How the Midterm Results Change AI

The midterms produced a blue wave, with Democrats set to control both chambers of Congress but short of a veto-proof Senate majority. Democrats gained 11 points in national averages, though their seat gains were smaller than the historic midterm losses for the party in power in 2010 and 2018. The election was less about AI in general than about data centers and their effects on electricity costs, local control, inequality, surveillance, and the influence of wealthy technology figures.

Voters ranked AI well below the economy: a Pew poll found 44% considered AI very important, compared with 84% for the economy, while 61% disliked or strongly disliked candidates who supported local data centers. NBC exit polling found 56% considered data centers a crucial or major voting factor, and Democrats were twice as likely as Republicans to cite them as a major factor. Data centers still ranked behind cost of living, government corruption, the war in Iran, and immigration enforcement.

Sherrod Brown made data centers central to his Ohio campaign against John Husted, arguing that residents should not subsidize their costs and should decide whether projects are built locally. Brown reclaimed his Senate seat after losing it in 2024; the episode describes him as a consistent big-tech critic and says crypto-PAC spending contributed to his earlier defeat. Ohio had around 200 data centers, mostly small, and 46 planned projects totaling 16 gigawatts, including Meta’s Prometheus facility outside Columbus and OpenAI’s planned 10-gigawatt Pike County campus. In Texas, Gina Hinajosa defeated Greg Abbott with a platform calling for a data-center moratorium until regulations were written, an end to tax subsidies, and protection of local access to electricity and water. Abbott had halted new permits in September pending an audit, but the episode characterizes the move as too late to prevent his defeat. Texas had more than 470 gigawatts of proposed data-center capacity, about three times the size of its current grid and new capacity.

John Ossoff used data centers to attack big-tech influence, political corruption, and rising energy costs, arguing that Georgia families should not pay more to subsidize data-center electricity use. He said Trump’s “big beautiful bill” gave developers $70 billion and linked that spending to rising healthcare costs and hospital closures; in April, he launched an investigation into data-center electricity use. The episode suggests his limited AI-policy activity could change, given the issue’s prominence in his campaign and discussion of a possible 2028 presidential run. Abdul El-Said linked data centers, Flock cameras, and AI-powered warfare as symbols of state oppression, citing job losses, economic disruption, loss of control, and bioweapon risks among public concerns; his future policy positions remained uncertain. In California’s 11th District, Scott Wiener had a commanding lead with about half the votes tallied, but the race had not been called at recording time. He had made AI regulation a campaign theme after sponsoring SB 1047 in 2024, which Governor Gavin Newsom vetoed.

The results reinforce that data-center developers can no longer ignore local communities, a shift already underway. Developers are increasingly using fewer NDAs, committing to pay for infrastructure and electricity, and offering community incentives. Governors have the final say on state data-center policy, and Democrats governed nine of the 10 largest states after the election. A comprehensive federal AI framework or a top-down moratorium on data centers or frontier-model training was considered unlikely. More plausible incremental measures include whistleblower protections for AI-lab employees, requirements codifying voluntary frontier-model testing, incident reporting, stronger model-builder liability, and possibly data-sovereignty rules; the direction of export controls remained hard to predict.

The episode characterizes the results as an “AI earthquake” that showed voters want rules for AI and data centers, while also saying the election largely confirmed a direction already underway. AI executives could face many congressional hearings, though the host expected limited impact. Debate may cover the pace of AI development and whether to prohibit techniques such as recursive self-improvement, with common-sense incremental rules potentially providing bipartisan ground. The main risk raised was a wave of data-center moratoria that could make Wall Street withdraw funding: without financing, AI buildout could halt, neoclouds could fail, and a broader downturn could follow. That outcome could shape the 2028 debate over AI, jobs, cybersecurity, risk, and a potential bubble, although the speaker saw substantial room for political alignment unless opposing AI development becomes a winning issue.