The 2026 US midterms were the first American election in which artificial intelligence was a ballot-box issue rather than a sidebar — and the verdict was blunt. The industry's money did not buy the races it was spent on, and its most visible physical footprint, the data center, became the most effective attack line of the cycle.
Democrats won control of Congress, according to TIME. An NBC exit poll found that 56 percent of voters called AI and data centers a "crucial or major factor" in their vote for Congress. Democratic campaigns spent more than $100 million on broadcast ads casting Republican opponents as too close to Big Tech and data-center developers, and won in the states where that message landed hardest: Ohio, Kansas and Iowa.
The backlash is local and specific. Counties from Edgewater, Florida, to Bloomington, Illinois, passed data-center bans or moratoriums. The mayors of Piqua, Ohio, and Yukon, Oklahoma, were recalled over earlier support for data-center projects. Blue Rose Research's comparative polling found Americans would rather live near a prison, a methadone clinic or a nuclear power plant than an AI data center. AI still ranks only 29th of 39 issues in the firm's tracking — but its salience rose faster over the past year than any other issue, including war in the Middle East.
South Carolina's Senate race became the exemplar. Darline Graham, backed by two AI industry PACs to the tune of $4.8 million, lost to Annie Andrews, who nicknamed her "Data Center Darline" and campaigned for a two-year moratorium on new projects. In Ohio, Senator-elect Sherrod Brown ran ads calling Jon Husted "the face of data centers in Ohio" and won by five points. The National Republican Senatorial Committee had already conceded the point in an August memo: "more than any other thing in this race, data centers are the anchor hanging around Husted's neck."
On the spending side, the money itself became the liability. Leading the Future, launched in August 2025 with more than $100 million behind it, spent $27.38 million in primaries — including $8 million opposing New York Assembly member Alex Bores — then largely sat out the general election. One survey cited by TIME found three-quarters of likely voters said they would be less likely to support a candidate who had taken money from an AI super PAC; donors including OpenAI president Greg Brockman and investor Ron Conway reportedly pulled back from promised contributions.
That gap let the other side spend. Anthropic put more than $40 million into Public First Action, money restricted from electoral activity, while Anthropic employees separately donated about $5 million to campaign efforts. Guardrails Alliance, which exists to make AI money politically toxic, spent more than $1 million apiece opposing two candidates who both lost. Its political director, Charlie Blaettler, described the strategy without ornament: "Our goal is to make AI money politically toxic."
The aggregate numbers are brutal for the industry. According to Citation Needed's tally, crypto and AI interests together spent roughly $250 million this cycle; of the twelve competitive races called so far, the side they backed lost nine, with $74.7 million spent on the losing side against $28.3 million on the winning side. In the 40 safe races where they spent, their picks went 40 for 40 — which is rather the point. Backing near-certain winners inflates a win rate without buying influence in the contests that decide control.
Fear of AI-generated disinformation also failed to become a decisive factor. Deepfaked videos of Maine's Troy Jackson circulated in the final week; he won by more than seven points. In Wisconsin, Republican Derrick Van Orden posted deepfakes of his opponent saying things she never said; he lost.
What comes next is a Congress that can subpoena more easily than it can legislate. Sweeping federal AI law would need President Trump's signature or a veto-proof majority, and neither looks likely. The more consequential fight is over state law: Republicans had sought to preempt state AI rules, and that effort is now harder to pass. "If Congress is gridlocked, what happens in California will govern AI for the U.S. and the world," Scott Singer of the Carnegie Endowment told TIME, referring to SB-53, the first US law governing frontier systems.
The industry is already repositioning. OpenAI plans a policy and public-relations offensive that includes rebranding data centers as "AI compute campuses," Axios reported, while Meta is reportedly assembling a bipartisan Washington team.
Blue Rose's Ali Mortell offered the line the industry should probably frame: "This was the first cycle where AI showed up in the political discourse, and it's arguably the least important that it will ever be again."
