What's driving this is substitution more than sentiment. In Weigel's account, data centers have become the thing voters point at when they mean their power bill [8]. That is a harder position to argue out of than ordinary siting opposition, because the industry is being judged on a line item it does not set, cannot itemise for a household, and has no way to promise will fall.
The arithmetic of the figure quoted in the interview is worth stating plainly. If two-thirds to three-quarters of Americans say they do not want a data center near them [1], the pool of residents who are supportive or still undecided tops out somewhere between a quarter and a third [13]. A community-relations budget can work a persuadable majority, but it has no way to conjure one where none exists.
Sequencing is the part that should interest anyone with a project in development. The governor-level reversals in Texas and Pennsylvania arrived after the county races and the commission races, not before them [5][3][6], and Weigel's read is that the Georgia Public Service Commission contests supplied the impetus for Republicans who were more exposed on pro-data-center policy to start dialing back [7]. Political positions can reverse inside a single election cycle [2], while interconnection queues, transmission upgrades and twenty-year power contracts move on a much slower clock, so the asymmetry sits with the party holding the capital commitment.
The evidence behind this is thinner than a data set: it is one reporter's impression from the road, published as an edited podcast excerpt [12], and there is no named survey behind the two-thirds-to-three-quarters range, nor any count of projects actually cancelled or delayed [14]. That is a fair gap, and the honest answer is that we do not yet know the project-level cost. What is checkable is that two named executives who had backed these builds no longer do [5], and the error is expensive in only one direction: a site selection can be paused, but a substation, once built, stays built.
The causation dispute inside the source changes what a leadership team would actually spend money on. OpenAI has said it found Chinese users creating meme posts, and X said the same the week before the interview [10]; Weigel says he is sure outside actors are in the conversation but that people were not getting angry because of Chinese bots, and he has not seen them [11]. He attributes the anger to broad unease about technology, sharpened by AI's own advocates explaining which jobs will go [9]. If the foreign-influence reading is right, the remedy is platform enforcement. If Weigel is right, the remedy is rate design, cost allocation and who visibly pays for new generation. A company that funds the first while the second is the problem will keep losing hearings it thought it had won.
The board-deck version treats opposition as local, manageable, and already priced into the schedule as a contingency. It is incomplete: the movement first showed up in candidates of both parties in Loudoun County running on electricity and bills [4], and it has since travelled to statewide races and utility commissions. This quarter, the decision is narrow, which jurisdictions to file in and what the ratepayer story looks like when a commissioner is up for election. This decade, the question is whether the capex plans of the AI buildout assumed a permitting environment that voters have begun to reprice, and on that the record does not yet say.