Voters Reject AI Data Centers: What the November 2026 Ballot Measures Mean

On November 3, 2026, voters in dozens of American towns decided whether AI data centers could be built near them. By early the next morning, the side against data centers had won 33 of the 34 measures with complete counts. Pooled across the 231,724 votes counted, 72.1% went to restricting data centers (see Sources, [1]).

The industry has spent two years solving one bottleneck after another: chips, memory, power and cooling. This vote added another one. Here is what was on the ballot, why residents said no, and what it means for the AI buildout.

What was on the ballot

The tracker followed 45 local measures on November 3; Ohio had 20 of them (see Sources, [1]). Ballotpedia, which counts slightly differently, listed 41 measures in 10 states: a yes vote on 28 would limit data center development, and a yes vote on 13 would facilitate it (see Sources, [3]).

The measures came in several forms:

The megawatt caps say the most about scale. Nvidia is preparing its data center power design for racks that draw 1 MW each, starting in 2027 (see Sources, [9]). A 25 MW cap therefore allows about 25 of those racks, and a 1 GW AI campus would be 40 times the limit.

Almost one-sided

Megawatt-cap measures went 11 for 11 to the restrict side, counting those still leading (see Sources, [1]). The pattern was not new: six data center measures decided earlier in 2026 all went the restrict side's way, with 66% to 90% of the vote (see Sources, [1]).

The only win for supporters came in Harwood, North Dakota, where voters rejected a charter amendment that would have required independent review before large industrial projects, 75.6% to 24.4% (see Sources, [1]). Stateline likewise reported that only one measure, in North Dakota, went the way of supporters of data center construction (see Sources, [2]).

Why people say no

A national UMass Amherst poll of 1,000 adults, fielded August 21–26, 2026, found that 11% would support an AI data center in their local community and 65% would oppose one; 49% were strongly opposed. The margin of error was 3.5 points (see Sources, [4]).

Opposition crossed party lines: 76% of Democrats and 52% of Republicans opposed a local data center, with 18% of Republicans in support (see Sources, [4]). Asked to explain in their own words, respondents most often cited:

• Environmental harm and natural resources: 28%
• Local disruption, land use and poor community fit: 15%, with noise the most frequent complaint
• Rejection of AI and distrust of the industry: 13%
• Higher utility bills: 10%

Twelve percent gave economic reasons in support (see Sources, [4]).

The power bill question

Much of the eastern U.S. is served by PJM, a grid operator that runs a capacity auction paying power plants to be available at peak. The capacity price was $28.92 per megawatt-day for the year starting June 2024, $269.92 for 2025/26 and $329.17 for 2026/27, the last of these under a price cap (see Sources, [5]). That is more than 11 times the 2024/25 price.

PJM's independent market monitor estimated that data centers caused 63% of the price increase in the 2025/26 auction, about $9.3 billion recovered from customers through higher electric rates. Capacity prices are estimated to add about $16 a month to the average residential bill in Ohio and $18 in western Maryland (see Sources, [5]).

The evidence on household bills is not settled. A Rutgers analysis this year concluded that data centers are not raising bills "in a robust, clearly detectable way. At least not yet" (see Sources, [6]).

For what a given load means in electricity use and cost, try our AI data center power and cost calculator.

The money at stake

Wells Fargo estimates that 27% of planned U.S. data center gigawatts are in Texas, 7% in Georgia, and 5% each in Ohio and Pennsylvania (see Sources, [7]). Ohio, with its 5% share, held 20 of the November 3 measures (see Sources, [1]).

The industry spent to win these votes. Tech companies and their trade groups gave more than $400,000 to defeat local measures in Ohio, according to county campaign finance reports (see Sources, [8]). In Oregon, Ohio, a city of fewer than 20,000 people, opponents of a 25 MW cap raised $335,000, including $250,000 from Vantage Data Centers, while supporters of the cap reported $4,600 (see Sources, [8]). The cap was still passing, with 60.7% on a partial count (see Sources, [1]).

Markets reacted the next day: CoreWeave was off 7% on Robinhood and Nasdaq-100 futures were down 0.8% (see Sources, [7]). States, meanwhile, still compete for the investment: as of April 2026, 38 had created tax incentives for data center construction (see Sources, [3]).

What it means

For the AI buildout, permission is becoming a constraint alongside chips, memory, power and cooling. A company can buy GPUs and sign power contracts, but a town can still vote no, and in 2026 almost every town that voted did. The next test comes on December 15, when Mandan, North Dakota votes on banning data centers of 25 MW or more (see Sources, [3]).

For how power limits are reshaping the hardware itself, see 800-Volt Power for AI Data Centers and AI Has a Power Problem.

This article is for educational and informational purposes only. It is not legal, political, financial or investment advice.