The AI industry spent the past year assuming Washington would leave it alone. The US midterm elections, in which Democrats took control of both the House and the Senate, have upended that assumption. According to The Verge's reporting, holding AI companies accountable was a recurring theme across many of the winning campaigns, and the first real effects are likely to land on data centres rather than on models.
The ballot results were blunt. Of 44 local data centre moratoria on the ballot, 43 passed. In Texas, the fastest-growing area for data centre construction, governor-elect Gina Hinojosa has said she will halt all data centre approvals until the legislature passes laws to protect residents, and similar promises have been made in several other key states. An NBC News poll of voters on election day found 76% opposed data centres in their own communities, with noise and rising electricity bills among the complaints. Campaigners on both sides describe the data centre backlash as unusually bipartisan.
The industry's response has been quiet in public and nervous in private. OpenAI and Anthropic did not comment to The Verge. The Wall Street Journal reported that chief executives have discussed significant compute constraints after the results, particularly at OpenAI, and Meta's Dina Powell McCormick said many companies are considering building data centres abroad. Elon Musk, who had spent close to $450 million backing Republican campaigns, deleted his call to vote red. Many incoming leaders have promised to call AI executives to testify.
Policy experts quoted in the piece expect the federal push to block state AI laws to be finished, and more state-level accountability and safety bills to follow. Senator Mark Warner, re-elected after campaigning on AI safety, and a bill introduced by Senator Bernie Sanders and Representative Greg Casar to ban superintelligence point to the direction of travel. Analysts at Yorkville Ives & Co. warned investors of major hurdles for tech legislation and a slowdown in plans for up to 1,000 US data centres, while noting this might help avoid a bubble.
For people using the tools we review, the near-term effect is indirect but real: compute is what limits how much a provider can serve, and slower or pricier US data centre construction tends to show up as tighter usage caps and higher prices. The caveat is that campaign promises are not laws. New legislatures still have to agree on details, courts may intervene, and companies will lobby hard; how much of this mood becomes binding rules is the question for 2027.