Google has announced that nuclear energy will partly power its future AI efforts.
A new deal will see Google purchase energy from “multiple small modular reactors” developed at Kairos Power in California.
Google plans to have the first of its reactors up by 2030, with additional ones due up and running between 2030 and 2035.
The power provided won’t cover the entire bill for Google’s AI operations, though.
The deal is estimated to provide the company with up to 500 megawatts (MW) in energy generation. For comparison, Google expended 25 terawatt-hours (TWh) of energy in 2023, according to its own environmental report. So if its energy use stays the same, the nuclear energy will account for less than a sixth of the company’s needs, before accounting for efficiency loss and varying demand levels.
Michael Terrell, Google’s Senior Director of Energy and Climate, wrote in The Keyword blog: “The grid needs new electricity sources to support AI technologies that are powering major scientific advances, improving services, for businesses and customers, and driving national competitiveness and economic growth.”
How much energy does AI use?
A single request to the ChatGPT chatbot is estimated to consume up to 2.9Wh, just over a fifth the power an iPhone 16’s 13.7Wh battery can store.
AI requests consume 10-30 times more power than specialised software — the International Energy Agency (IEA) estimates a Google search consumes around 0.3Wh, for example. However, that was before Google added AI summaries to its search results.
While some suggest AI is a route to tackling climate issues, its high energy cost is confounding.
A report published by the Guardian in September 2024 also claimed the emissions figures for the biggest tech companies — Google, Apple, Microsoft, Meta, and Amazon — were up to 7.62 times higher than the companies’ reports claimed.
