AI’s Surging Power Demands Put Spotlight on Reliable Energy Sources

Artificial intelligence is a power-hungry industry whose energy needs are exploding. Utilities across the nation are seeking record rate increases, driven in part by the massive electricity requirements of AI data centers.

The demand for reliable, around-the-clock baseload power has triggered a backlog of orders for new gas turbines, underscoring the need for dependable energy sources to keep this sector running.

A new report from the Electric Power Research Institute and Epoch AI paints a clear picture: AI’s power appetite isn’t slowing down. Training today’s large, advanced AI models already requires between 100 and 150 megawatts each. By 2030, that figure is projected to soar past four gigawatts per model. This is more than many cities consume in total.

The report calls this “the fundamental driver of the growth in power demand for AI training,” fueled by increasingly powerful AI chip clusters whose rapid scale-up has outpaced efficiency gains.

Currently, total AI power capacity in the U.S. stands at around five gigawatts. By the end of the decade, it could reach 50 gigawatts.

50 GW is enough to power 37.5 million to 50 million homes.

Historically, AI training has needed large, localized power supplies. But as these needs escalate, building ever-larger dedicated data centers is not a sustainable solution.

Meeting the challenge will demand innovative strategies and a recognition that gas-fired generation, with its ability to provide steady, on-demand baseload power, will be essential to keeping the AI revolution moving.

Tech leaders and utility managers realize that wind and solar can contribute to the power mix, but they are insufficient on their own for the 24/7 reliability AI data centers require. Until long-duration energy storage scales significantly, baseload from natural gas (and in some regions nuclear or hydro) will remain essential.