AI data centers consume record electricity as companies race to build massive facilities, prompting urgent focus on energy-saving strategies

U.S. data centers consumed some 224 terawatt-hours of electricity in 2025—more than 5 percent of the country's electricity use, up from an estimated 1.9 percent in 2018, before the mainstream surge of generative AI.

Processing a median-length text prompt with Google's AI assistant Gemini consumes around 0.24 watt-hours. Training GPT-4 guzzled between 50 and 60 gigawatt-hours of electricity. New AI-focused data centers often consume a gigawatt or more of electricity, compared to pre-AI data centers that consumed around 100 megawatts.

Without energy-saving strategies, according to a 2025 estimate, U.S. data centers could release the equivalent of 24 to 44 megatons of CO2 annually. Much of this demand is met by fossil fuel plants because data centers are often built in places without abundant renewable energy sources, and tech companies' offset strategies keep emissions in stasis rather than reducing them to net zero.