Did data centers swing the midterms? Not the way you think.

Were the 2026 midterms a referendum on data centers?

The morning after, plenty of people said yes. The Wall Street Journal called data centers "the big loser." The Washington Post called it "a nationwide backlash." Candidates fought at the stump over the right approach and the future of these facilities.

According to the data, that story is only partially right. We looked at every race where both candidates took a public position on data centers, analyzed against Spark's database of thousands of data center projects. In the end, we found, data centers did matter, but not how the headlines suggest.

Come January, Democrats will hold the governor's office in Texas, Pennsylvania, Georgia, and Ohio, four of the busiest data center states in the country. Governors shape the tax breaks, grid connections, and permitting rules that decide whether a project pencils out. In Texas alone, data center tax breaks run more than $1 billion a year. This new class of governors will decide the future of data centers.

Where candidates stood didn't matter

Voters don’t like data centers. But surprisingly, a position on the issue didn't win them any more votes.

We looked at 92 Senate, governor, and House races where both candidates took a public position on data centers. Some candidates wanted a moratorium, some wanted restrictions, and some wanted more data centers built faster. Voters didn't seem to care.

Across the country, Democrats ran about 6 points ahead of their baseline in presidential elections since 2016. That held whether a candidate backed a ban, restrictions, or more construction. Measured against their own party's average, the average candidate landed within a point, regardless of their preferred approach.

So voters weren't grading candidates on their data center talking points. But that doesn't mean data centers didn't matter.

The states building the most saw the biggest swings

In states with more data center activity, governor's races swung harder toward Democrats. In the states with the least activity, the average race moved 6.5 points toward Democrats. In the states with the most, it moved 13.3 points, about twice as much.

We measured the buildout by counting every data center project with permitting, construction, or opposition activity from January 2025 through Election Day. Then we compared that to how far each governor's race moved from its own average.

The more data centers in a state, the bluer the wave. Effects are consistent after controlling for partisan lean, population density, and incumbent governor.

And these states weren't already trending together: data center activity didn't predict how states shifted between the 2020 and 2024 presidential elections.

But the effect is only at the state level, not any more local

Yet voters living next to data centers didn’t behave any differently from those living further away. In any given state, voters’ proximity to a data center had no measurable effect. Data center activity in your state feels local, regardless of how close your home actually is to a facility.

Within the same state, the most active districts swung only about 1 point more toward Democrats than the least active, too small to tell apart from noise

Instead, the effect showed up at the state level, in the races that decide data center policy: tax breaks, grid connections, and who pays for new transmission lines.

The key states: Texas, Pennsylvania, Georgia, and Ohio

You can see this most clearly in four key gubernatorial races for data centers. In all four of these, Democrats ran far ahead of the national average, no matter which candidate was tougher on data centers.

For comparison, the average governor's race swung about 9 points. In low-activity states like Massachusetts and New Hampshire, Democrats gained little or nothing.

Texas: both candidates ran against data centers

Texas has been building data centers faster than any other state, as we wrote in July. It's also been paying for them, with more than $1 billion a year in tax breaks.

Gina Hinojosa ran on a statewide moratorium, saying Gov. Greg Abbott had made Texas "the wild west of data centers." But Abbott had seen which way the wind was blowing. In August, he halted approvals for new data center grid connections, pending an ERCOT audit. He also called for data centers to bring their own power and pay for infrastructure.

The timing raised eyebrows. Abbott announced the pause three months before the election, and it only lasts until ERCOT finishes its audit, with a goal of this December. Several developers told us they read it as a campaign move, not a real change in policy. Either way, voters didn't give him much credit for it.

Abbott tried to outrun his own data center record, but it didn’t work. Come January, Gov. Hinojosa will have the opportunity to implement her pause, halting construction in the biggest data center state in the country.

Pennsylvania and Iowa: where the Republican was the tougher one

If voters were punishing Republicans for being "on the side of data centers," Pennsylvania and Iowa should have been the exceptions. In both states, the Republican was more anti-data center than the Democrat.

In Pennsylvania, Stacy Garrity called for a pause on all new data center development. Gov. Josh Shapiro stopped short of a moratorium, instead issuing an executive order requiring new projects to meet power and water limits and win local approval. Shapiro won by 22 points, one of the biggest swings of the night.

Iowa is a smaller data center market, but it tells the same story. Republican Zach Lahn backed a moratorium, an end to tax incentives, and floated taxing data centers five times over. Rob Sand wanted restrictions, not a pause, and won by 11 in a state Trump carried by 13.

Georgia: a pro-data center record didn’t hurt

If any race fit the "pro-data center Republican" story, it was Georgia. Democrat Keisha Lance Bottoms called for a statewide pause on new data center construction. Republican Rick Jackson opposed a moratorium, and he's an investor in a Texas data center project himself.

Jackson still lost, but by less than you'd expect. Georgia's governor's race swung 11 points toward Democrats, about average and well below other key states, like Texas or Pennsylvania. On the same ballot, Senator Jon Ossoff won his Senate race handily, by 18 points.

The Republican with the closest ties to the industry did better than the Republicans who promised pauses.

Ohio: the middle ground didn't help either

Ohio voters have seen the backlash up close. Wood County's fight with Meta started quietly: the $750 million project barely made the local news when it was announced in 2023. By this February, nearly 200 residents were packing a county meeting to stop it. It's one of dozens of fights across the state.

Democrat Amy Acton responded by calling for a "conditional moratorium." Vivek Ramaswamy said Ohio needs more data centers, but promised a day-one executive order halting approvals until the legislature passed his conditions. In the end, it didn’t help Ramaswamy; Ohio swung 17 points toward Democrats, right in line with other high-activity states.

What it means for developers

So the headlines were half right.

Voters didn't reward candidates for saying the right thing about data centers. Republicans who promised pauses lost by as much as Republicans who promised more construction. But in the states where the buildout was most intense, voters swung hardest toward Democrats.

Voters are thinking of data centers at the state level. When a facility crops up in their state, they worry about their electric bills, regardless of whether it’s in their community. And they responded by voting for Democrats, regardless of which data center policies those Democrats actually promised to deliver.

For developers, that means two things:

1. State politics is a major site-selection variable. The incoming governors in Texas, Georgia, and Ohio, plus Pennsylvania's re-elected Josh Shapiro, all ran on tighter rules. These states will be ground zero in 2027 for incentive fights, grid-connection reviews, and ratepayer protections.

2. Candidate rhetoric is a weak signal. A pro-business candidate promising a pause and a Democrat promising restrictions may end up in the same place. Pay more attention to what governors actually do in office than to what they said on the trail.

If you want to know where data center policy is shifting in your state or county, get in touch.

Methodology

This analysis draws on Spark's internal project-tracking database: data center projects with any permitting, construction, opposition, or legal event dated January 1, 2025 through November 3, 2026, assigned to counties, states, and 119th Congress districts (counties split across districts are weighted by land area). Activity is measured as projects per 100,000 2024 presidential voters.

Election results are unofficial counts as of November 4, 2026, limited to races with more than 50% of the expected vote reported. Each race's swing is its 2026 Republican-minus-Democratic margin compared to its presidential baseline: the average 2016–2024 presidential margin for statewide races, and 2020–2024 margins on 2026 district lines for House races.

Candidate positions were researched from campaign sites, news coverage, debates, and voting records, and coded as favoring a ban or moratorium, restrictions, or more data centers. The position analysis covers 92 races where both major-party candidates took a public position (excluding Montana's Senate race, where an independent split the Democratic vote). The governor analysis covers 32 of those races; tier averages exclude Vermont, where outlier Republican Phil Scott ran far ahead of his party. The all-races analysis covers 265 contested House and Senate races, excluding House races in states that redrew their districts for 2026.

Effects were estimated with OLS regressions controlling for presidential baseline, population density, race type, and incumbent party, with standard errors clustered by state where applicable. Within-state comparisons use state fixed effects. A placebo test found that data center activity did not predict the 2020-to-2024 presidential shift. These results are correlational: data center activity isn't randomly assigned, and other differences between states may contribute.