In the U.S. election on Nov. 3, 2026, 47 data center-related ballot measures were held across 10 states, and the side favoring construction limits led in most of them. According to the tracker Data Centers on the Ballot, unofficial counts as of 12:45 p.m. Eastern Time on Nov. 4 showed that the pro-restriction side led in 44 of the 45 measures with results. The only one won by the side allowing development was in Harwood, North Dakota. Combining all votes, the restriction side took 72.8%, underscoring strong local wariness toward data center construction. However, what residents backed was not a uniform construction ban. The measures ranged from limits based on a facility's power capacity to mandatory voter approval and nonbinding advisory votes. Media reports also cite different totals, from 29 to 47, so understanding the results accurately requires distinguishing the time of each count and the content of each question.
Restriction side leads in 44 of 45 measures, with 72.8% of combined votes
Of the 47 measures that Data Centers on the Ballot tallied for Nov. 3, 45 had a determined leader by midday on Nov. 4. The restriction side led in 44, the pro-development side won one, and results for two measures in Maine, in Wiscasset and Brownfield, were still unavailable. To allow comparison of differently worded questions, the site classifies each side of every question as either "restriction" or "pro-development." For example, a "no" vote on a question asking whether to allow data center construction counts as a restriction vote. The site also defines a "lead" as a margin of at least 2 percentage points, and all results listed are unofficial, pending certification.
Adding up the 45 measures, the restriction side received 185,668 votes and the pro-development side 69,517, giving the restriction side 72.8%. This is calculated against 255,185 total votes and is based on unofficial counts as of 12:45 p.m. ET on Nov. 4. The figure includes 10 nonbinding advisory votes, and counting continues in six measures, so the final share may change.
Meanwhile, the per-question restriction share published by the tracker has a median of 76.58% and a mean of 73.5%, which differ from the 72.8% obtained by simply pooling votes. The median and mean give every question equal weight regardless of turnout, whereas the pooled share is heavily influenced by results in municipalities with more voters.
For example, in Hubbard, Ohio, 2,622 votes were cast in favor of an ordinance banning data centers outright, against 802. After Trumbull County election officials counted all 158 precincts, support reached 76.58%. By contrast, in Janesville, Wisconsin, 26,711 votes were cast in total, with 16,779 (62.82%) backing the restriction side. Janesville alone accounts for roughly 10% of the total votes across the 45 measures, so its comparatively low support rate pulled down the pooled share. Even so, restriction-side support exceeded 70% whether measured by median, mean or pooled total.
From bans above 25 MW to mandatory referendums, the restrictions vary widely
The 47 measures include several that apply restrictions based on a facility's power capacity. The tracker recorded a megawatt (MW) threshold for 16 measures, and 12 of them ban data centers above a certain size. The breakdown is 10 measures banning facilities above 25 MW (nine in Ohio and one in Ottawa, Kansas) and two banning those above 10 MW (Emporia and El Dorado, Kansas). The restriction side led in all 12, though counting continues in Ottawa and Oregon, Ohio. The ordinance in Urbana, Ohio, also includes a provision exempting self-use facilities under 7.5 MW from the ban.
| Role of threshold | Count | Municipality / standard |
|---|---|---|
| Ban on facilities above a certain size | 12 | Above 25 MW: nine in Ohio and Ottawa, Kansas / Above 10 MW: Emporia and El Dorado |
| Condition requiring voter approval | 1 | Grove City, Ohio: 20 MW (counting continues) |
| Subject to a 180-day construction moratorium | 1 | Sabattus, Maine: 20 MW |
| Condition for tax incentives (bond and noise-mitigation guarantees) | 1 | Perry, Ohio: 50 MW |
| Standard requiring independent review | 1 | Harwood, North Dakota: 100 MW |
Importantly, not all 16 measures with MW thresholds are construction bans. Twelve are direct bans, while the remaining four set conditions such as voter approval or independent review. An ordinance banning construction above 25 MW, for instance, does not cover facilities at or below 25 MW. The Emporia, Kansas, question is stricter: it bans facilities in all of the city's zoning districts if their power demand exceeds 10 MW or their daily water use exceeds 100,000 gallons (about 380 cubic meters). Meanwhile, Grove City's 20 MW is the threshold for holding a referendum and Harwood's 100 MW is the threshold for requiring independent review; neither immediately bans facilities above the line.
Some restrictions have no power-capacity standard at all. The Hubbard, Ohio, ordinance bans data center construction in the city entirely, regardless of size. In Janesville, Wisconsin, a measure requiring a referendum on developments worth more than $450 million on the city-owned former GM/JATCO site was approved, although the text itself never mentions data centers.
Of the 44 measures where the restriction side led, 34 were binding votes, while the other 10 were advisory votes conveying residents' wishes to local governments. In 13 measures framed as "a yes vote allows development," the "no" votes prevailed in every case. In an advisory vote in Joliet Township, Illinois, 14,207 voted against and 3,800 in favor. The U.S. outlet Stateline also reported that the votes included measures of differing natures: outright bans, temporary construction moratoriums, energy-use caps, pre-construction voter approval and advisory questions.
Restrictions backed even where Amazon and Meta plan sites; Harwood the lone exception
Even in municipalities where major tech companies plan data centers, support for construction limits was the majority. In Sunbury, Ohio, where Amazon Data Services plans a data center worth about $2 billion, 2,809 votes were in favor of a question banning facilities above 25 MW, against 763. In Piqua, Ohio, where Meta plans a campus of about 607 acres (about 246 hectares), 4,061 votes backed banning facilities above 25 MW, exceeding the 2,027 against. However, it is unclear whether Piqua's restriction applies to Meta's already approved plan.
The Janesville, Wisconsin, vote also came against the backdrop of a large data center plan. According to Data Centers on the Ballot, the aim was to block a planned 800 MW, roughly $8 billion facility conceived by Viridian Partners. However, local reports say the city council passed a resolution in June to let its agreement with the developer lapse, so the plan had effectively been withdrawn before the election.
In contrast, the only one of the 45 decided measures where a restriction proposal was rejected was in Harwood, North Dakota. A charter amendment that would have required independent review of large industrial projects of 100 MW or more failed, with 105 votes in favor and 325 against. The area adjoins an AI data center campus of about $3 billion and 280 MW being advanced by Applied Digital. Both Stateline and Data Centers on the Ballot report that Harwood was the only case in this election where the pro-development side won.
Comparing the thresholds with actual development sizes shows the scale of what is being regulated. In Ohio, 25 MW became the dividing line for bans in many ordinances, while the campus next to Harwood is 280 MW and the project envisioned in Janesville reached 800 MW. However, ordinance thresholds and developers' stated facility capacity are not necessarily calculated using the same definition, so simple comparison requires caution.
Moreover, even when the restriction side wins a vote, the ordinance does not necessarily take effect immediately. In Emporia, Kansas, the unofficial final count in Lyon County was 4,346 votes in favor of construction limits and 2,891 against. But according to the tracker, even after the majority is formally confirmed, publication, enforcement and implementation of the ordinance are suspended until a court issues a final ruling.
Why reports cite 29 to 47 ballot measures
Reporting on this election gives widely differing counts of data center ballot measures. Comparing the main tallies shows that the numbers are affected not only by when the count was taken but also by differences between the number of municipalities and the number of questions, and by the range of states covered.
| Source | Time (ET) | Count | What was counted |
|---|---|---|---|
| Ballotpedia | Sept. 21 (pre-election) | 34 | Measures scheduled for Nov. 3. 38 across eight states for all of 2026 |
| Stateline | Article at 12:48 a.m. on Nov. 4 | 33 | Measures where tighter restrictions passed or were leading. About 12 others not yet announced |
| Washington Post | Article at 6:39 a.m. on Nov. 4 | At least 29 | Municipalities where bans or restrictions passed |
| Data Centers on the Ballot | 12:45 p.m. on Nov. 4 | 47 | Questions held in 44 municipalities across 10 states. Restriction side leads in 44 |
Particularly significant is the difference between the number of municipalities and the number of questions. Stateline reported that votes were held in 21 Ohio municipalities, but Data Centers on the Ballot says 21 is the number of questions, covering 18 municipalities. Ballotpedia's pre-election tally put Ohio at 17 municipalities and 20 measures, so the numbers do not match here either. In Wilmington, for example, separate votes were held on four districts, so four questions were counted from one municipality. In this way, one ballot measure does not necessarily correspond to one municipality.
Stateline's 33 also cites the Data Centers on the Ballot tally, so the two are not independent findings. The gap between 33 and 44 is thought to stem mainly from the timing of the counts. Meanwhile, the Washington Post's "at least 29 municipalities" is lower even though it was published after Stateline's article. This is probably because the paper counts by municipality, and because "at least 29" is a lower bound of what could be confirmed rather than a definitive total.
Differences in the states covered also matter. The 38 for all of 2026 that Ballotpedia published in September covered eight states: Ohio, Illinois, Wisconsin, Kansas, California, Michigan, Nevada and Florida. The 47 in Data Centers on the Ballot include eight in Maine, two in New Jersey and one in North Dakota, none of which are among those eight states. The site counts a broad range of questions directly related to data centers, including construction bans and power caps as well as rezoning of sites, development agreements, voter-approval systems and advisory votes.
Therefore, "44 of 47 measures led by the restriction side" is an unofficial result based on the questions the tracker had identified as of midday Nov. 4, and does not mean that 44 municipalities passed restrictions. Counting also continues in six measures: Alhambra, Ottawa, Downe, Boulder City, Grove City and Oregon. In Grove City, the restriction side's share is a relatively close 54.6%, and in Boulder City, Nevada, counting mail-in ballots is expected to take several days.
65% oppose local construction; restrictions could spread statewide
Residents' wariness about data centers is not limited to this vote. In an NBC exit poll, 56% of voters said "AI and data centers" were either the deciding factor or a major factor in their vote in congressional elections: 23% said "deciding" and 33% said "major factor." However, the survey asked about the political importance of AI and data centers together, and did not directly measure support or opposition to building facilities.
In a survey of 1,000 people conducted through YouGov from Aug. 21 to 26 by the University of Massachusetts Amherst (UMass Amherst), only 11% supported building AI data centers locally, while 65% opposed. Classifying open-ended responses on reasons, concerns about the environment and resource use accounted for 28% of all respondents, land use and community impact 15%, distrust of AI and related industries 13%, and economic burdens such as higher utility bills 10%. Associate Professor Alexander Theodoridis of the university noted that many Americans may oppose not data centers themselves but their being built where they live.
Similar friction is occurring in Japan. According to government materials, as of 2023 about 90% of the country's data centers by floor area were concentrated in the Tokyo and Osaka areas. Jiji Press reported that in March 2026, residents in Inzai, Chiba Prefecture, filed a lawsuit seeking cancellation of the building confirmation for a data center planned near a train station, and a similar suit was filed in neighboring Shiroi. In response, the Japan Data Center Council issued guidelines in May urging thorough consultation with nearby residents and local governments when considering facility sites.
In Tsukuba, Ibaraki Prefecture, Australia's Goodman is also planning a data center development that would ultimately reach 1 GW and about 3 trillion yen in investment. The company has begun construction on a first-phase 50 MW facility, but according to Reuters, local residents and others have criticized it because the roughly 45-hectare site is adjacent to homes, a hospital and care facilities. Not only a facility's power and water demand but also land use and impacts on the surrounding environment are becoming important issues in advancing development.
Still, the results of these U.S. ballot measures cannot be taken as proof that all voters rejected data center construction itself. The 25 MW and 10 MW standards were set by those who proposed the ordinances, not chosen by voters after comparing multiple sizes. And while some votes backed outright bans, 10 were nonbinding advisory votes. Whether residents want construction halted entirely, limited by size or given a role in the development process must be judged according to the content of each question.
Going forward, the focus will be on formal certification of the results and the implementation of the ordinances. Data Centers on the Ballot plans to update its tally hourly through Nov. 13, so the results of the remaining six measures may change. The Emporia ban ordinance also awaits a court ruling.
In Ohio, meanwhile, Conserve Ohio is seeking a constitutional amendment to ban data centers above 25 MW statewide. The group missed the July 2026 signature deadline but aims to put the measure to a vote in November 2027. If the amendment is put to voters and approved, the 25 MW construction limit that has so far been adopted municipality by municipality would extend statewide. This election highlighted that conflicts over data center siting may grow beyond individual municipalities into a state-level policy issue.