The nation’s grid operators are now forecasting 8.2 percent growth in electricity demand over just the next five years, largely driven by the rapid addition of AI data centers. That’s equivalent to hooking up nearly 50 million homes to the grid by 2029.

It’s no secret, U.S. power supplies are already alarmingly tight, particularly in periods of peak demand. For the nation’s utilities and energy and reliability regulators, the answer to meeting that demand is increasingly running through the coal fleet.

Federal Energy Regulatory Commission Chairman Mark Christie recently pointed out in a Twitter thread, at moments of peak demand during the nation’s latest brush with severe cold, dispatchable generation provided the lion’s share of power. He wrote, “We need to stop the premature retirements of dispatchable generation and build more, otherwise we freeze in the dark. That is reality.”

Utilities are getting the message. One after another, they are pushing back or cancelling coal plant retirements. Georgia Power is the latest utility to recognize it needs its coal capacity to meet the soaring demand now on its doorstep. In a new integrated resource plan modelling its future investments, Georgia Power indicated that it is going to be upgrading and extending the life of units at two large coal plants to at least 2035. Those plants, Bowen and Scherer, had been scheduled for retirement in 2028.

Why the change? Georgia Power is facing enormous demand growth—growth that is coming overnight. The utility sees 8,200 megawatts of electricity demand growth by 2030 — about 50 percent more than current peak demand. And while electrification and the addition of major new manufacturing projects are driving some of the new demand, a utility executive told the Atlanta Journal-Constitution that 80 percent of the projected demand growth is coming from data centers.

More Efficiency May Mean Even Greater Power Demand

Of course, electricity forecasts are an imperfect science, especially when built around the uptake of a nascent technology. But underestimating the demand from AI could be a huge, missed opportunity with deep consequences for our economy.

While there has been some recent debate about just how much demand there will be from AI, especially after the emergence of Chinese DeepSeek AI, with much improved energy efficiency, we need to be careful to assume that more efficient AI means less power demand.

As Mark Mills, a leading thinker on the subject, recently observed, “the idea that a leap in energy efficiency solves the AI power-demand challenge is a misread of reality.” He goes further:

“By that logic, investors would have dumped shares in turbine manufacturers, airframe suppliers, and oil companies circa 1958, when PanAm began commercial service with the then-revolutionary Boeing 707. That aircraft technology leap brought not only far greater fuel efficiency than anything previously seen in commercial air service but also more than triple the passenger capacity. The result wasn’t fewer airplanes and less fuel used, but more of both. And that, in turn, drove the expansion in the associated aviation infrastructures.”

In other words, the more efficient AI becomes, the greater its application and speed of its growth. More energy efficiency could very well mean more energy use. Meeting it will require the dispatchable, fuel secure and readily available coal fleet.