Voters Reject Data Centers in 44 of 45 Ballot Measures Across Ten States

Voters approved restrictions on data center development in 44 of an estimated 45 ballot measures across ten US states and 40 localities this week, a near-uniform rejection of the facilities that now underpin AI infrastructure. Only one pro-development outcome survived: in Harwood, North Dakota, a proposal to add a review process for industrial projects, including data centers, failed, according to Ballotpedia data. A handful of measures remained unsettled at the time of reporting.

The votes cap a year of escalating friction between hyperscale buildouts and the communities asked to host them. New Street Research analyst David Barden put a number on the cost in a note released Wednesday: "AI capex is running at $690-725B in 2026, but local opposition blocked or delayed ~$198B of U.S. data center projects in the first half of 2026 alone. On Nov. 3, data center development restrictions won or led in 39 of 45 local votes."

That $198 billion figure is worth sitting beside the total: blocked or delayed projects equal roughly 27% of a full year of AI capital spending, based on New Street's own range. The scale of the loss is no longer a rounding error for an industry that treats siting as a solved problem.

Where did the restrictions pass?

Winning measures spanned bans, direct limits and the rejection of pro-development proposals across California, Florida, Illinois, Kansas, Maine, New Jersey, Nevada, Ohio and Wyoming. Some states had already paused projects or passed moratoriums before Tuesday, so the ballot results extend an existing trend rather than start one.

Ohio stood out. Twenty-one of all data center-related measures appeared on that state's ballots — nearly half the national total — on a night when voters also elected Amy Acton as their first Democratic governor in nearly two decades. Acton, who beat Republican Vivek Ramaswamy, supports a conditional moratorium on new data center projects. Ohio voters also returned Sherrod Brown to the US Senate; Brown, who defeated incumbent Republican John Husted, opposed large tax breaks for data center projects and criticized Husted's record of supporting them during the campaign. The statewide results and the local measures point the same direction.

How fast is public opinion moving?

Quickly, by polling standards. A tracking poll cited by New Street Research shows opposition to data centers rising from 42% in August 2025 to 75% in August 2026 — a 33-point jump in twelve months. The sentiment is bipartisan: 69% of Democrats, 54% of Republicans and 53% of Independents say they oppose the facilities, with opposition above 60% "in cities, suburbs, and rural areas alike," according to the analysis.

Attitudes toward AI itself track the same curve. "73% of Americans are worried about increased use of AI (up from 68% in 2023), and 39% say AI is having a negative impact on society vs. 11% a positive one. This gap is widening. Zero of 25 countries surveyed show 'excitement' as the dominant AI sentiment," Barden writes.

That last figure is the one operators should weigh most carefully. It is one thing for US zoning fights to slow construction; it is another for negative AI sentiment to dominate across 25 surveyed countries, which suggests the problem is not localized permitting friction.

What does the industry do now?

Barden argues operators must "change the story" around AI, "from AI being pushed onto communities, to AI demand being pulled by consumers who are already benefiting." His framing is economic rather than rhetorical: "Data center demand is pull-based, not push. Consumers want what comes out of the data center. Like power industry issues to date, throttling data center development leads to higher prices for end-users and revenue (potentially) for data center incumbents. Navigating the thorny reality of cost-benefit trade-offs will distinguish leaders in the industry."

There is a self-interested caveat embedded in that argument, and readers should notice it: incumbents facing constrained supply benefit from higher prices, so "throttling" is not uniformly bad news for companies that already hold capacity. Barden's note does not address which incumbents stand to gain.

The counterfactual he offers — that communities restricting data centers will pay more for the services they demand — is a plausible economic claim, but it remains a claim. The note offers no elasticity estimates or price forecasts to substantiate it.

Is there an escape route?

Elon Musk offered one, from a weakened position. The SpaceX CEO spent over $445 million in an unsuccessful attempt to secure Republican majorities on Tuesday, and used election night to revive a longer-horizon plan. "Data centers in space looking pretty appealing right now," Musk posted on his X platform early Wednesday morning. SpaceX has said it hopes to begin deploying orbital AI data centers in late 2027 — a date that has yet to be matched by a demonstrated launch architecture, and one that sits well beyond the window in which the roughly $198 billion of blocked and delayed capacity would have come online.

For now, the near-term arithmetic is unforgiving. Public opposition has climbed 33 points in a year, every state that voted on restrictions except one chose them, and Ohio — the state with the most measures on the ballot — has just elected a governor who favors a conditional moratorium. Barden's own polling record suggests the trend is widening, not peaking, which means the $690-725 billion of planned 2026 AI capex will meet a siting environment measurably harder than the one the industry budgeted for at the start of the year.