Google to Limit AI Due to Energy Shortage

Google has agreed to limit the energy consumption of its data centers used for training AI models, responding to pressures on power grids.

Agreements with Indiana Michigan Power and the Tennessee Power Authority stipulate reductions in energy consumption during periods of high demand. Google states that these agreements help reduce the need for new electricity infrastructure.

A study in the journal Joule estimates that by 2027, AI data centers will consume 85 to 134 terawatt-hours annually. Goldman Sachs predicts that by 2030, AI will require an additional 370 terawatt-hours per year, equivalent to the total consumption of Britain and the Netherlands in 2023.

Google admitted that its emissions increased by 13% in 2023, making the company's climate goal of zero emissions by 2030 unlikely.