Since January, campaigns and outside groups have put more than $45m into advertising that mentions data centres, across governor, House and Senate races. From January to August the split was $22m-plus behind Republicans and $21m-plus behind Democrats.
Both sides are making the same argument: the facilities get too many tax breaks, they push up electricity bills, and they need reining in.
A number that means the opposite of a controversy
Political money is normally a map of disagreement. Two parties spending comparable sums on an issue usually means they are spending it against each other.
Forty-five million dollars split down the middle to make one case is a different signal. It says the polling is not close, and both sides have concluded the only risk is being second to say it. Reuters/Ipsos in June had 77 percent of Americans worried about data centres raising electricity costs, with two-thirds of Democrats and half of Republicans opposed to one locally. Gallup earlier in the year had roughly seven in ten opposed to construction in their area.
Nobody is buying advertising in defence of the buildings.
Which means the interesting question is not whether policy turns against data centres. It is what the turn can actually reach.
The capital is already in the ground
Here is the timing problem, and this paper has spent the fortnight documenting its parts.
Aggregate hyperscaler capital spending is running toward $770bn this year, at around 94 percent of operating cash flow, with the balance financed by bonds issued long against useful-life estimates the issuers set themselves. The racks are ordered years before the power arrives, which is why a data centre is a chip buyer before it is a power buyer. The electrical apparatus is on lead times measured in quarters, which is why Eaton is doubling US enclosure capacity. Grid interconnection queues run for years.
Every one of those commitments predates the politics by a considerable margin. A project that has land, permits, an interconnection agreement and a signed tax abatement is not reachable by a campaign advertisement, and mostly not by the legislature that follows it either. Contracts and vested permits are durable in precisely the circumstances where a legislature would like them not to be.
So the political energy will land on the next cohort — projects not yet permitted, abatements not yet granted — while the ones people are actually angry about proceed on schedule.
What the fight is really about
Strip the advertising down and the two complaints are specific, and only one of them is about data centres at all.
Tax abatements are a local fiscal question. A county grants a decade of property tax relief to attract a facility that employs relatively few people once built, and residents discover that the assessed value is on the roll and the revenue is not. That argument is winnable at the county level and is being won already in several places.
Electricity bills are a rate case. When a large load connects, somebody pays for the transmission and generation it requires, and the question of whether that cost sits with the new customer or is spread across every ratepayer is decided by a public utility commission, in a proceeding, using cost-allocation methodology that almost nobody attends.
That is where this issue will actually be resolved. Not in a midterm campaign — in dozens of state utility commission dockets about large-load tariffs, most of which are open now and none of which will produce an advertisement.
The awkward part for whoever wins
A candidate elected on this has promised something the office may not be able to deliver, and the two available levers cut in different directions.
Denying abatements is straightforward and forfeits the investment, which is a real trade a community is entitled to make. Making large loads pay their full cost of interconnection is the better policy and is harder, because the utility's incentive runs toward building rate base and spreading the cost, and because the developer will argue that a cost-reflective tariff simply sends the facility to the next state.
Both of those arguments will be made in a venue with no press coverage, by lawyers, about tariff schedules, after the election that was fought on the subject is over.
What to watch
Not campaign advertising, which will keep rising through November and tells you only what is popular.
Watch state utility commission decisions on large-load tariffs — whether new data centre connections are charged the incremental cost of the capacity they require, or whether that cost enters general rates. That single methodological choice determines whether the seventy-seven percent worried about their electricity bill were right to be.
It is being decided this year, in public, in documents anybody can read, and it will not be mentioned in a single one of the advertisements.
The figure of more than $45m spent since January on political advertising mentioning data centres in gubernatorial, House and Senate races, and the near-even split of over $22m for Republican candidates and over $21m for Democrats with independents accounting for the rest, are from an NPR analysis of AdImpact data published on 5 September 2026. The June Reuters/Ipsos finding that 77 percent of Americans are concerned about data centres raising electricity costs, with two-thirds of Democrats and half of Republicans opposed to one locally, and the Gallup finding that roughly seven in ten Americans oppose local construction, are as reported in that coverage. Capital spending and bond issuance figures referred to here are from this publication's reporting of 2 and 4 September 2026. The analysis is our own.
Topics aidata centrespoliticsenergy
