Every use of AI requires massive amounts of data, meaning as AI has surged, companies have been building more and more data centers across the country. Those data centers also require lots of energy to operate, and that means they could soon require more energy than what’s available on the grid.

If AI’s energy demand outstrips electricity supplies, that has real impacts for Americans—like a higher risk of electricity outages, and higher energy costs. By 2029, consumers and small businesses could see their electricity bills increase 70% because of surging energy demand from AI data centers, according to a new report by the Jack Kemp Foundation, a Washington, D.C. think tank created by former Republican Representative Jack Kemp.

That means consumers will bear the brunt of AI’s increasing energy use, says Ike Brannon, coauthor of the study and a senior fellow at the Jack Kemp Foundation. This is already beginning to play out in Northern Virginia, which currently has the highest concentration of data centers in the world. In Virginia, data centers could use almost half of the state’s total electricity by 2030.

The surge of data centers in Northern Virginia—driven by the federal government and national security agencies—is already leading to utility increases. “We haven’t really increased [energy] supply all that much around here, but demand is going up,” Brannon says. In the areas of the state served by utility company Dominion Energy, one price metric for peak demands in 2025 increased from $29 to $444 per megawatt-day.

Without ways to increase the energy supply—like by building new power plants or more transmission lines—Bannon says the only way to deal with the demand is through price increases. Usually, electricity demand surges over only a few hours, like when people come home from work and turn their lights and TVs on. “But because the demand from AI is constant—these data centers run 24/7, demand doesn’t cease—the peaks and valleys you normally see across the 24 hours of a day, they’re getting narrower and narrower,” he says. “There’s just less capacity in the system.”

The Jack Kemp Foundation’s report uses Northern Virginia as a case study for this issue, since it has such a high concentration of data centers—nearly 300 across a few counties and more in the works. But that state isn’t alone; regions across the Mid-Atlantic, South Carolina, and even Oregon are beginning to face this issue. Monthly residential bills for households served by transmission company PMJ in Maryland, Ohio, West Virginia, and more are already expected to increase between 11% to 19% starting in June. (Average increases are around 2%.)

The report suggests a few policy actions, including limiting the subsidies these data centers get from local governments; encouraging data centers to be more energy efficient, and charging higher fees for AI data centers so they bear the additional costs of the energy they consume, rather than passing that cost on to customers.

Bannon also points to the need for more power plants, and says efforts to stop new energy sources, including nuclear plants, hinders a region’s ability to meet this increased demand. In Pennsylvania, a unit of the Three Mile Island nuclear power plant is set to be restarted in order to power Microsoft’s data centers; while nuclear fuel doesn’t create greenhouse gas emissions, some experts say there are still risks concerning nuclear waste, and that there are other clean energy options that can be prioritized.

There’s also the issue of transmission infrastructure. The country’s outdated power grid is already nearing a crisis point, and increased energy consumption from AI could exacerbate that issue. The report emphasizes the need to ease permitting for electricity generation and transmission infrastructure, in order to update the grid.