On Thursday, November 5, two sessions after Election Day, the AI trade had its worst day in years. The Nasdaq 100 fell more than 6% and erased $2.3 trillion in market value, according to Bloomberg, while the Philadelphia Semiconductor Index dropped 12%. The Dow lost 1,833.06 points, or 3.44%, and the S&P 500 fell 4.36%. The trigger was not an earnings miss but a ballot, and the question it raised is who pays for the power.
Thursday, by the numbers
The breadth was the story. All 30 Dow components fell, Bloomberg reported; the Nasdaq 100 closed nearly 15% below its October 6 record, the Russell 2000 slid about 4.5% into correction, and the VIX jumped 11.57 points to 28.18. The chip index now sits more than 38% under its June 22 peak, a bear market by any definition. Bloomberg cited forced selling by multi-manager hedge funds and a buyer's strike among long-only funds; The Wealth Alliance's Eric Diton summed it up: "This selloff is extreme."
Fox Business had the Nasdaq Composite down 5.52%, into correction, with Nvidia off 9.26% and IREN, Applied Digital and Nebius each down more than 14%. TheStreet's market wrap puts the Nasdaq 100's drop at 6.3% and says the S&P 500 technology sector fell more than 7%. On history the sources differ: TheStreet says worst since April 2025 for both indexes, Bloomberg says the Nasdaq 100's worst since March 2020.
It started Wednesday, and it started with power
The first reaction to the results was a rotation, not a rout. On Wednesday, Reuters reported Amazon and Microsoft up 6% each and Alphabet up 4.6%, while data-center operators sank (CoreWeave down 9.5%, Applied Digital 8.8%, Cipher Digital 21%) and the power suppliers behind them fell harder: Vistra 15.4%, GE Vernova 11.5%, Constellation Energy 9.2%. B. Riley Wealth's Art Hogan told Reuters that existing data centers gain value if overbuilding is avoided; Siebert Financial's Mark Malek argued that fewer tax incentives and costlier power and water would squeeze returns, leveraged builders first.
On Thursday that distinction dissolved. eToro's Jakub Rochlitz told Fox Business that markets now see the AI infrastructure buildout as "slower and more expensive," and noted that some Democrats had called for restrictions or pauses on new projects. TheStreet also cites a Citrini Research note heralding the "end of the AI trade as we know it"; the note itself could not be reviewed for this piece.
The tape was already fragile. On October 30 the S&P 500 fell 1.8%, its worst day since June, per the AP, after Intel plunged 28.6% and Amazon disclosed $42.4 billion of AI-driven cash burn over twelve months.
What the winners actually proposed
The positions on record are specific. In Texas, Democrat Gina Hinojosa, who unseated Gov. Greg Abbott 57% to 41% in unofficial results, per Scripps News, proposed a moratorium on new construction, stricter water and power-use rules and an end to industry tax breaks. Ohio governor-elect Amy Acton backs a conditional moratorium: new facilities would cover all their utility costs, use union labor and build on old industrial sites rather than farmland. Her framing: "Ohio is open for business, but we are not up for sale." Georgia's Keisha Lance Bottoms wants a statewide pause to study effects on bills and water. In Ohio's Senate race, Sherrod Brown, who won, said data centers should pay their full utility costs and accused Jon Husted of "sweetheart tax breaks" for tech giants; Iowa governor-elect Rob Sand pledged to end tax breaks and block nondisclosure agreements.
The case for the buildout was made just as plainly. Husted had backed data-center development and argued its economic benefits outweigh the tax breaks; Abbott had pitched Texas as the epicenter of AI development, though in September he paused state permits pending a grid audit, which Hinojosa dismissed as an election-year pause. The map did not split cleanly by party either. In Wisconsin, Democrat David Crowley won while calling the state a potential global AI hub and opposing automatic rollbacks of incentives. In Florida, Republican Byron Donalds beat Democrat David Jolly, who had promised a day-one moratorium on hyperscale data centers; Donalds argued for local control and said companies that fail to protect ratepayers and water supplies would not build.
The backdrop is not in dispute. US electricity prices are up roughly 10% since early 2024, Texas holds about 27% of planned US data-center capacity by Wells Fargo's estimate, and Ohio's data-center incentives total $1.6 billion, per Scripps. TheStreet described the Democratic winners as anti-AI candidates; the proposals themselves, as reported, concern permitting, utility costs, water and tax incentives.
Friday: a jobs beat and a partial bounce
At 8:30 a.m. on November 6, the Bureau of Labor Statistics reported October payrolls up 165,000 with unemployment at 4.1%. August and September were revised up by a combined 84,000, and average hourly earnings rose 0.4% to $38.00, up 3.1% on the year. The Dow Jones consensus was 59,000, per TheStreet. For this readership: the information sector was flat (about 1,000 jobs lost), while professional, scientific and technical services added 33,000.
Stocks took the number and little else. Late in the session, TheStreet had the Nasdaq up 1.09%, the S&P 500 up 0.31% and the Dow down 0.46%, with the VIX down 14% to 24.19. Chips led: the iShares Semiconductor ETF was up 3.27% in the afternoon and AMD up 10.77% in the morning. The Composite recovered less than a fifth of Thursday's drop.
The 30,000-foot read
Thursday was not a verdict on AI. It was a repricing of the physical layer beneath it. If new capacity must pay full freight for power and water, wait on grid audits and lose tax breaks, the buildout gets slower and dearer, and the valuations that assumed otherwise get marked down first. Wednesday's split, owners up and builders down, was the more precise reading; Thursday was the indiscriminate one, and Friday a breath rather than a change of mind.
For people who build software, the second-order effect matters more than the index. Cheap marginal compute has been underwritten by a capex wave that assumed land, power and permits would keep arriving on schedule. If capacity costs more at the margin, it shows up eventually in inference prices, GPU availability and the hiring budgets of companies whose roadmap is one long API call. What you control is how much of your product's unit economics assumes compute only gets cheaper. What to watch: hyperscaler capex guidance next quarter, the first permitting moves of the new governors and state utility regulators, and whether the chip index finds a floor. One good jobs report answers none of that.
Sources: Bloomberg via Honolulu Star-Advertiser (Nasdaq 100 decline and $2.3 trillion, March 2020 comparison, chip index and bear market, Russell 2000, VIX close, all 30 Dow components, fund selling, Diton quote); Fox Business (Nov 5 Dow, S&P 500 and Composite closes, Nvidia and AI infrastructure stocks, Rochlitz quote); TheStreet (Nasdaq 100 6.3%, tech sector, April 2025 comparison, Citrini note, anti-AI characterization, Nov 6 index, ETF, AMD and VIX moves, payrolls consensus); Reuters via The Star (Nov 4 rotation, Hogan and Malek); AP via Action News Jax (Oct 30 sell-off, Intel, Amazon cash burn); Scripps News (Texas, Ohio and Georgia governor results and proposals, Abbott permit pause, Acton quote, electricity prices, Texas capacity share, Ohio incentives); NewsNation (Brown, Husted, Sand, Crowley, Donalds and Jolly positions and results); BLS Employment Situation, October 2026 (payrolls, unemployment, revisions, earnings, sector data). CAVEATS: The Citrini Research note is single-sourced via TheStreet and was not reviewed directly. Nov 6 index moves are TheStreet's late-session figures (3:38 p.m. ET update), consistent with the quote tables on the Fox Business page but not checked against official exchange closes; the ETF and AMD figures are intraday. The 6.3% Nasdaq 100 figure and the tech-sector decline are TheStreet's alone (Bloomberg says more than 6%); S&P 500 decline is 4.36% per Fox Business, rounded to 4.4% by TheStreet. Worst-day comparisons differ between Bloomberg and TheStreet, as noted. The Texas result is unofficial per Scripps. Reuters' Nov 5 roundup of data-center restrictions could not be opened and is not cited. Wire copy from Bloomberg, Reuters and AP was read via syndication partners. Election Day was Tuesday, Nov 3; the rout came two sessions later.