The $1 billion Microsoft–G42 data centre planned for Olkaria didn’t stall just because Kenya’s grid was too small. According to Bloomberg, citing people familiar with the private negotiations, Microsoft and its Abu Dhabi-based partner G42 had asked the Kenyan government to commit to paying for a fixed amount of computing capacity every year. When the government couldn’t provide that commitment at the level Microsoft wanted, the talks broke down.
That single detail reframes the story we told yesterday, the one President William Ruto told this week, and the one that has dominated coverage since. The grid story is true. A 1 GW facility was always going to be uncomfortable in a country with around 3,200 MW of installed capacity. But phase one was 100 MW, and Olkaria’s geothermal complex generates roughly 950 MW. Phase one was always within reach. It was supposed to be operational by May 2026, two years after the Washington announcement. That deadline has now passed. The binding question on phase one was never really the power. It was who carried the risk if East African demand for Azure capacity took longer to materialise than Microsoft’s capex schedule could absorb.
Microsoft wanted the Kenyan government to be that anchor. The government couldn’t take on the commitment Microsoft was asking for.
What Microsoft was actually asking for
Bloomberg’s phrasing is specific. Microsoft and G42 wanted the government “to commit to paying for a certain amount of capacity annually.” In commercial cloud and infrastructure deals, this kind of arrangement usually shows up as a minimum revenue commitment or a capacity reservation. The customer agrees in advance to pay for a defined block of compute, storage, and network capacity over a fixed period, regardless of how much actually gets used. The provider gets a guaranteed revenue floor in exchange for sinking the capital into the facility.
For a hyperscale build, this is not unusual. A facility of this scale costs about a billion dollars and takes years to deliver. Without a sizeable anchor tenant, the financial model becomes uncomfortable. The cleanest way to fix that is to make the host government the anchor tenant. From Microsoft’s side, asking for that guarantee is rational. From Kenya’s side, providing it is not straightforward, particularly at the level Microsoft wanted.
There is a familiar shape here. Kenya has spent the last decade trying to renegotiate its way out of legacy Power Purchase Agreements with Independent Power Producers, where Kenya Power agreed to pay for contracted electricity whether or not it was consumed. Those take-or-pay clauses have been blamed for keeping electricity tariffs high, and the Ruto administration came to power partly on the promise of cleaning them up. A new multi-year commitment with similar economic logic, signed in dollars, to a Microsoft-G42 vehicle, was always going to be a hard sell inside government. This particular comparison is ours, not Bloomberg’s. But it explains why the request landed where it did.
What this says about the public framing
Ruto’s “we’d need to shut off power for half the country” line, delivered in Nairobi this week, was technically accurate at full 1 GW build. It was also a cleaner thing to say in public than the truth that emerged today. Power capacity is a story Kenyans understand. A bilateral negotiation that broke down over a multi-year payment guarantee to a foreign cloud venture is harder to explain in a single sentence. It is also harder to explain right now, because Kenya is currently negotiating a new IMF programme after the previous one ended in April 2025, and the IMF has recently pushed Kenya to classify more government commitments as debt. A new dollar-denominated capacity guarantee to a Microsoft-G42 vehicle would have walked straight into that scrutiny.
The official line so far has been that nothing is wrong. On Saturday, a day before today’s Bloomberg report, Special Tech Envoy Philip Thigo pushed back on coverage that described the project as suspended, saying the President’s remarks had been misinterpreted and that the project remains a flagship initiative. Thigo’s statement focused on energy. It did not mention financing.
What to actually watch
Whether the deal is dead or paused depends on which version of “broke down” you believe. Negotiations break down all the time and get restarted in different shapes. A smaller phase-one commitment, routed through KenGen or a state-backed cloud entity rather than directly through Treasury, would be politically easier and fiscally smaller. If KenGen, a state-owned utility, signed the capacity commitment instead of Treasury, the obligation would sit on KenGen’s balance sheet rather than the national one. That removes the optics of a direct sovereign guarantee, even though the government still ultimately stands behind KenGen.
Meanwhile, the workloads keep moving. Oracle’s Nairobi cloud region went live on iX Africa earlier this year, anchored to a 4.5 MW carrier-neutral facility that doesn’t ask the government to underwrite anything. Carrier-neutral means iX Africa sells capacity to multiple cloud providers and enterprises rather than committing the whole facility to one anchor tenant. That spreads commercial risk across many customers instead of placing it on a single host government. Africa Data Centres is expanding on the same model. EcoCloud Olkaria is still on track at a much more realistic 60 MW.
Today’s reporting tells us the original framing of this story was incomplete. The grid genuinely cannot support a 1 GW facility. That part is true and worth understanding. But the deal also failed because Microsoft wanted Kenya to take on a long-term financial obligation that the government, given its current fiscal position and its political history with similar guarantees, was not in a position to take on. Both things were happening at the same time. The energy explanation became the public one because it is the easier one to communicate, and because power infrastructure is something the government can credibly promise to expand. A capacity-payment guarantee is harder to revisit because the disagreement is not about whether Kenya can build something. It is about whether Kenya can sign something. Until that question gets answered, even a smaller version of this project will be difficult to revive.
