The morning after America’s midterm elections, the financial markets delivered their own verdict. The Nasdaq composite fell more than 5.5% on Thursday, erasing an estimated $2.5 trillion in market capitalisation in a single trading session — its worst day since 2025. Nvidia shed more than 9% of its value, wiping over $400 billion off the firm alone. AMD and Micron each fell by more than 17%. SpaceX and Tesla dropped 8% and 11% respectively. The so-called Magnificent Seven tech firms combined lost more than a trillion dollars in a day. The proximate trigger: Democratic wins in races where candidates had campaigned explicitly on slowing artificial intelligence development, with dozens of ballot measures targeting data-centre construction across multiple states. Texas governor-elect Gina Hinojosa, whose campaign centred on a moratorium on new data-centre construction, flipped the state blue for the first time since 1976.
The received wisdom
The mainstream reading of Thursday’s crash is sympathetic to the Democrats and sceptical of the market. The argument runs like this: AI development has proceeded at a breakneck, arguably reckless pace, with minimal public consultation and enormous environmental footprints. Communities in Texas, Georgia and elsewhere have watched data centres consume water, strain power grids and drive up energy costs while delivering few local jobs. The backlash, in this reading, is a healthy democratic corrective — communities asserting sovereignty over their own infrastructure. The stock market, meanwhile, was always inflated by speculative enthusiasm. OpenAI president Greg Brockman’s admission at the WSJ Tech conference that the industry had failed to bring communities along — that you “can’t just tweet and say: ‘This is going to be good’” — underlines how badly Silicon Valley miscalculated the political mood. Meta’s Dina Powell McCormick calling for the industry to “double and triple down on telling the story” is precisely the kind of tone-deaf corporate response that confirms the progressive critique.
A different read
There is something to all of that. But the scale of Thursday’s correction, and the political coalition that produced it, deserves a more careful autopsy than the progressive narrative allows.
Start with what actually happened on Tuesday. Democrats did not run primarily on a platform of regulating AI in the abstract. They ran on cost of living, on what BBC News characterised as “food, fuel and taxes.” The anti-data-centre measures were, in many cases, local ballot initiatives riding a wave of economic anxiety rather than a coherent industrial policy. The market, in pricing them as a categorical threat to AI infrastructure, may be conflating a protest vote with a mandate.
More troubling, however, is the question of whether the protest vote is right for the wrong reasons. The AI bubble — and there are genuine grounds for thinking it is a bubble — was not primarily inflated by community goodwill. It was inflated by the assumption that capital expenditure on data centres would generate returns commensurate with the investment. Susannah Streeter of Wealth Club put it plainly: “fears that the vast sums poured into AI infrastructure may not deliver the returns investors have been banking on are now colliding with the prospect of a sharp slowdown in data centre construction.” The political shock may simply have been the pin that pricked a balloon already overinflated on its own terms.
This is the uncomfortable position for both sides. Progressives celebrating the stock market fall as a comeuppance for Silicon Valley arrogance should be wary of what they are actually celebrating: a repricing of productive capacity in the economy, a slowdown in the development of technology that — for all its genuine risks — will determine whether the West keeps pace with China’s state-directed AI programme. Beijing does not hold referendums on data-centre construction. The Philadelphia Semiconductor Index fell more than 12% on Thursday. Every percentage point of that fall is a relative gain for SMIC and Huawei.
Equally, however, the tech right — Musk’s $450 million Republican campaign bet, now looking badly misjudged — should ask why the backlash was so severe. The answer has something to do with the industry’s long failure to engage seriously with distributional questions: who pays the energy costs, who drinks the water, who gets the jobs. Musk spent a reported $45 million backing Ken Paxton for the Texas Senate seat and watched Texas go blue. That is not just an electoral miscalculation. It is evidence that the libertarian tech vision has run out of political runway in the communities where it sought to plant its infrastructure.
The deeper historical parallel is instructive. The 1970s saw a similar backlash against the petrochemical industry — not because oil was not economically valuable, but because communities felt they bore the costs while distant shareholders reaped the gains. The regulatory response (the Clean Air Act, NEPA) was messy, economically costly in the short run, and ultimately forced the industry to internalise externalities it had been exporting. Something similar may be underway with AI infrastructure. The question is whether the political class managing this correction is capable of distinguishing between legitimate regulation and simple obstruction — a distinction that, judging by Thursday’s ballot measures, it has not yet earned the right to make.
What to watch
Four signals will determine whether Thursday was a one-day correction or the beginning of a structural repricing. First, whether the new Democratic Congress moves quickly on AI-specific legislation, or lets the ballot-measure energy dissipate into committee hearings. Second, whether Nvidia’s earnings guidance in Q1 2027 shows actual capex slowdowns from hyperscalers, or merely spooked sentiment. Third, whether the Trump administration — now facing a hostile Congress — attempts to use executive authority to override state-level data-centre restrictions, and whether courts let them. Fourth, and most consequentially for the long run: whether China’s AI programme accelerates into the gap, prompting a national-security recalibration of the kind that reversed the semiconductor export controls debate in 2023.
— J