Last month, a bipartisan group of senators introduced the long-awaited Bipartisan American Affordability and Jobs Act, or BAAJA. The sweeping bill would make significant changes to the nation’s infrastructure permitting processes. Buried inside the 400-plus-page bill is a little-known section that would fill a major gap in our understanding of the data-center boom: it would direct the Energy Information Administration (EIA) to begin collecting data on how much electricity data centers use.
Remarkably, no federal agency regularly measures data-center electricity usage. In a new report published this month by Advancing American Freedom, I highlight the consequences of this data gap. Policymakers are debating rules for data centers without reliable data on how much power the facilities actually consume. BAAJA can help close that gap—and give America a clearer picture of the data-center boom and its implications for the nation’s electric grid.
As I explain in my report, reliable information about data centers’ electricity usage is sparse. In March 2026, EIA launched voluntary pilot field studies to gather information about the facilities’ electricity consumption. The studies targeted 196 companies in Texas, Washington State, Virginia, and the District of Columbia. Beyond those pilots, policymakers rely on estimates from Lawrence Berkeley National Laboratory, the Electric Power Research Institute, and others.
This comes at a time when America’s electricity demand is rising after years of stagnation. Overall U.S. electricity consumption grew more than sevenfold from 1950 to 1980 and nearly doubled again by 2008. Consumption flatlined until 2020 but has grown upwards since. Berkeley Lab estimates that data centers accounted for less than 5 percent of American electricity usage through 2024, while projections for 2030 range from 9 to 17 percent.
EIA Administrator Tristan Abbey acknowledged the challenge when he announced the pilots, noting that the agency’s consumption surveys “were conceived decades ago.”
Unfortunately, utilities and grid operators don’t necessarily have better information. Georgia Power’s 2022 Integrated Resource Plan predicted less than 400 megawatts of peak load growth through the winter of 2030–31. Just one year later, the utility raised that estimate to 6,600 megawatts, almost 17 times the previous forecast. Its 2025 plan raised the projection again to 8,205 megawatts. PJM Interconnection, the nation’s largest grid operator, similarly went from forecasting 4,312 megawatts of growth to 2030 in 2021 to 29,739 megawatts in 2025, before revising downward for the first time in 2026.
This uncertainty can have real consequences. If utilities underestimate future electricity demand, reliability suffers. If they overestimate, ratepayers end up paying for unneeded generation and transmission.
States have their own difficulties accounting for data-center energy usage. Thirty-eight states offer tax exemptions for data-center equipment, often including electricity used by those facilities. Ohio estimated that its exemption would cost $130 million in fiscal year 2024 and $136 million in fiscal year 2025. In reality, the numbers were way off: the exemption cost the state $555 million and $1.6 billion, respectively—roughly four and 11.5 times the original projections. These figures were not publicly available on a state website but were instead first revealed by an Ohio media outlet earlier this year.
As Ohio state representative Adam Holmes, who chairs a legislative committee on data centers, put it: “Let’s figure out what the heck is going on, and do it in a public forum.” He’s right. States should publish what these exemptions actually cost and review them regularly.
Other states face the opposite problem: anticipating data-center growth that doesn’t materialize. Minnesota barely changed its estimates of data-center tax exemptions between 2022 and 2024. One might conclude that’s because Minnesota’s forecasters did a better job, but the data suggest otherwise. Between 2019 and 2025, Minnesota’s commercial and industrial electricity sales fell 5 percent, while sales in Virginia grew by 35.7 percent, and in North Dakota by 53.3 percent. Minnesota’s data-center boom never arrived.
Our Advancing American Freedom report offers several recommendations to address these challenges.
First, Congress should establish a statutory basis for data-center load measurement. In March, EIA began collecting some of this data on a voluntary basis, but Congress should make that collection broader and more permanent. Section 2114 of BAAJA would take an important step by requiring EIA to add data-center electricity demand to its annual survey of electricity suppliers. Since that survey is filed by suppliers, not customers, this change would impose no new reporting requirements on data centers or other large-load customers themselves.
Second, states should authorize consumer-regulated electricity (CRE), an approach laid out by Travis Fisher and Glen Lyons in a 2026 Cato Institute paper. Under their approach, private companies could build and operate their own power systems to serve new industrial customers, including data centers, through voluntary contracts. Because such a utility serves only new load, on an islanded system and at contract prices, inaccurate forecasts would not expose existing utility ratepayers to the costs of unnecessary infrastructure investments. Several states, including New Hampshire, Utah, and Ohio, have taken steps toward enabling CRE, while Arkansas Senator Tom Cotton has introduced federal legislation to facilitate it.
Another approach is to adopt large-load tariffs, which are special rates for large customers. These tariffs can require data centers and other major electricity users to cover the costs of new infrastructure needed to serve them. But such measures address only part of the problem. Ultimately, meeting growing electricity demand will require expanding the nation’s energy supply.
Lastly, states should reject data-center moratoriums. At least 16 states considered bills this year to halt construction of new data centers. These proposals range from one-year pauses to prohibitions running to 2030. New York’s proposal passed the legislature, Maine’s was vetoed by the governor, and Maryland’s failed. The premise of these bills is that the grid is not ready and that costs might shift to ratepayers. But the answer to a grid that is not ready is to make it ready—through permitting and interconnection reform, an all-of-the-above energy strategy, and policies that protect existing ratepayers.
As AI continues to proliferate, America needs policies that provide abundant and reliable electricity. But before that, policymakers need actual data. Section 2114 of BAAJA delivers that data. Congress should make sure that provision becomes law.
Kevin Dayaratna, Ph.D., is Vice President of the Institute for Statistical Policy Analysis at Advancing American Freedom.
