AI's power grab: The hidden toll on a handful of states

While anyone in the country can use ChatGPT, just a handful of states are shouldering the brunt of the energy burden required to power it.

Generative AI models answer millions of user queries from across the nation every day, yet only 15 states generate 80% of the energy required to power them: Virginia delivers the most electricity to data centers, followed by Texas, California, Illinois, and Oregon, according to a 2024 Electric Power Research Institute (EPRI) report. 

Data centers—facilities responsible for training and running generative artificial intelligence (AI) models like ChatGPT—are often built in areas that have dependable power, fast internet, low electricity and property costs, and favorable policies. As a result, certain counties and cities have become data center hubs, skewing electricity demand in those regions. 

U.S. data center electricity consumption is projected to rise by as much as 130% by 2030. To keep up, the electric grid must simultaneously expand alongside data centers. However, because data centers tend to be clustered together in specific locations, the cost of building new infrastructure falls disproportionately on homeowners in those places. The aging grid wasn’t designed to accommodate today’s rising household electricity needs, let alone the massive demands from these geographically concentrated data centers, so utility companies must invest in expensive upgrade projects. Those costs, which are ultimately passed on to customers, could cause electric bills to increase 70% by 2029.

“Data centers have to compete for electricity with everyone else, and that creates scarcity,” Yury Dvorkin, associate professor at Johns Hopkins Whiting School of Engineering, said to EnergySage. “We have essentially two problems: We don't have enough supply [to meet data center demand], and we don't have the omnipotent grid that can move power from location A to location B.”

For homeowners in data center-heavy regions, home solar panel systems offer a way to take control and avoid falling prey to impending electricity rate increases. By generating electricity independently, solar-equipped homes can reduce or eliminate their electric bills, insulating themselves from rising rates tied to utility infrastructure upgrades. 

As grid strain grows due to substantial demand from AI and other energy-intensive technologies, solar energy also provides resilience, helping to ease grid strain during periods of peak energy demand. 

“It’s a unique situation where we have an immediate shortage—which is fairly large—and quite a bit of uncertainty on how to move forward. A lot can be done on a mid- and long-term basis, but very little can be done in the short term because generation and transmission projects aren’t built quickly,” Dvorkin said.

“We need to think about fast transmission options as a short-term solution. We can deploy solar and other forms of distributed energy resources fast, and battery technology is available on relatively short notice,” he said.“This can help reduce the grid system's peak and provide relief.”

Data center power consumption by state

Northeast

Midwest

South

West

*Only states with significant data center load are included.

Just 15 states accounted for 80% of national data center electricity consumption in 2023, according to EPRI. Virginia—particularly the northern part of the state—leads the nation by a wide margin, with more than a quarter of its electricity supply used solely to power data centers.

The other top states for data center electricity use are Texas, California, Illinois, Oregon, Arizona, Iowa, Georgia, Washington, Pennsylvania, New York, New Jersey, Nebraska, North Dakota, and Nevada, in that order.

Data center electricity load in key states, 2023

As AI becomes a part of daily life, the data centers powering it are growing in both size and number. But this rapid growth, combined with lopsided electricity demand, puts significant pressure on the energy systems in the regions where these facilities are most concentrated.

To put it into perspective, a 100 megawatt (MW) data center consumes as much power as 80,000 homes, per EPRI. That’s a lot of energy. These high-capacity facilities create real congestion on the grid—especially when multiple centers are built in the same geographic area. That congestion strains existing infrastructure, increasing the risk of outages and overall grid instability for everyone in the area. 

Utility companies often point to this strain when proposing large-scale upgrades, and the cost of those investments is passed on to customers through higher electricity rates.

The “Big Beautiful Bill” makes a tough grid situation worse

Much of the U.S. electric grid is already outdated and overloaded, requiring substantial upgrades simply to meet today’s needs, let alone tomorrow’s. As data center growth accelerates, recent unfavorable legislative changes to federal clean energy policy will further push the grid’s limits in the coming years. 

The newly passed reconciliation package—branded the “Big Beautiful Bill” by President Trump—eliminates or phases out many significant federal clean energy tax credits that were previously helping drive down the cost of solar and battery storage projects, which in turn help power data centers and mitigate grid strain. Under the new law, the 30% residential solar tax credit will abruptly disappear on January 1, 2026—nearly a decade ahead of schedule and with no phase-down period. 

“These incentives aimed to make electricity produced by certain technologies cheaper. Remove those, and you accelerate the problem by returning the cost back up,” Dvorkin said. “Not only is demand driving the price, but the production itself becomes more expensive.”

In effect, the new law undermines the very technologies that make the grid stronger, more flexible, and cleaner, putting more stress on already insufficient infrastructure. 

Under Trump’s law, developers may find it harder to secure financing for clean energy installations—which will likely delay or cancel projects expected to help meet rising demand. For utility companies already citing reliability concerns, the rollback gives them further reason to double down on carbon-emitting fossil fuels that harm the environment.

As data centers drive up electricity demand, utility companies are investing heavily in new power plants, transmission lines, and substations—costs that are ultimately reflected in electricity rates. Homeowners are now increasingly on the hook for more expensive electric bills, especially in areas where this demand is most concentrated.

Rooftop solar offers a defense against these rising prices and instability. By generating their own solar electricity, homeowners can reduce or eliminate their electric bill while contributing to a more reliable grid.

“It takes years of contracting, permitting, construction, and delays to enable manufacturing and the supply chain for something like a gas turbine,”  Dvorkin said. “Deployable distributed energy resources can partially alleviate the strained capacity [from data centers] immediately: When you have 1,000 households delivering 10 kilowatts to the grid, you have the capacity equivalent to a small gas turbine.”

Rooftop solar can improve energy security for homeowners, too. When paired with battery storage, it can keep essential devices running during outages, offering peace of mind as the grid faces increasingly strained resources and growing blackout risks.

As AI’s appetite grows, homeowners who go solar can stay one step ahead—powered, financially protected, and off the hook—especially those who install systems by December 31, 2025. After this year, Trump’s clean energy policy, or lack thereof, will take effect, adding about $9,000 to upfront solar costs for homeowners. When it comes to solar, the AI boom isn’t waiting, and neither should you.