Against the backdrop of ITW Africa in Nairobi, the conversation about Africa’s digital future has shifted beyond connectivity to data centres, compute, sovereign AI and energy — but one crucial element remains missing: a local cloud platform layer that sits between newly arriving infrastructure and the institutions expected to use it.
John Njoroge examined this shift, arguing that capacity is not capability and that institutions must be architected to absorb the infrastructure being built. Building on that, Hans van Linschoten says the institutional work Njoroge describes cannot begin until a common cloud platform layer exists to provide identity, tenancy, storage, networking, monitoring, orchestration and billing as shared services.
What the cloud platform actually delivered
Van Linschoten notes that the global hyperscalers did not invent compute but rather delivered a self-service platform layer on top of existing racks and halls: instant environments, pay-as-you-go billing and scale without special permission. That self-service layer is taken for granted elsewhere but is largely absent across Africa. Fibre, subsea capacity and data centre halls are arriving, and many local applications are being built, yet there is little between raw infrastructure and applications.
Fragmentation as a symptom
When no common platform layer exists, each programme must build its own middleware — identity, tenancy, storage, networking, monitoring, orchestration, billing — often from different vendors, assumptions and funding cycles. Van Linschoten argues this drives the incoherence John Njoroge observes: ministries, agencies and hospitals end up with multiple unmatched systems because every project had to invent its own foundation.
Design constraints for an African cloud platform
Van Linschoten says the platform cannot be a copy of the hyperscaler model because their economics assume cheap power, cheap capital and dense demand that most African markets lack. Two non-negotiable constraints follow:
– Low capex: entry cost must be a fraction of a data centre build. In van Linschoten’s modelling, a production node capable of running cloud services and AI inference can be built for roughly $300–500k depending on minimum initial hardware. He adds that a quarter-rack inside an existing enterprise IT room is frequently sufficient; anything beyond a rack or two can be provided by containerised modules.
– Scalable by default: starting small requires multi-tenant, self-service, metered and billable platforms that add capacity without redesign or a second procurement, and that are AI-ready from the outset.
Van Linschoten points to ABQ Cloud Uganda as an example: it runs on this pattern, deployed at the Raxio facility in Namanve and at the Deep Tech Hub in Kampala.
Sovereignty and the build-operate-transfer model
A platform layer permanently operated by a foreign vendor cannot deliver true sovereignty, van Linschoten warns, because the platform sees tenants, workloads, data movements and permissions. He recommends a build-operate-transfer delivery model: build with experienced partners, operate jointly until local teams genuinely run the platform, then transfer assets and skills on a defined timetable. It is slower and less profitable for builders, but he argues it is the only model that survives changes of government, currency or vendor strategy.
Government as the primary buyer
Governments are the largest customers in most African markets, and fragmentation is especially costly there because different agencies hold fragments of the same citizen data with no trusted means of reconciliation. Van Linschoten says a national data exchange — providing provenance, access control, permissioning and an auditable record of system knowledge and agent permissions — can answer the institutional questions John Njoroge raised. He cites work by teams such as AI Studio Uganda as examples of this approach being built rather than just described.
No single owner if the ecosystem is to work
The platform layer can only function if it is not exclusively owned. Van Linschoten argues that a single vendor, telco or hyperscaler holding the governing platform would recreate the dependency the continent aims to escape. The practical requirements are stack neutrality, standards-based interconnect, published interfaces and customers’ free choice of models and workloads. He notes that his organisation opened its federation and billing layer to infrastructure stacks it did not build, accepting the cost of lock-in for others as the price of a real ecosystem.
Van Linschoten agrees with John Njoroge that the most important work begins after infrastructure is built, and adds that work cannot begin until the missing cloud platform layer is in place.
Hans van Linschoten is CEO of whitesky.cloud and Co-founder of AfriQloud.
