Data-centre politics: moratoriums win TX/GA/OH as Silicon Valley's $600mn midterm bet fails
What's the deal? Democratic candidates who campaigned on curbing data-centre development won the 3 November 2026 governor races in Texas, Georgia and Ohio, three hubs of the AI build-out, the Financial Times reports. In Texas, Gina Hinojosa defeated incumbent Greg Abbott after promising an immediate moratorium pending legal safeguards (Abbott had already directed the TCEQ in September to pause grid-connection permits pending an audit). In Georgia, former Atlanta mayor Keisha Lance Bottoms pledged a temporary statewide pause; Atlanta has about 2.9GW under construction, according to CBRE via the FT, overtaking Northern Virginia. In Ohio, Amy Acton beat Vivek Ramaswamy on a conditional freeze of new approvals unless strict energy and water criteria are met.
Local votes went the same way. Trenton, Ohio passed a measure that would stymie an Amazon campus, several more Ohio measures were on track to pass, and Edgewater, Florida approved a charter ban 77% to 23%, the FT reported. Before the vote, Amazon's planned $2 billion Sunbury, Ohio campus was already one of 41 data-centre ballot measures across 10 states (28 restrictive, 13 easing; Ohio alone had 21), with a local ban covering facilities above 25MW and a city moratorium running through January 2027, per reporting summarised via Ballotpedia and CNBC. In all, 26 gubernatorial candidates in 21 states backed some form of moratorium; an NBC exit poll found 56% said AI and data centres were a crucial or major factor in their congressional vote. Fortune and RFI reporting the same day corroborates the three governor wins and the Texas capacity stakes.
The ads that set the stage. Campaigns and outside groups spent $173 million on data-centre ads in the general election, almost all negative, according to AdImpact data reported by Axios before the vote (Democrats about $140M, Republicans $28M, independents $5M). A Pew Research Center analysis of Facebook and Instagram political ads found the tone flipped from mostly supportive in January to 57% critical and 3% supportive by October. Counter-readings from Infrastructure for America and the Blue Horizon Project argued data centres were a smaller share of spots and that many ads attacked tax breaks and power bills rather than the facilities themselves.
The money that didn't work. Silicon Valley donors put more than $600mn behind Republican candidates and still lost large majorities in both houses of Congress, the Financial Times reports (5 November). Elon Musk alone spent more than $470mn on House and Senate races, nearly half of it for Ken Paxton, who lost to Democrat James Talarico in Texas; an operative told the FT Musk "spent half a billion dollars and lost every race he showed up in", and Musk later deleted an X post urging "Vote red today!". Marc Andreessen and Ben Horowitz spent more than $90mn, including $50mn to Leading the Future, the main pro-AI PAC; Greg and Anna Brockman gave $25mn to Maga Inc and $25mn to Leading the Future; the Winklevoss twins (Gemini) spent more than $5mn. Fairshake, the crypto super PAC, spent more than $50mn on ads against Sherrod Brown, who still returned to the Senate, and Leading the Future's $2.5mn behind Republican Darline Graham in South Carolina failed in an upset that gave Democrats a Senate seat there for the first time in decades. An NBC exit poll found a majority of voters cited AI data centres as a crucial or major factor (alongside inflation and the Iran war); dozens of data-centre moratoriums passed. Democrats such as Greg Casar want AI bosses forced to testify under oath next year. Caveats on this scorecard: the cash versus independent-expenditure split is not broken out; key failure quotes come from an anonymous operative and an advocacy group; and the spend was still a tiny fraction of the donors' wealth.
Market reaction. After the results, AI data-centre suppliers sold off while large clouds gained: Credo fell about 10%, Coherent about 7% and Dell about 9%, while Amazon rose about 6–7% and Microsoft about 6.7%, per market reporting. Nebius also fell about 6% even as broader indexes rose. Dan Ives of Yorkville Ives & Co. called Texas the "biggest worry" of the night, citing roughly 25–30% of planned US data-centre capacity. Goldman Sachs chief economist Jan Hatzius said Democratic wins at state and local level are likely to bring short-term moratoria that could affect about 45% of planned US capacity expansion. Bank of America analysts argued that regulatory roadblocks could lift utilisation and support pricing if paid-compute demand stays strong — AWS revenue up 39%, Azure 48%, Google Cloud 103% in the periods they cited — with a trade-off of lower peak earnings but a longer profit cycle.
Who benefits — competing views. Some investors read a slower US build-out as good for hyperscalers that already have capacity and rent it out, and harder for developers and neoclouds still needing new permits and power. That is a market thesis, not a settled outcome: neoclouds such as CoreWeave, Nebius and Nscale, and operators still raising sites, remain exposed to permitting and credit risk, while Oracle, Crusoe and others with projects in contested states face the same political and grid constraints. Elon Musk, after Democratic wins in Texas, posted that "Data centers in space looking pretty appealing right now" (@elonmusk) — his view, not a forecast. The counter-case is that delayed capacity could also slow AI demand growth and cut orders for the very suppliers and clouds that finance the build-out.
Caveats. Governors' powers depend on state law and some will need their legislatures. The pledges differ in strength (immediate moratorium, temporary pause, conditional freeze). Several Ohio measures were only on track when the FT went to press; ballot tallies and hyperscaler responses were still forming. Bloom Energy has projected Texas near 30% of US AI electricity demand by 2028, per secondary reporting.
Why it matters. Amazon, Meta, Microsoft and Alphabet plan up to $740bn of capex in 2026, and Nvidia has forecast $3tn to $4tn of global AI infrastructure spend by 2030, per the FT. Permitting and power access, not just chips and capital, now carry political risk for hyperscalers, neoclouds and every startup whose plan assumes fast new capacity in the US — and the same midterm that produced the moratoriums also wiped out a $600mn+ Silicon Valley PAC bet that cash could blunt the backlash. New York already paused 50MW+ data centres for a year in July.
