Sometime in 2026, Nigeria is expected to commission its first facility purpose-built for AI workloads — a data centre with the power density and cooling capacity required to run the compute-intensive models that are reshaping software. The fact that this is a milestone says something important about the state of digital infrastructure in Africa's largest economy. It also raises a set of questions that matter to every operator building anything that relies on cloud computing, AI, or data processing in the region.
The infrastructure gap is not abstract
Most African operators interact with the data centre problem indirectly. They experience it as API latency when using cloud services hosted in Europe or the United States. They experience it as higher costs for cloud compute, because the nearest hyperscaler region is thousands of kilometres away and that distance shows up in pricing. They experience it as regulatory friction when data residency requirements — increasingly common across African jurisdictions — conflict with cloud architectures that store data offshore by default.
The aggregate effect is a structural disadvantage. African companies building software products operate on infrastructure that is geographically distant from their users, priced at rates designed for markets with different cost structures, and subject to regulatory complexity that their counterparts in North America or Europe do not face. This is not insurmountable — millions of successful African software businesses operate on AWS or Google Cloud infrastructure hosted in Ireland or Virginia — but it is a friction that compounds.
Why AI makes the infrastructure gap more acute
For most categories of software, the geographic distance of compute infrastructure is an inconvenience, not a crisis. A few hundred milliseconds of additional latency does not break a CRM or an inventory management system. But AI inference — particularly for real-time applications like conversational interfaces, fraud detection, or image processing — is latency-sensitive in ways that other software is not.
More significantly, the economics of AI compute are different from the economics of general cloud computing. AI workloads are GPU-intensive, power-hungry, and generate significant heat. The facilities required to run them at scale need reliable, affordable power — which is the specific constraint where Africa's infrastructure gap is most acute. Running a GPU cluster in an environment with expensive, unreliable power is either prohibitively costly or operationally fragile. Both outcomes undermine the viability of AI-native products built in Africa for African markets.
This is why the arrival of AI-capable data centre infrastructure in Nigeria — and the similar investments underway in South Africa, Kenya, and Egypt — matters beyond the technical announcement. It is the precondition for a generation of AI-native African products that run on local infrastructure, with local data, at latencies that make real-time AI applications viable.
What operators should know
The shift is not immediate. Building data centre infrastructure takes years. Power constraints do not resolve quickly. But the direction is clear, and the pace is accelerating. For operators making architectural decisions about their products today, a few things are worth factoring in.
Data residency requirements are going to tighten across more African jurisdictions, following the pattern set by Nigeria's NDPR and Kenya's Data Protection Act. Products that store all data offshore by default are building a compliance liability that will eventually need to be resolved. Designing for data residency from the start is cheaper than retrofitting it later.
Local cloud providers — Rack Centre in Nigeria, Liquid Intelligent Technologies across multiple markets, Yamifi and others — are worth evaluating seriously alongside the global hyperscalers. They may not match AWS or Google Cloud on breadth of services, but for specific use cases where data residency, latency, or cost are primary constraints, they are increasingly viable.
And for products being built with AI at the core: the infrastructure landscape in 2028 will look meaningfully different from the infrastructure landscape in 2024. Building products that assume current infrastructure constraints as permanent is a mistake. Building products that are designed to take advantage of improving infrastructure — without depending on it being perfect today — is the right frame.
The data centre problem in Africa is being solved. Not as fast as the demand requires, but faster than it has been solved before. Operators who understand the infrastructure layer are better positioned to make decisions that age well.