The thesis is clean, which is part of the problem. Kenya has abundant geothermal power in the Rift Valley — cheap at wholesale, low-carbon, and available around the clock. Site data centers next to it, the story goes, and Kenya becomes the region’s compute hub on the back of the cleanest baseload in Africa. It is repeated often enough to sound settled. It is not.
Why the load stays in Nairobi
Power is one input to a data center, and rarely the one that decides location. Compute wants to sit close to fibre routes, to peering and the internet exchange, to the customers it serves, and to the people who operate it. In Kenya, those things are in Nairobi. A facility at the geothermal fields buys cheap electricity and pays for it in latency, connectivity, and operational distance.
So the capacity that is actually live — a modest but real and growing base across a handful of operators — sits in and around Nairobi, next to the fibre and the customers, not at the wells. The map of where the power is cheapest and the map of where the data centers are built do not overlap.
The flagship that tested the thesis
The clearest test of the geothermal-siting story was the large hyperscale campus proposed at Olkaria — a roughly billion-dollar project explicitly built on the site-at-the-source logic. It stalled, reportedly on grid-capacity and payment-guarantee questions rather than on ambition. The arithmetic underneath is sobering: a 100 MW facility would draw a large share of national peak demand, and the grid cannot yet spare that block of firm capacity on demand. Cheap energy at the wellhead is not the same as deliverable, firm, interconnected power at the scale a hyperscaler needs.
The market that is actually bankable
The real market is not the gigawatt headline. It is the steady stream of modular facilities — built in single-digit to low-double-digit megawatt increments, in Nairobi, by operators expanding as demand and power allow. That is where the deals are, and it is a different underwriting case from the one the siting thesis implies.
There is a genuine mechanism worth watching. A special-economic-zone arrangement offering direct power-purchase agreements at the geothermal fields — bypassing the distributor — changes the economics for a large, power-hungry, latency-tolerant load, and it has drawn scoping interest. It is real. It is also unproven at scale, and one stalled flagship is not yet a counter-example in either direction.
What this means for a deal
Geothermal siting is a credible thing to analyse and an easy thing to overstate. For anyone underwriting a data-center or power story here, the honest position is the careful one: treat the cheap-clean-power advantage as latent, not realised; separate deliverable firm capacity from wholesale price; and build the case on the modular market that exists rather than the gigawatt one that has been announced. Telling that story straight is not pessimism. It is the version that survives diligence.