Against the backdrop of ITW Africa in Nairobi last week, the conversation around Africa’s digital future has moved further up the stack. John Njoroge examined that shift in a newly published piece, looking beyond connectivity to data centres, compute capacity, sovereign AI and energy.
Connectivity is no longer the frontier. Data centres, compute capacity, sovereign AI and energy are. His harder question is what happens after the infrastructure is built, and his answer is that capacity is not capability: institutions have to be architected to use what is being poured into the ground.
He is right, and the argument deserves to travel further than LinkedIn. I want to add the part that sits between his two halves — because there is a layer missing, and until it exists the institutional capability he describes cannot be built at all.
What the hyperscalers actually sold
The digital transformation of the past fifteen years is usually credited to cloud computing. That is imprecise. What AWS, Azure and Google Cloud delivered was not compute. Compute already existed, in racks, in colocation halls, in university basements. What they delivered was a self-service platform on top of it: swipe a card, get an environment in minutes, deploy in an afternoon, pay for what you consume, scale without asking anyone’s permission.
Every SaaS company, every digital government service, every AI application built since has been built on the assumption that this layer is simply there. It is so reliably present elsewhere that it has become invisible, which is why it rarely features in discussions of what Africa needs.
It is not there on this continent in any comparable form. Raw infrastructure is arriving – fibre, subsea capacity, data centre halls, the kind of investment John cites. Applications are being built, many of them excellent and genuinely local. Between the two, there is very little.
Fragmentation is a symptom, not the disease
John describes institutions becoming more digital without becoming more connected: a ministry deploys one platform, an agency builds another, a hospital introduces a third, and over time the organisation holds more systems and less coherence. His diagnosis is exact. I would push the causation one step further back.
This is not primarily a sequencing failure or a question of poor coordination. It is what necessarily happens when there is no common platform layer. Each programme has to construct its own middle – identity, tenancy, storage, networking, monitoring, orchestration, billing – because nothing underneath supplies it. Those middles are built by different vendors, on different assumptions, at different times, under different funding cycles. They do not reconcile, and no amount of integration budget applied afterwards will make them reconcile cleanly.
You cannot reasonably ask an institution to be coherent when every project it has funded was obliged to invent its own foundation first.
What the layer has to look like here
It cannot be a copy of the hyperscaler model. Their economics assume cheap and reliable power, cheap capital and demand density sufficient to justify hundred-megawatt sites. Most African markets have none of the three. A quarter-billion-dollar GPU build is not the answer for a country whose entire public-sector workload would comfortably fit in a few racks, and framing the gap that way puts the starting line somewhere almost no institution can reach.
Two design constraints follow, and neither is negotiable.
Low capex. The entry cost has to be a fraction of a data centre build. In our own modelling, a production node capable of running cloud services and AI inference can be built for roughly anywhere from $300-500k depending on minimum requirements for the initial hardware.
A quarter rack inside an existing enterprise IT room is frequently sufficient; anything beyond a rack or two is covered by a containerised module rather than a new facility. This matters because it changes who is able to start: a national research institute, a mid-sized bank, a regional health authority, a university.
Scalable by default. Precisely because you start small, the platform has to add capacity without redesign, re-platforming or a second procurement. Multi-tenant from the first node. Self-service. Metered and billable, so cost can be attributed to the department that incurred it. AI-ready from the outset, so an institution’s first inferencing workload does not require the whole conversation to start again.
This is not theoretical. ABQ Cloud Uganda runs on exactly this pattern, deployed at the Raxio facility in Namanve and at the Deep Tech Hub in Kampala.
Who owns it
A platform layer permanently operated by a foreign vendor is not sovereign, whatever the data residency terms say. That layer sees everything: which tenants exist, which workloads run, which data moves, which permissions were granted, what changed and when. Sovereignty asserted at the infrastructure layer while the platform layer above it is rented is the weaker half of the claim.
The delivery model that resolves this is build-operate-transfer. Built with partners who have done it before, operated jointly until the local team genuinely runs it rather than nominally owns it, then transferred on a defined timetable with the skills transferred alongside the assets. It is slower than the alternative and less profitable for the party doing the building. It is also the only version that survives a change of government, a change of currency or a change of vendor strategy.
Who buys it
Government is the largest customer in most African markets by a considerable margin. That is inconvenient in procurement terms and decisive in strategic ones, because government is also where fragmentation is most expensive: the ministry, the hospital, the school network and the revenue authority each hold a fragment of the same citizen’s reality with no trusted means of reconciling them.
This is where John’s institutional question becomes a concrete instrument rather than a principle. A national data exchange – provenance, access control, permissioning, and an auditable record of what a system knows and what an autonomous agent is allowed to do – answers the ‘know your agent’ list he sets out, item by item. It is one of the more credible answers available, and work by teams such as AI Studio Uganda shows what it looks like when it is built rather than described.
No one can hold an exclusive position
The last point is the hardest, and it is a condition rather than a recommendation. The layer only works if it is not owned. If a single vendor, a single telco or a single hyperscaler ends up holding the platform layer that governs what every institution knows, the continent will have reconstructed the dependency it set out to escape and painted a local flag on it.
In practice that means stack neutrality as an architectural commitment rather than a marketing line: federation that accepts infrastructure stacks other than your own underneath it, standards-based interconnect, published interfaces, and the customer’s free choice of models and workloads on top. We have made that decision ourselves, opening our federation and billing layer to infrastructure stacks we did not build. It costs us lock-in. That is the price of the ecosystem being real rather than rhetorical.
Africa’s genuine advantage in this cycle is that it is building this layer late enough to build it open. The rest of the world acquired its platform layer from three American companies and is now spending considerable political energy trying to loosen that grip. There is no
obligation to repeat the sequence – but avoiding it requires participants willing to rise above their individual ambitions, and that is a harder ask than any of the engineering.
John is right that the most important work begins after the infrastructure is built. I would only add that it cannot begin until the missing layer is in place.
Hans van Linschoten is the CEO of whitesky.cloud, a European sovereign cloud platform, and Co-founder of AfriQloud, a pan-African sovereign cloud and edge AI infrastructure initiative.
Real ESG impact doesn’t happen in panels alone, it happens in the rooms where financiers, operators, and policymakers actually align. Our GreenShift Forum 2026 cuts the noise, bringing together the people rewiring Africa’s sustainability and energy frameworks for one focused day in Nairobi. Secure your seat.
Go to TECHTRENDSKE.co.ke for more tech and business news from the African continent and across the world.
Follow us on WhatsApp, Telegram, Twitter, and Facebook, or subscribe to our weekly newsletter to ensure you don’t miss out on any future updates. Send tips to editorial@techtrendsmedia.co.ke
