A new study is the first to quantify AI's climate impact across the entire power sector, finding that productivity gains AI delivers to fossil fuel extraction outweigh the emissions savings it enables in renewables. Researchers modelled 64 scenarios and concluded that net carbon pollution would rise, undercutting claims that AI is primarily a boon for the green energy transition. This matters because most existing research has focused narrowly on comparing data centre energy use against AI's emissions savings, ignoring the extra pollution generated as AI boosts productivity in drilling and gas extraction.

The study found net annual emissions increased by 0.47 to 1.8 gigatonnes, equivalent to roughly 1-5% of the energy sector's yearly total, and emissions only fell in scenarios where AI did not improve fossil fuel productivity. For a break-even outcome, renewable productivity gains from AI would need to outpace fossil fuel gains by at least fourfold, researchers said, and co-author Holly Alpine noted fossil fuel AI applications are already deployed at scale via real industry contracts, while renewable uses remain largely experimental. The International Energy Agency estimates AI could raise recoverable oil and gas reserves by 5% and cut deepwater project costs by 10%, with companies including Saudi Aramco and Equinor already crediting AI for boosting well discoveries, and Rystad Energy projecting AI could generate close to £370bn in cumulative value for fossil fuel firms by 2030.