AI and data centers mattered to 56% of voters, but trailed other issues

Voters could assign importance to multiple issues. These are separate responses on the same scale, rather than shares of a single-choice ranking. AI and data centers came last among these five tested issues; the results do not rank them against every possible voter concern.

The sharper evidence of resistance comes from the separate construction question: 76% opposed a new local data center and 24% favored one. On AI itself, 57% expected more harm than help to society, 24% more help and 18% neither. Yet neither party commanded much confidence: 43% chose neither as better able to handle AI, compared with 29% choosing Democrats and 27% Republicans.

The survey was conducted by NBC News with other news organizations across 45 key states, combining Election Day interviews with mail, phone and online methods to include early and mail voters. The tables credit survey provider SSRS. They do not give sample sizes or margins of error for the AI questions, so the figures should not be read as exact population measurements or evidence of a meaningful two-point party confidence gap.

Restrictions advanced, but industry backing was no automatic liability

Election-night reporting said at least 35 local ballot measures banning or limiting data center construction nationwide appeared set to pass. A separate account described bans or other restrictions passing in 15 Ohio municipalities. The Ohio figure is not an additional national tally, and the broader count was provisional.

The electoral stakes extended to candidates' records. Sherrod Brown, the former Ohio senator whom cryptocurrency interests helped defeat in 2024, won his Senate race after campaigning against rising energy costs associated with AI data centers. The election report put cryptocurrency PAC spending against him at $55 million. That race brought infrastructure affordability into a contest already shaped by technology-sector money.

Political funding was itself divided: the report also described donors working in and around frontier AI labs contributing millions to pro-regulation candidates. It identified victories for Leading the Future, a pro-AI political action committee that backed candidates in both parties. Florida's incoming Republican governor, Byron Donalds, had repeatedly supported data center construction and won with millions in contributions from the committee and developers. Leading the Future also backed Democrat Jared Moskowitz, who won the U.S. House race in Florida's 25th Congressional District against a Republican supported by pro-regulation committee Public First Action.

The size and timing of the industry's financial effort also need separating. An October 31 report said Leading the Future had launched with $100 million in pledges, but its network had spent less than $10 million in the general election, compared with roughly $26 million in the primaries. The report cited federal filings for October spending of $4.1 million through affiliated committees despite nearly $24 million in cash on hand. Committee spokesperson Jesse Hunt described a strategy spanning multiple election cycles to cultivate allies in Congress. Pledged money was not equivalent to spending against candidates in Tuesday's contests.

JLARC’s 2024 modeling estimated monthly increases of $14–$37 by 2040 for a Dominion household using 1,000 kWh. These projections use constant 2024 dollars and exclude distribution charges and some transmission costs. Source: Virginia Joint Legislative Audit and Review Commission.

Why fair current rates do not settle the power-cost dispute

In the exit poll, 86% were very or somewhat concerned about utility costs, including electricity, natural gas and water. That answer does not identify the cause of higher bills. An earlier Virginia legislative review supplies a more specific account of the infrastructure tradeoffs.

The Joint Legislative Audit and Review Commission, the state's legislative oversight body, studied the industry in 2024 after previously reviewing its tax incentive in 2019. Its setting matters: the report identified Northern Virginia as the world's largest data center market. Its researchers commissioned independent economic and energy modeling.

Its cost study found that rates examined at the electric utility Dominion Energy and two member-owned electric cooperatives appropriately allocated existing costs to responsible customers, including data centers. But meeting rapidly growing demand would require new generation and transmission, while tighter power supplies and greater reliance on imports could raise energy prices for other customers.

For a Dominion residential customer using 1,000 kilowatt-hours a month, the review estimated that data center growth could add about $14–$37 to monthly generation and transmission charges by 2040, in constant 2024 dollars. The lower figure came from a scenario with half of unconstrained demand; the upper end assumed unconstrained demand and compliance with the state's clean-energy requirements. Unconstrained demand means demand before limits such as the ability to build enough infrastructure. Even the lower-growth scenario was judged difficult to supply.

These were modeled increases, not observed bills or a national forecast. They excluded distribution charges and some transmission costs, and assumed the rate-allocation practices examined in 2024. They do not establish what customers are paying in 2026.

There are substantial benefits on the other side of the ledger. The review found construction supplied most of the industry's economic impact. Industry representatives described about 50 full-time workers, roughly half contractors, at a typical 250,000-square-foot facility. Separately, they described about 1,500 workers on site at peak construction. An individual building generally took 12–18 months to construct; peak construction staffing is not a permanent workforce.

Local tax receipts can be large: the study put data center revenue in Loudoun County at $733 million, or 31% of total local revenue, in fiscal 2023. Incentives reduce other tax collections. Qualifying data centers and tenants reported $928.6 million in sales-tax savings statewide that year, including state, local and regional portions. The county receipts and statewide exemption measure different taxes and geographies; subtracting one from the other would not give a net fiscal balance.

The review also documented the competition for investment: in 2024, 34 states offered data center sales-tax exemptions. Some AI workloads, such as large language model training, could operate farther from customers than services requiring very fast responses. Local restrictions can therefore affect where facilities are built as well as whether projects proceed.

JLARC’s 2024 maps show zoning beside an existing data center near Great Oak in Prince William County and a planned facility near Bren Mar in Fairfax County. Source: Virginia Joint Legislative Audit and Review Commission.

The next decisions are about siting and financial obligations

Final local results and the restrictions they enact will show how much proposed development the vote actually constrains. The national poll supplies no choice among outright bans, temporary pauses, stronger siting standards or different utility contracts.

The Virginia review illustrates alternatives to blanket restrictions. It recommended clearer local authority to require sound studies and address persistent low-frequency noise, while noting that most facilities had not generated noise complaints. It also identified separate rate classes, revised cost allocation and contractual minimum payments as ways to reduce risks to other electricity customers. Another risk needs attention alongside rising demand: infrastructure can leave customers with stranded costs if forecast data center demand fails to materialize.

The review warned that slowing growth in Virginia might still leave its customers exposed to higher regional energy prices if development shifted to other states sharing the PJM power market. A local construction restriction does not by itself settle regional electricity costs.

For officials considering those choices, the unresolved test is concrete: which projects can secure acceptable sites and sufficient power, what revenue and lasting employment they offer, and which customers remain liable for infrastructure costs if developers use less electricity than promised.