An Amazon Web Services data center is located next to single-family homes in Stone Ridge, Virginia, part of Northern Virginia’s vast data center market.
Getty Images
Americans are increasingly worried that data centers could drive up their electricity bills. Virginia just became the first state to tax the power they consume. The levy won’t lower household bills by itself, but it signals that states are beginning to question whether the AI buildout is an automatic economic win.
For a decade, the pitch to communities courting data centers was simple: the industry would bring investment, jobs and tax revenue, while the community supplied land and electricity. On July 1, Virginia changed the terms.
The state at the center of the world’s largest data center market began charging operators 1.1 cents for every kilowatt-hour they consume. It appears to be the first state tax in the country levied directly on data center electricity use.
The belief that data centers are almost pure economic upside for host communities is starting to crack, especially in places already dealing with rapid development. The pressure isn’t only environmental. The math behind the old deal may be shifting, and voters are reacting before regulators and utility plans have fully caught up.
The old math actually worked
The honeymoon wasn’t irrational. For most of the last decade, a big, steady electricity customer was good for everyone else’s bill.
A June 2026 EPRI working paper found that from 2015 to 2024, data center growth was linked to a modest decline in average retail rates. The logic is simple. The grid carries large fixed costs that must be recovered regardless of how much electricity customers use, so a big, steady customer spreads them across a bigger base and average prices drift down. States weren’t crazy to compete. The catch is that this only works while the system has spare capacity.
Then demand began outrunning the grid
A federal laboratory analysis projects that U.S. data center electricity use could reach 325 to 580 terawatt-hours by 2028, up from 176 terawatt-hours in 2023. When a single campus can require power on the scale of a city, utilities must build new substations, transmission lines and generation capacity, often held up by transformer lead times that now stretch into years.
A separate Dallas Fed working paper models what may come next. Its model estimates that existing data centers have already raised wholesale electricity prices by 3% to 5% nationally, with larger effects in major data center corridors. Under its moderate buildout scenario, wholesale prices could rise about 20% by 2028. Under a high-utilization scenario, the increase could approach 50%.
Together, the studies show why data centers may have lowered retail rates in the past but raise costs as the grid becomes more constrained. The widely repeated claim that data centers lowered electricity bills came with an important qualifier: so far.
The fight turns on utility rules most voters never see. When a utility upgrades its system for a giant customer, that cost doesn’t automatically fall on the customer who triggered it. Depending on the contract and the regulator, some of those costs can enter the broader rate base and be recovered from other customers.
Utilities can impose line-extension charges, minimum-demand requirements or special large-load tariffs, but getting those protections approved can take years. Without those protections, other customers can end up covering part of the infrastructure built for a server farm they may never use. Gallup found that 71% of Americans oppose building an AI data center in their local area, including 48% who are strongly opposed.
The backlash is in the numbers
Data Center Watch, an advocacy tracker rather than an audited database, says opponents blocked or delayed at least 75 data center projects worth about $130 billion in the first quarter of 2026, roughly matching its tally for all of 2025. The group also reports that organized opposition more than doubled and now spans 49 states, and that legislators filed more than 300 data-center bills in the year’s first six weeks.
North Carolina offers another example. Its new budget repealed the electricity exemption on power bought by data centers while keeping capital incentives, a sign that states are reconsidering which incentives still make economic sense.
What the tax does, and doesn’t
Just over a penny per kilowatt-hour sounds trivial. At scale it isn’t. A 100-megawatt site running flat out owes about $9.6 million a year, and a gigawatt campus nears $96 million.
Virginia limited the impact by keeping the equipment sales-tax exemption, limiting general-fund revenue to $600 million a year and refunding excess collections to operators. The tax is currently authorized from July 1, 2026, through June 30, 2028.
But this shouldn’t be confused with direct relief for consumers. A tax on a data center’s electricity and a lower bill for the household down the road are different things. Virginia’s money goes to the state, not to ratepayers.
Whether household bills rise or fall depends on the rate-case machinery: how the utility classifies large loads, and whether contracts require large customers to cover the infrastructure built for them. Steel mills and chemical plants have fought over that for decades, because a utility can’t recover a stranded substation if the big customer leaves. Electricity isn’t simply a commodity that can be bought on demand.
The takeaway
The era of treating power as a cheap, invisible input to the AI boom is ending. For years, the public conversation focused on capital and chips. Increasingly, the limiting factors are grid capacity, local politics and who pays for new infrastructure.
One growing constraint is a community’s willingness to carry the load and whether residents believe they’ll be left paying part of the bill. If you’re underwriting a data center, or a deal that leans on one, power can’t be a fixed line at the bottom of the model. It needs its own risk column.
Virginia’s tax won’t lower household bills on its own. But it marks the end of the era when data center developers could treat power as cheap, abundant and politically invisible.