The $1bn budget was split between Microsoft and G42, with the Kenyan government providing land, power-purchase terms, and regulatory facilitation. Geothermal power, available cheaply and abundantly in Kenya, was central to the pitch.

Microsoft has not formally cancelled the project. The Kenyan Ministry of Information told local press that the project remains live: “It is not failed or withdrawn.”

Bloomberg's sources frame the situation as a delay, not a death; the companies may rescope the facility to a smaller footprint that does not require the same offtake guarantee.

What is clear is that the original commercial structure does not currently work.

The offtake demand is the unusual element. Hyperscalers do not typically ask host governments to guarantee compute purchases; the same offtake question Western utilities are answering, where utilities and offtakers commit to long-term capacity contracts that anchor a build.

Microsoft applied the same logic in East Africa because the local enterprise market on its own is not large enough to justify a $1bn facility, and because financing partners are increasingly insisting on guaranteed revenue floors for projects in higher-risk geographies.

Kenya's government, facing a tight fiscal environment and contesting IMF programme conditions, could not commit to the multi-year buy-in.

The political layer is real but secondary. Kenya is finalising a national AI strategy and has framed digital infrastructure as a development priority; cancelling the deal carries domestic political cost.

Equally, committing public budget to multi-year compute contracts when health, education, and infrastructure spending are under pressure is hard to defend. The mismatch is structural rather than personal.

G42's position is interesting. The UAE-backed firm has been on a global build, with significant capacity announcements in the US, Italy, and France over the past eighteen months.

G42's separate US data-centre footprint makes the East Africa pause look less like a strategic retreat than a regional reset. G42 has more cash, more flexibility about siting, and more leverage with non-US partners than its Microsoft co-investment historically allowed.

A scaled-back Kenya facility funded primarily by G42, with Microsoft as a service tenant, is a structure both parties could accept; whether they will is uncertain.

Microsoft, separately, announced a $329m expansion in South Africa last month, partly framed as a hedge against the East African delays.

South Africa's commercial market is larger, the regulatory environment more predictable, and the offtake question more easily answered by existing private-sector demand.

The expansion is incremental rather than transformational, but it ensures Microsoft's Azure Africa footprint continues to grow even if Kenya does not deliver on the original timeline.

For Kenya, the broader cost is reputational. The Microsoft-G42 facility was meant to anchor an East African digital hub spanning Rwanda, Uganda, Tanzania, and Ethiopia, with downstream effects on submarine cable landings, fintech build-out, and AI talent pipelines.

The stall does not foreclose those, but it requires the country to find a new anchor tenant or shift the model. The closest substitutes, Equinix and Africa Data Centres, do not bring the same hyperscaler relationship.

Whether Microsoft and G42 return to the table depends on whether Nairobi can offer a structure that gives the operators revenue certainty without the kind of guarantee that would breach its fiscal constraints.

A shorter contract, a smaller initial scope, a different financing partner: any or all of those could revive the deal. None of them is the original.

The Kenya story is also a useful data point for other African governments competing for hyperscaler builds. Nigeria, Egypt, and Senegal are all in different stages of negotiation.

The terms Microsoft attached to Kenya, including offtake guarantees, will be familiar to each of them when their own negotiations turn serious.