Voters in several states this week approved ballot measures meant to slow the construction of electricity-hungry data centers. That detail, noted by CNBC in a piece about renewable energy, got little attention amid the larger story of a Republican rout, though it deserves more. Nobody has produced evidence that Americans cast ballots over chatbots or model safety. In some places they did vote against the buildings that house the servers, which is a different kind of AI politics and may prove the more durable one.
Three lessons follow from the results, none of them certain, and together they sketch the likely shape of AI policy over the next two years.
The electricity bill is the new entry point
For most of the past three years, the public argument about artificial intelligence has been conducted in the vocabulary of experts: existential risk, copyright, alignment, jobs. The midterms suggest voters care about something more tangible. According to CNBC, outrage over high energy prices and over data centers ran through the campaign, strong enough to produce multiple state-level initiatives restricting their growth.
The logic isn't hard to follow. A household watching its power bill climb does not need a briefing on transformer architecture to conclude that a new campus down the road is part of the reason. Justina Lai, chief impact officer at the wealth advisory firm LNW, told CNBC that national electricity demand is rising partly because of data centers, while the grid is aging and constrained. Whether the bills are actually rising because of those facilities is something analysts argue about, and the evidence differs by region. Politics rarely waits for the argument to settle.
This matters for AI policy because it moves the fight from Washington's committee rooms to county zoning boards and state capitals, where officials control land use, water permits and utility rate cases. If voters have decided those are the levers worth pulling, the industry's most important regulators over the next two years may not be federal at all.
Gridlock with a veto pen
The second lesson is arithmetic. Democrats now control both chambers, according to CNBC's account, and President Trump retains the veto. Overriding him takes two-thirds of each chamber, a bar that, as the BBC observed in a different context, a Democratic Congress cannot clear on its own. The same report noted that a Democratic majority could still slow or stall things it dislikes without reaching that threshold, through its control of funding, hearings and the legislative calendar.
Applied to AI, the picture is mostly one of friction. Sweeping federal AI legislation, whether a licensing regime or a liability framework, is unlikely to reach the president's desk in a form he would sign. A Democratic majority can still hold hearings, subpoena documents, attach conditions to spending bills and demand answers from the companies and agencies involved, and it should be expected to. That is less dramatic than a new statute, but companies tend to dread it more than they admit.
There is a wrinkle on the other side. CNBC reported that some Republicans may be more willing to side with Democrats against the president on energy costs, precisely because their own voters are angry. A veto-proof coalition on a narrow, pocketbook-flavored AI measure, say one forcing data centers to bear more of the cost of the grid upgrades they require, is not fanciful. A broad one remains very unlikely.
A bipartisan permitting-reform bill is waiting on Senate approval. According to CNBC, a number of progressive Democrats would prefer to postpone a vote until the incoming Congress is seated in January, giving them a chance to revise language they see as tilted toward fossil fuel projects at the expense of renewables. How that bill ends up will largely determine how quickly new generation and transmission can be built to feed the data center boom. It is an AI bill in everything but name.
A crowded agenda and an AI policy afterthought
The third lesson concerns attention, which is scarce. In the BBC's reporting, Mr Trump attributed the Republican defeat partly to the war with Iran, while Vice-President JD Vance named the American relationship with Israel as one of four reasons voters abandoned the party. Foreign policy, war costs and energy prices dominated the postmortem, and artificial intelligence barely appears in it.
That should temper expectations on both sides of the debate. Safety advocates hoping that a new majority will make AI oversight a first-tier priority are likely to be disappointed, at least at first, since a party that has just won on affordability and war fatigue will reasonably spend its early months there. Industry figures hoping the issue will fade are misreading where the pressure comes from. It is arriving through utilities, land use and household budgets rather than through a congressional AI caucus.
The administration's own posture will matter as well. Executive orders on energy policy can continue regardless of what Congress does, as CNBC pointed out, and there is no obvious reason the White House would treat AI differently. Whether it uses that power to speed data center construction or to answer voter anger is unknown. The two impulses pull in opposite directions, and the president has not said which wins.
What to watch from here
Barclays analysts, led by Peter Crampton, wrote that the results reduce the risk of political interference with long-term clean energy economics, and clean energy stocks rallied the day after the vote. A similar test is coming for AI companies, though nobody can say yet how the market will price it. If investors begin treating local opposition to data centers as a real constraint on growth rather than a nuisance, capital spending plans will shift, and the shift will show up in earnings calls before it shows up in any bill.
The near-term markers are modest and concrete. The Senate calendar will show whether the permitting vote slips into January. The committee staffing choices will show where the new majority intends to press hardest. And it will become clear soon enough whether the state data center measures spread to places that did not vote on them this year. AI policy in this Congress will probably be written less by lawmakers debating the technology than by constituents objecting to what it costs to run. Those who wanted the debate to be about the technology itself may have to wait their turn.