd8329d15a9f2ed1e2b20">[60]</a>.</p><p>Globally, the IEA counted electric heavy-truck sales nearly tripling from about 84,000 in 2024 to 230,000 in 2025<a href="https://www.iea.org/reports/global-ev-outlook-2026/trends-in-other-ev-modes">[61]</a>, and truck-as-a-service operators like WattEV in California already price battery-electric haulage per mile at parity with diesel<a href="https://wattev.com/wattev-orders-50-volvo-vnr-electric-trucks/">[62]</a>, while Germany’s Hylane leases DHL its electric Mercedes trucks strictly pay-per-kilometre, breakdown risk included<a href="https://www.hylane.de/news/dhl-daimler-truck-und-hylane-schliessen-partnerschaft-bei-vollelektrischen-lkw--k8v8d8xrwda4jlvxqlvvr458">[63]</a>.</p><p>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<a href="https://im-mining.com/2026/02/18/xcmg-delivering-seven-xde260-mining-trucks-to-exxaro-in-south-africa/">[64]</a>; Zambia hosts a copper mine running what its contractor calls the first large-scale electric mining truck deployment on the continent<a href="https://www.africanews.com/2025/07/21/china-ships-31-electric-mining-trucks-to-africa-for-zambian-copper-mine//">[65]</a>; DHL, Unilever and Volvo launched Africa’s first electric Superlink truck in South Africa in August 2025<a href="https://www.dhl.com/za-en/home/press/press-archive/2025/dhl-supply-chain-launches-africas-first-electric-superlink-truck-with-unilever-and-volvo.html">[66]</a>; DSV followed with electric Volvos in December<a href="https://www.dsv.com/en-za/about-dsv/press/news/za/2025/11/volvo-8-ton-electric-truck-launch">[67]</a>. Volvo’s electric excavator cut hourly operating costs 74% in a 90-day pilot with Skanska<a href="https://www.volvogroup.com/en/news-and-media/news/2024/may/volvo-ce-shares-latest-electrification-and-innovation-news-at-act-expo.html">[68]</a>.</p><p>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.</p><h3>Layer 4: Cooling. Buy the Cold, not the Chiller</h3><p>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<a href="https://doi.org/10.1016/j.rser.2021.110856">[30]</a>.</p><p>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<a href="https://www.ellenmacarthurfoundation.org/es/ejemplos-circulares/refrigeracion-como-servicio-kaer">[69]</a>.</p><p>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<a href="https://energypartners.co.za/energy-partners-aspen-pharmacare-sign-r300m-caas-agreement/">[70]</a>. Its book includes a R360 million servitisation deal with dairy group Clover projecting R792 million of savings over the contract life<a href="https://businessreport.co.za/energy/2023-05-31-energy-partners-invests-r360m-in-clover-energy-servitisation-solution/">[71]</a>. The refrigerant transition mandated by the Kigali Amendment<a href="https://ozone.unep.org/kigali-amendment-overview">[72]</a> (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.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Eoqhgx6s7m5AFKXL.jpeg" /><figcaption>Aspen Pharmacare’s Gqeberha manufacturing site, where Energy Partners is building a centralised ammonia cooling plant. Image credit: Energy Partners official project release.</figcaption></figure><p>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<a href="https://be-cis.com/energy-as-a-service/cooling-as-a-service-solution/">[73]</a><a href="https://britchamvn.com/wp-content/uploads/2024/08/BECIS_Slide_EWG.pdf">[79]</a>, 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<a href="https://pidg.org/pidg-august-energy-30-million-platform-eaas/">[74]</a>, with modelled energy cost savings of around 34% on combined cooling and power.<a href="https://pidg.org/pidg-august-energy-30-million-platform-eaas/">[74]</a> And the grandfather of the whole idea, Veolia, has been selling outsourced industrial utilities, cooling included, to corporates for decades.<a href="https://www.veolia.com/en/solutions/integrated-utilities-management">[75]</a></p><p>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.<a href="https://be-cs.co/energy-as-a-service/">[39]</a><a href="https://be-cis.com/energy-as-a-service/cooling-as-a-service-solution/">[73]</a></p><p>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<a href="https://ca.kaeser.com/en/about-us/press/press-releases/l-sigma-air-utility.aspx">[77]</a>, because compressed air is famously the most expensive utility per unit of useful work in most plants<a href="https://www1.eere.energy.gov/manufacturing/tech_assistance/pdfs/compressed_air1.pdf">[76]</a>, and almost nobody owns the problem.</p><h3>The entire factory can become a service contract</h3><p>Put the layers together and the shape of the opportunity becomes clearer.</p><p>The factory’s structure can use lower-carbon materials.<a href="https://www.worldcement.com/special-reports/15072026/building-at-scale/">[78]</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*Zff5t41jz8R-23hI.jpeg" /><figcaption>CBI Ghana’s calcined-clay plant at Tema. Image credit: Fuller Technologies official project feature.</figcaption></figure><ol><li>Its daytime electricity can come from onsite solar.</li><li>Additional renewable electricity can be wheeled from a remote project.</li><li>Batteries can manage power quality, peaks and outages while reducing generator runtime.</li><li>Sustainable biomass can provide process steam.</li><li>Biomass CHP can produce both heat and electricity where the load supports it.</li><li>Cooling can be procured as a guaranteed temperature service.</li><li>Forklifts, trucks and heavy equipment can be electrified under long-term fleet-service contracts.</li><li>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.</li></ol><p>The company no longer has to appoint nine contractors and then discover that nobody is responsible for the total result.</p><p>It could instead have one decarbonisation programme, one technical baseline and one commercial architecture.</p><p>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:</p><p>We finance the infrastructure.<br>We deliver the output.<br>We guarantee the performance.<br>You pay for what your factory uses.</p><p>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.</p><p><strong>One accounting footnote,</strong> 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.<a href="https://www.ifrs.org/news-and-events/updates/ifric/2023/ifric-update-march-2023/">[84]</a></p><p>The claim here is about upfront cash, ownership burden and risk allocation (not an unsolicited accounting opinion, I am definitely not an accountant).</p><h3>Why someone else may finance your factory more cheaply than you can</h3><p>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.</p><p><strong>First</strong>, 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.</p><p><strong>Second</strong>, 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.</p><p><strong>Third</strong>, the provider absorbs and manages risks that the customer currently pays for badly: performance, maintenance, spare parts, fuel aggregation, battery degradation and residual value.</p><p><strong>Fourth</strong>, 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.</p><p>The $30 million BECIS bioenergy facility is one example.<a href="https://be-cis.com/green-loan/">[80]</a> MIGA’s guarantees for Konexa are another.<a href="https://www.miga.org/project/konexa-nbp2-and-solar-project-0">[19]</a> They demonstrate how contracted corporate demand can be turned into financeable infrastructure cash flows.</p><p>The multinational’s creditworthiness effectively becomes infrastructure.</p><p>But cheaper must be a test, not a slogan.</p><p>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.</p><p>The proposed service price must beat that number on an agreed set of assumptions.</p><p>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.</p><p>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.</p><p><strong>My commercial rule is simple: I will not ask the sustainability department to justify a project that the operating economics cannot support.</strong></p><p>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.</p><h3>The honest ledger</h3><p>Now the part a marketing department would usually remove.</p><p>I cannot point to a publicly documented West African factory that has placed every layer in this article under one master contract.</p><p>There are facilities that have combined several of them. Kasapreko has both financed solar and a signed biomass-steam project.<a href="https://crossboundaryenergy.com/project/kasapreko-phase-2/">[22]</a><a href="https://be-cs.co/ghana-gets-greener-energy-becs-signs-major-deal-with-kasapreko-plc/">[32]</a> Clover has outsourced power, steam and refrigeration.<a href="https://businessreport.co.za/energy/2023-05-31-energy-partners-invests-r360m-in-clover-energy-servitisation-solution/">[71]</a> Nigerian Breweries is using site-side systems and wheeled renewable power.<a href="https://www.miga.org/project/konexa-nbp2-and-solar-project-0">[19]</a> Bisedge is electrifying industrial material handling.<a href="https://gcpf.lu/gcpf-invests-in-african-electric-logistics-leader-bisedge/">[52]</a> CBI Ghana is lowering the carbon content of construction materials.<a href="https://www.worldcement.com/special-reports/15072026/building-at-scale/">[78]</a></p><p>The pieces exist.</p><p>The integrated category is still forming.</p><p>Nor does “end to end” mean a multinational has reached literal net zero across every Scope 1, 2 and 3 emission.</p><p>A food company may decarbonise its factory while emissions remain in fertiliser, agriculture, packaging and third-party logistics.</p><p>A data centre can address electricity, cooling, backup power and construction materials while server racks, semiconductors and networking equipment remain deeply carbon-intensive.</p><p>A mine can electrify light vehicles, loaders and some processing loads while its heaviest haul routes remain difficult.</p><p>A cement plant can reduce clinker, substitute fuels and improve efficiency, but conventional cement chemistry still releases process emissions.</p><p>Batteries are excellent at seconds, minutes and hours. They are not automatically the cheapest answer to several consecutive days without grid power or sunlight.</p><p>Biomass is useful only where the feedstock is genuinely sustainable, traceable, locally available and contractually secure.</p><p>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.</p><p>And not every “green” product is cheaper. Some lower-carbon materials still carry a premium, particularly where supply is immature.</p><p>So the defensible claim is not:</p><p><em>“Every African industry can become absolutely zero-carbon tomorrow at a lower cost.”</em></p><p>It is:</p><p><strong><em>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.</em></strong></p><p>That is still a very big claim.</p><p>It is also one I am prepared to put into a model.</p><h3>The actual pitch</h3><p>Here is the disclosure I have delayed less than most people would.</p><p>I sell this for a living.</p><p>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.</p><p>But my work in Africa today is wider than boilers.</p><p>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.</p><p>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.</p><p>I am not claiming that I have already closed every layer under one agreement.</p><p>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.</p><p>A technical baseline. A risk allocation. A tariff. A performance guarantee. And a counterparty prepared to write the cheque.</p><p>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.</p><p>I will bring the engineering options and the term sheets. Your signature on an offtake agreement is the entire ask.</p><p>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.</p><p>If any layer of this stack is sitting on your to-do list behind “get through this quarter,” find me on LinkedIn.</p><p>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.</p><p>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><p><strong><em>P.S.</em></strong> Full disclosure: 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.</p><p><strong><em>A confession:</em></strong><em> 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.</em></p><p><strong><em>P.P.S.</em></strong> If you think I am wrong about any of this, especially the “cheaper” claims, say so in the comments. Being corrected in public is the cheapest consulting I ever receive.</p><p><strong><em>P.P.P.S.</em></strong> If this was useful, 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.</p><p><strong><em>Next up:</em></strong><em> 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.</em></p><h3>References</h3><p>Repeated citation numbers point to the same source.</p><h3>Energy economics</h3><ol><li><a href="https://nerc.gov.ng/media/april-2024-supplementary-order-to-the-multi-year-tariff-ordermyto-2024/">NERC</a> — April 2024 Supplementary MYTO Order (Band A)</li><li><a href="https://microdata.nigerianstat.gov.ng/index.php/catalog/158/related-materials">Nigeria National Bureau of Statistics</a> — Automotive Gas Oil (Diesel) Price Watch</li><li><a href="https://www.mondaq.com/nigeria/oil-gas-electricity/1769114/the-nigerian-midstream-and-downstream-petroleum-regulatory-authority-nmdpra-establishes-2026-domestic-base-price-and-wholesale-pricing-framework-for-natural-gas-for-the-strategic-sectors">NMDPRA pricing analysis</a> — 2026 domestic base price for natural gas</li><li><a href="https://www.imf.org/-/media/files/publications/cr/2026/english/1ngaea2026001.pdf">IMF</a> — Nigeria: 2026 Article IV Consultation</li><li><a href="https://www.cbn.gov.ng/MonetaryPolicy/Activities.html">Central Bank of Nigeria</a> — 2025 Monetary Policy Committee decisions</li><li><a href="https://nairametrics.com/2025/09/08/nigerian-businesses-struggle-to-service-loans-as-interest-rates-hit-36/">Nairametrics</a> — Nigerian commercial lending rates</li><li><a href="https://www.bog.gov.gh/economic-data/interest-rates/">Bank of Ghana</a> — Interest-rate data</li><li><a href="https://www.irena.org/Publications/2026/Jul/Renewable-Power-Generation-Costs-in-2025">IRENA</a> — Renewable power generation costs in 2025</li></ol><h3>Electricity, solar and storage</h3><ol><li><a href="https://crossboundaryenergy.com/crossboundary-energy-secures-40-million-equity-norfund-solar-portfolio-africa/">CrossBoundary Energy</a> — $188m C&amp;I renewable-energy portfolio</li><li><a href="https://www.miga.org/press-release/miga-support-over-100-energy-projects-20-african-countries">MIGA</a> — $495m framework supporting CrossBoundary Energy projects</li><li><a href="https://www.shell.com/what-we-do/renewable-power/renewable-power-news-releases/shell-completes-acquisition-of-solar-solutions-provider-daystar-power-group.html">Shell</a> — Acquisition of Daystar Power</li><li><a href="https://www.africa-newsroom.com/press/daystar-power-announces-completion-of-acquisition-by-shell?lang=en">Daystar Power</a> — No-upfront-capex service model</li><li><a href="https://www.empowernewenergy.com/post/empower-new-energy-and-afrigreen-enter-into-a-financing-agreement-to-fund-solar-and-storage-systems">Empower New Energy and AFRIGREEN</a> — Justrite solar-plus-storage financing</li><li><a href="https://www.mirova.com/sites/default/files/2024-07/PR_Mirova-Gigaton-provides-USD-20%20million-debt-funding-West-Africa-renewable-energy-specialist-ManoCap-Energy.pdf">Mirova</a> — $20m ManoCap Energy debt financing</li><li><a href="https://grips-energy.com/scl-pv-project/">GRIPS Energy</a> — SCL solar project in Senegal</li><li><a href="https://about.bnef.com/insights/clean-transport/lithium-ion-battery-pack-prices-fall-to-108-per-kilowatt-hour-despite-rising-metal-prices-bloombergnef/">BloombergNEF</a> — 2025 Lithium-Ion Battery Price Survey</li><li><a href="https://www.woodmac.com/reports/energy-markets-belching-in-the-background-sizing-africas-distributed-diesel-power-landscape-and-displacement-opportunity-150023037/">Wood Mackenzie</a> — Africa’s distributed diesel-generation landscape</li><li><a href="https://www.kamoacopper.com/en/kamoa-copper-and-crossboundary-energy-sign-agreement-for-a-groundbreaking-baseload-renewable-energy-system-in-the-drc/">Kamoa Copper and CrossBoundary Energy</a> — Solar-plus-storage baseload project</li><li><a href="https://www.miga.org/project/konexa-nbp2-and-solar-project-0">MIGA</a> — Konexa renewable supply for Nigerian Breweries</li><li><a href="https://www.etana.energy/article/etana-energy-sibanye-stillwater-ppa">Etana Energy and Sibanye-Stillwater</a> — Wheeling PPA</li><li><a href="https://thevault.exchange/?get_group_doc=245%2F1771568795-ssw-presentation-H2-YE-2025-20Feb2026.pdf">Sibanye-Stillwater</a> — 2025 results and energy-cost savings</li><li><a href="https://crossboundaryenergy.com/project/kasapreko-phase-2/">CrossBoundary Energy</a> — Kasapreko solar project</li><li><a href="https://www.daystar-power.com/en/">Daystar Power</a> — Corporate service model</li><li><a href="https://be-cs.co/del-monte-kenya-partners-with-becs-to-commission-major-807kw-solar-power-plant/">BECS</a> — Del Monte Kenya 807kWp solar commissioning</li><li><a href="https://be-cs.co/finlays-project/">BECS</a> — James Finlay Kenya 2.0MW solar commissioning</li><li><a href="https://www.the-star.co.ke/news/2023-10-16-james-finlay-commissions-24mwp-solar-power-projects">The Star</a> — James Finlay solar-project coverage</li><li><a href="https://www.standardmedia.co.ke/business/business/article/2001483576/tea-firm-commissions-24-megawatt-solar-plant">Standard Media</a> — James Finlay solar-project coverage</li><li><a href="https://www.linkedin.com/posts/lorraine-hutchison-ceng-mice-75046150_it-was-great-to-attend-this-site-visit-in-ugcPost-7296162086437453827-mkXC/">LinkedIn</a> — Junction Mall Nairobi rooftop-solar site visit</li><li><a href="https://www.linkedin.com/feed/update/urn:li:ugcPost:7296097010401660928/">LinkedIn</a> — Junction Mall Nairobi rooftop-solar project post</li></ol><h3>Industrial heat and bioenergy</h3><ol><li><a href="https://doi.org/10.1016/j.rser.2021.110856">Review</a> — Energy use and decarbonisation in food and beverage manufacturing</li><li><a href="https://www.iscc-system.org/">ISCC</a> — Sustainability certification system</li><li><a href="https://be-cs.co/ghana-gets-greener-energy-becs-signs-major-deal-with-kasapreko-plc/">BECS</a> — Kasapreko biomass steam-purchase agreement</li><li><a href="https://boilerworldupdate.com/bioenergy-steam-solution-biomass-boilers-and-energy-as-a-service-in-southeast-asia/">Boiler World Update</a> — BECIS biomass Steam-as-a-Service</li><li><a href="https://www.sciencedirect.com/science/article/pii/S0973082609600168">Study</a> — Bagasse cogeneration in Mauritius</li><li><a href="https://expogr.com/kenyaenergy/detail_news.php?newsid=4419&amp;pageid=2">Tropical Power</a> — Gorge Farm Energy Park</li><li><a href="https://www.revistamundoeletrico.com.br/geracao/biomassa/edp-entrega-caldeira-de-biomassa-a-vapor-para-a-fabricante-de-cacau-e-chocolate-barry-callebaut/">EDP and Barry Callebaut</a> — Itabuna cocoa-husk biomass project</li><li><a href="https://www.nestle.pk/sites/g/files/pydnoa361/files/2025-01/press-release-KBF.pdf">Nestlé Pakistan</a> — Kabirwala biomass boiler</li><li><a href="https://www.diageo.com/en/news-and-media/stories/2023/eabl-met-with-king-charles-iii-on-his-visit-to-kenya">Diageo</a> — Biomass energy at East African Breweries</li><li><a href="https://be-cs.co/energy-as-a-service/">BECS</a> — Energy-as-a-Service solutions</li><li><a href="https://be-cs.co/sustainable-biomass-maximising-emission-reductions-in-industry/">BECS</a> — Sustainable biomass, 150 t/h pipeline and 85 to 95% GHG reduction</li><li><a href="https://www.engineeringnews.co.za/article/john-thompson-becs-to-deliver-biomass-based-steam-for-africas-industries-2025-08-12">Engineering News</a> — John Thompson and BECS biomass-steam partnership</li><li><a href="https://be-cis.com/becis-1st-biomass-power-plant-project-developed-for-heineken-in-cambodia/">BECIS</a> — Heineken Cambodia biomass plant</li><li><a href="https://www.khmertimeskh.com/501156393/heineken-opens-kingdoms-largest-biomass-plant/">Khmer Times</a> — Heineken opens Cambodia’s largest biomass plant</li><li><a href="https://be-cis.com/case-study/heineken-indonesia/">BECIS</a> — Heineken Indonesia bioenergy case study</li><li><a href="https://be-cis.com/danone-builds-bioenergy-plant-to-mitigate-climate-change-and-endorse-circularity/">BECIS</a> — Danone Prambanan bioenergy and circularity case study</li><li><a href="https://www.nestle.com.ph/media/news/cabuyao-biomass-boiler">Nestlé Philippines</a> — Cabuyao biomass-boiler inauguration</li><li><a href="https://www.philstar.com/business/2026/04/28/2523996/nestl-unveils-new-biomass-facility-cabuyao-plant/amp/">Philstar</a> — Nestlé Cabuyao biomass facility</li><li><a href="https://www.bworldonline.com/corporate/2026/04/28/745844/nestle-phl-launches-biomass-facility-to-support-net-zero-goal/">BusinessWorld</a> — Nestlé Philippines Cabuyao biomass facility</li><li><a href="https://enertech.com.ph/elementor-8513/">Enertech</a> — Nestlé Cabuyao 18 TPH biomass-boiler project</li></ol><h3>Electric mobility and machinery</h3><ol><li><a href="https://doi.org/10.21203/rs.3.rs-9213519/v1">Research Square</a> — Comparative field study of electric and diesel forklifts</li><li><a href="https://nigeria.endeavor.org/the-lifelong-build-the-next-chapter-for-christian-wessels-and-jasper-graf-von-hardenberg-2/">Endeavor Nigeria</a> — The founders behind Bisedge and Daystar Power</li><li><a href="https://gcpf.lu/gcpf-invests-in-african-electric-logistics-leader-bisedge/">GCPF and responsAbility</a> — Investment in Bisedge’s electric logistics fleet</li><li><a href="https://www.bisedge.net/?trk=organization_guest_main-feed-card-text">Bisedge</a> — Electric logistics clients and service model</li><li><a href="https://www.prnewswire.com/news-releases/bisedge-secures-20-million-investment-from-metier-private-equity-to-expand-electric-intralogistics-footprint-302765546.html">Metier Private Equity</a> — $20m investment in Bisedge</li><li><a href="https://www.energiseafrica.com/news/a-greener-more-prosperous-future-for-nigeria">Energise Africa</a> — Bisedge forklift fuel and emissions savings</li><li><a href="https://www.basi-go.com/pay-as-you-drive-">BasiGo</a> — Pay-As-You-Drive model</li><li><a href="https://www.basi-go.com/in-the-news/4ytd8iccdo21tbjwg812upo8f6ka2n">BasiGo</a> — Financing and uptime support</li><li><a href="https://africasignal.com/basigo-driving-africas-electric-bus-revolution/amp/">Africa Signal</a> — BasiGo fleet deployment and operating savings</li><li><a href="https://newsroom.kplc.co.ke/articles/kenya-power-kicks-off-customer-transition-to-e-mobility-tariff">Kenya Power</a> — E-mobility off-peak tariff</li><li><a href="https://apnews.com/article/e91ff820907dd8329d15a9f2ed1e2b20">Associated Press</a> — African electric motorcycles and battery swapping</li><li><a href="https://www.iea.org/reports/global-ev-outlook-2026/trends-in-other-ev-modes">IEA</a> — Global EV Outlook 2026: heavy-truck sales</li><li><a href="https://wattev.com/wattev-orders-50-volvo-vnr-electric-trucks/">WattEV</a> — Battery-electric trucking economics</li><li><a href="https://www.hylane.de/news/dhl-daimler-truck-und-hylane-schliessen-partnerschaft-bei-vollelektrischen-lkw--k8v8d8xrwda4jlvxqlvvr458">Hylane and DHL</a> — Pay-per-kilometre electric trucks</li><li><a href="https://im-mining.com/2026/02/18/xcmg-delivering-seven-xde260-mining-trucks-to-exxaro-in-south-africa/">International Mining</a> — Exxaro and XCMG mining trucks</li><li><a href="https://www.africanews.com/2025/07/21/china-ships-31-electric-mining-trucks-to-africa-for-zambian-copper-mine//">Africanews</a> — Electric mining trucks deployed in Zambia</li><li><a href="https://www.dhl.com/za-en/home/press/press-archive/2025/dhl-supply-chain-launches-africas-first-electric-superlink-truck-with-unilever-and-volvo.html">DHL, Unilever and Volvo</a> — Africa’s first electric Superlink</li><li><a href="https://www.dsv.com/en-za/about-dsv/press/news/za/2025/11/volvo-8-ton-electric-truck-launch">DSV</a> — Electric Volvo truck deployment in South Africa</li><li><a href="https://www.volvogroup.com/en/news-and-media/news/2024/may/volvo-ce-shares-latest-electrification-and-innovation-news-at-act-expo.html">Volvo CE</a> — EC230 Electric pilot results</li></ol><h3>Cooling and industrial utilities</h3><ol><li><a href="https://www.ellenmacarthurfoundation.org/es/ejemplos-circulares/refrigeracion-como-servicio-kaer">Ellen MacArthur Foundation</a> — Kaer Cooling-as-a-Service case study</li><li><a href="https://energypartners.co.za/energy-partners-aspen-pharmacare-sign-r300m-caas-agreement/">Energy Partners</a> — Aspen Pharmacare cooling-as-a-service project</li><li><a href="https://businessreport.co.za/energy/2023-05-31-energy-partners-invests-r360m-in-clover-energy-servitisation-solution/">Business Report</a> — Energy Partners and Clover servitisation agreement</li><li><a href="https://ozone.unep.org/kigali-amendment-overview">UNEP Ozone Secretariat</a> — Kigali Amendment overview</li><li><a href="https://be-cis.com/energy-as-a-service/cooling-as-a-service-solution/">BECIS</a> — Cooling-as-a-Service</li><li><a href="https://pidg.org/pidg-august-energy-30-million-platform-eaas/">PIDG and August Energy</a> — $30m Energy-as-a-Service platform</li><li><a href="https://www.veolia.com/en/solutions/integrated-utilities-management">Veolia</a> — Integrated industrial utilities management</li><li><a href="https://www1.eere.energy.gov/manufacturing/tech_assistance/pdfs/compressed_air1.pdf">U.S. Department of Energy</a> — Cost of compressed air in industrial plants</li><li><a href="https://ca.kaeser.com/en/about-us/press/press-releases/l-sigma-air-utility.aspx">Kaeser</a> — Compressed air supplied as a metered utility</li></ol><h3>Cement and buildings</h3><ol><li><a href="https://www.worldcement.com/special-reports/15072026/building-at-scale/">World Cement</a> — CBI Ghana’s calcined-clay plant</li></ol><h3>Finance, standards and integrity checks</h3><ol><li><a href="https://britchamvn.com/wp-content/uploads/2024/08/BECIS_Slide_EWG.pdf">BECIS</a> — 300MWp-plus solar portfolio presentation</li><li><a href="https://be-cis.com/green-loan/">BECIS, Clifford Capital and Pentagreen</a> — $30m green-loan facility</li><li><a href="https://be-cis.com/bii-and-fmo-back-becis-with-joint-us50-million-financing-facility/">BECIS, BII and FMO</a> — $50m financing facility</li><li><a href="https://be-cis.com/new-equity-financing/">BECIS</a> — New equity financing</li><li><a href="https://be-cis.com/us45-million-in-new-equity/">BECIS</a> — $45m in new equity</li><li><a href="https://www.ifrs.org/news-and-events/updates/ifric/2023/ifric-update-march-2023/">IFRS Foundation</a> — IFRS 16 lease definition and control rights</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=24e4f6ef9810" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>