The 2026 US midterms delivered a clear verdict on the physical footprint of the AI boom: dozens of ballot measures limiting datacentre expansion passed across the country, and candidates who ran against unchecked tech infrastructure won. The cost-of-living crisis driven by Trump's war in Iran was the primary driver of the Democrat wave, but opposition to AI's material imposition on communities — noise, light pollution, energy consumption, water use — was a distinct and potent secondary force. The market had already been wobbling. The Nasdaq peaked a month before election day, and the IMF had flagged "growing evidence" that successful AI rollout "may not require investment on the massive scale being undertaken by 'hyperscaler' firms." The midterm results appear to have catalysed a broader repricing. Capital Economics analyst Joe Maher warned that the election "could plausibly become the catalyst that leads to a more long-lasting unwind of the AI equity boom," given that sentiment was already fragile and the market was showing late-cycle signs. The structural mismatch at the heart of the AI trade is stark. TS Lombard economist Dario Perkins described the industry in late 2026 as "$200bn in current revenues supporting more than $1tn in future commitments from just two companies" — Anthropic and OpenAI. Corporate users have already begun pulling back: firms are slapping limits on staff use of costly products like Claude, and some — including Revolut — are turning to much cheaper self-hosted models to cut coding costs tenfold. Texas became the sharpest flashpoint. Elon Musk poured billions into trying to buy the electorate there, and senator-elect James Talarico made datacentre opposition a centrepiece of his campaign. The anecdote he deployed — a mother whose son thought it was daytime because of neighbouring datacentre lights — is the kind of concrete, visceral complaint that no amount of productivity-gains rhetoric can neutralise. The IMF added another uncomfortable data point: lower-cost Chinese AI models are demonstrating growing capabilities, and circular financing within the AI sector has become rife. The premise that only American hyperscalers building at massive scale could deliver AI's benefits is being challenged from below and from abroad simultaneously. Two things are true and in tension: AI is a real technology that will yield productivity gains, and the current investment thesis — which requires carpeting the US in debt-funded datacentres on the assumption that demand will materialise — is not supported by the evidence. The article's framing, via Perkins, is precise: people outside the tech-finance ecosystem "just weren't that interested in AI" in ways that justified trillion-dollar bets. The question now is whether this is a speed bump or a structural turn. The Teddy Roosevelt analogy — trust-busting the oil and railway monopolies — is aspirational, but the immediate effect is concrete: local opposition will slow permitting, raise costs, and force the AI lobby to demonstrate public value rather than assume it. The gauntlet has been thrown.
