Reading East Africa’s compute landscape without mistaking signals for orders

A regional compute map becomes useful when it helps someone decide what to verify next. Putting organisations on a page is only the beginning.

One organisation can play several roles

A cloud integrator may sell GPU capacity, host customer equipment and purchase infrastructure for its own service. A data-centre operator may provide power and space without selling the GPU service running inside it. Exclusive labels obscure these relationships.

SSA’s index therefore records capability separately from organisation type. Servernah’s published A100 offering is evidence of advertised GPU services. Colocation and announced future supply remain separately labelled. Current inventory, booking dates and commercial terms still require a quote. Servernah ↗

A workload establishes relevance

Sunbird AI’s model documentation describes a concrete regional-language application. That makes it relevant to compute demand analysis. It does not establish that the organisation is currently buying capacity, that its budget is approved, or that it would purchase from a new provider. Sunbird documentation ↗

The same discipline applies to tenders. A published requirement is stronger evidence of a defined buying process than a broad AI announcement. Once a tender closes, its status must change; an old notice cannot remain an open sales lead indefinitely. An award establishes a different fact again.

Keep four questions separate

1. Technical relevance: what workload or infrastructure is evidenced?
2. Commercial intent: is there an identified purchase, approved budget or procurement process?
3. Delivery: who can supply the required service, where and when?
4. Counterparty diligence: what financial and contractual evidence supports the proposed transaction?

A financing announcement does not answer all four. Neither does a large parent company, a national AI strategy or a data-centre power rating. Preserve the legal entity, source date, claim scope and remaining question.

Make the coverage boundary visible

The current SSA index reviews selected organisations in Kenya, Uganda, Tanzania, Rwanda and Ethiopia. It is a bounded public-source dataset, not a complete census of African compute or a credit rating. Its records distinguish documented activity from watchlist entries and do not claim verified GPU offtake contracts.

Useful updates add evidence: a service specification, a quote, a procurement status change or a clearly attributed financial disclosure. Increasing the number of logos without improving qualification makes the map less useful.

Inspect the current evidence map ↗ or use the project workbench to connect a regional lead to a concrete workload and acceptance plan.