DEV Community

Cover image for Data centers lost 40 of 41 votes — America PAC won 3 of 44
frank chu
frank chu

Posted on

Data centers lost 40 of 41 votes — America PAC won 3 of 44

Data centers lost 40 of 41 votes — America PAC won 3 of 44

Voters went against data centers in 40 of 41 local ballot measures on November 3 — all 27 restrictive measures passed, and all 13 pro-development measures failed. The only place the industry won was Harwood, North Dakota, where voters rejected a measure that would have added review requirements for big industrial projects.

The ballot math

Ballotpedia tracked 41 local data-center measures across 10 states, and CNBC reported the tally. The breakdown is the story: 27 measures that limited development — bans, moratoriums, new restrictions — all passed. 13 measures that would have favored development — approving projects, easing rules — all failed. Ohio was the center of gravity with 21 of the 41, and data centers lost every one; in Piqua, a ban went into the city charter with about 67% of the vote. Five full bans passed (one each in California and Florida, three in Ohio), three six-month moratoriums passed in Maine, and Kansas approved a ten-year moratorium on facilities using more than 100,000 gallons of water a day.

An NBC exit poll found 56% of voters called AI and data centers a crucial or major factor in their vote. And the counterweight, worth stating plainly: more than 830 restrictive measures have passed this year, according to Heatmap's count, but 40 bans are 0.2% of America's 19,000 municipalities. This was a sweep of the places that voted, not a national ban — and one strategist noted the results had "basically no correlation to how red a county was."

The Texas math

Gina Hinojosa beat Greg Abbott 56.69% to 42.02% — 4,080,339 votes to 3,024,099, a margin of 14.67 points and 1,056,240 votes. She takes office at noon on January 19, 2027, the first Democrat to hold the office since Ann Richards left in 1995 and the first Latina governor of Texas. She campaigned on freezing new data centers until the state writes rules for them.

The money makes the margin stranger. Abbott's campaign spent $281 million; Hinojosa's spent $36.8 million — a 7.6-to-1 advantage, and he lost by nearly 15 points. Per vote, that is $92.92 for Abbott against $9.02 for Hinojosa: he spent 10.3 times more per vote and still lost by more than a million.

The Musk math

Elon Musk gave at least $445 million to federal committees through mid-October, per CNN's analysis of FEC filings — the largest disclosed individual donor in modern American politics. CNBC's later tally put the total at $474 million, including $224 million in Texas alone. Most of it flowed through two channels: $216 million to America PAC, the super PAC he built in 2024, and at least $168 million through the Senate Leadership Fund and other groups.

America PAC spent about $87 million across 44 federal races. As of Wednesday morning after the election, Republicans it backed had won 3, lost 29, with 12 uncalled. More than $80 million — 93% of its spending — went to races Republicans lost. In Texas, where Musk spent the most, America PAC poured more than $33 million into electing Ken Paxton to the Senate; James Talarico won. Talarico had raised $146 million to Paxton's $22.3 million — the one race where the other side outspent Musk's machine, 6.5 to 1.

For scale: $474 million is about 0.05% of Musk's roughly $924 billion net worth. The richest man in the world spent half a billion dollars on an election and, in the races his own super PAC spent on, won 3 of 44.

The industry's answer is a PDF

Two days after the election, OpenAI published a National Community Compact and a Compute Campus Policy Blueprint: future campuses it owns and operates will pay the full cost of their electricity and grid infrastructure, publish site-level water and power use, construction plans and independent noise studies, and stop asking public officials to sign NDAs (with narrow exceptions for security information and trade secrets). Futurum reports the compact also includes accepting up to 1,000 MW of curtailment in Georgia.

Three things the announcement does not change. First, the timing is misleading: chief global affairs officer Chris Lehane said the blueprint had been in development since August, and Amazon had already stopped using NDAs with government agencies in October — OpenAI is following, not leading. Second, the compact covers only future campuses OpenAI owns and operates; it says nothing about the leased and partner capacity where most of its compute actually lives. Third, compliance is self-reported in annual reports OpenAI writes itself. There is no independent verification and no penalty for falling short.

That is the shape of the new constraint. A blueprint is a document; a permit is a decision by a city council. The election moved the second, not the first.

What actually changed

The permit wall is now a two-sided squeeze. On one side, voters: 40 of 41, in red counties as well as blue ones. On the other side, capital: two days after the vote the Nasdaq posted its worst day since April, and specialized data-center operators fell double digits (Hut 8 -23%, IREN -11%, per The Decoder) while Microsoft, Amazon and Alphabet rose. The market is pricing the same thing the voters said: the next bottleneck for AI infrastructure is not chips or power plants, it is permission.

The question I keep coming back to: if 40 of 41 votes went against you, and the richest donor in modern politics just spent $474 million and won 3 of 44 races, is the problem really that communities do not understand data centers — or that the industry keeps asking for trust it has not earned? OpenAI's blueprint is a start. But a voluntary promise from one company, covering only the campuses it owns, enforced by its own annual report, is not a permit. The next two years will be decided in council meetings and statehouses, not in blog posts.

Sources: TechCrunch, "After the midterms, things are looking rocky for Elon Musk and data centers" (Nov 7, 2026); Ballotpedia data via CNBC on the 41 ballot measures; CNN, "How one billionaire could 'singlehandedly' alter the course of the midterms" (Oct 30, 2026, FEC filings through mid-October); CNBC's post-election tally of Musk spending ($474M) and America PAC's race record, as reported Nov 4; Associated Press results for the Texas governor's race via People (56.69% / 42.02%, 4,080,339 / 3,024,099 votes); OpenAI, "Setting a new standard for responsible AI infrastructure" (Nov 5, 2026); Futurum Group on the 40-of-41 tally and the 1,000 MW Georgia curtailment; Heatmap on the 830+ restrictive measures in 2026. All percentages and per-vote figures are my own arithmetic on those published numbers.

Top comments (0)