Microsoft’s $1bn Kenya data centre stalled over capacity payment dispute

TL;DR:

• A Microsoft data centre project in East Africa, announced as a $1 billion partnership with UAE-based AI firm G42, has been delayed by disagreements with the Kenyan government over Microsoft’s request for guaranteed annual capacity payments, Bloomberg News reported on Sunday.
• Microsoft and G42 may ultimately decide to scale back the project, according to Bloomberg, though Kenya’s information ministry says the talks are still active and not failed or withdrawn.
• Resultsense view: this is the first widely-reported case of a hyperscaler AI data centre stalling on commercial terms rather than power, planning or political risk — and the structure of the dispute (capacity guarantees) is the same financial mechanism UK government would face if it tried to underwrite domestic AI Growth Zone build-out at scale.

The Kenya facility, announced in May 2024 during Kenyan President William Ruto’s state visit to Washington, was designed to run entirely on geothermal power and provide UK-style cloud-region access to Microsoft Azure for East Africa. Microsoft, G42 and Kenya’s Information Ministry did not immediately respond to Reuters requests for comment, and Reuters could not independently verify the Bloomberg report.

What is being disputed

Microsoft and G42 asked the Kenyan government to commit to paying for a certain amount of compute capacity annually — effectively a minimum revenue floor — but the talks broke down when Kenya could not provide guarantees at the level Microsoft requested, the Bloomberg report said. Kenya’s principal secretary at the Ministry of Information, John Tanui, told Bloomberg the project “is not failed or withdrawn” and that “the scale of the data center they wanted to do still requires some structuring”, with power requirements still under discussion.

Capacity guarantee structures of this kind are common for large telecommunications and infrastructure projects, but rare for cloud compute build-out, where hyperscalers typically take demand risk themselves and price for it. The Kenya dispute is significant because it shifts that risk allocation: Microsoft, with G42’s backing, is effectively asking the host government to act as anchor customer for the facility.

Why hyperscalers are negotiating differently

The shift reflects the underlying economics of AI-era data centres. Power requirements, cooling needs and equipment costs have climbed sharply, and the certainty of long-term off-take has become more important to project finance than it was for general-purpose cloud build-outs. In emerging markets with smaller domestic enterprise IT spending, that off-take case is harder to make without government commitments.

The Kenya facility’s all-geothermal power configuration is a genuine advantage in any future AI workload that has to defend its emissions profile, so the project’s strategic logic remains intact even as the commercial structure is being renegotiated.

UK relevance

The dispute is more relevant to UK AI infrastructure debates than the geography might suggest. The UK government’s AI Opportunities Action Plan has committed £2 billion to expand UK compute capacity twentyfold by 2030, with five AI Growth Zones and up to £500 million for a Sovereign AI Unit. Any equivalent UK push to anchor large data centre projects will face a comparable financial-structure question: how much demand risk does the public sector take, how much do hyperscalers absorb, and where does long-term capacity pricing land?

Kenya’s experience is a useful data point: when a hyperscaler asks a government to underwrite capacity guarantees, the government can say no without killing the project — but the timetable slips, and the project’s scale is renegotiated downwards.

Looking forward

Watch for whether Microsoft and G42 announce a scaled-down Kenya facility or restructure the deal entirely. Either way, this case is now a working example UK officials and procurement teams should study before signing any anchor-tenant or capacity-floor commitments with hyperscalers for domestic AI Growth Zones.