The AI industry keeps talking about bigger models, faster chips, and trillion-parameter futures. What it talks about far less is the absolutely absurd amount of electricity needed to keep all of this running. That reality just hit a major roadblock in Kenya, where Microsoft’s proposed $1 billion AI data center project is reportedly facing resistance after government officials warned that the facility could consume so much power it might require “switching off half the country” to keep it operational.
Microsoft’s Kenya AI data center reportedly needs more power than the grid can comfortably handle
The project, announced in partnership with Abu Dhabi-based AI firm G42, was originally intended to bring a large Azure cloud and AI region to East Africa, powered by geothermal energy from Kenya’s Rift Valley. Initial plans reportedly targeted around 100MW capacity, with long-term ambitions stretching toward 1GW.
However, that scale is now becoming the biggest issue. Kenya’s peak electricity demand, as per Bloomberg’s report, already touched roughly 2,444MW earlier this year, meaning a fully scaled 1GW AI facility could consume an enormous chunk of the nation’s available power infrastructure. Negotiations between Microsoft, G42, and Kenyan authorities have reportedly stalled over power guarantees and infrastructure concerns, though officials insist the project has not been canceled outright.
The AI boom is quietly turning into an energy crisis nobody fully prepared for
Honestly, Kenya’s situation feels less like an isolated problem and more like a preview of what the global AI race could start looking like very soon. AI data centers are becoming so power-hungry that entire countries are beginning to rethink whether their grids can realistically support these projects without affecting ordinary citizens.
The uncomfortable reality is that AI’s massive energy demands are becoming harder to ignore globally, with data centers already consuming a significant share of electricity in major markets. Kenya simply ended up saying out loud what many countries will probably have to confront soon: powering the AI boom is starting to look like an infrastructure problem as much as a technology one.