Brief us

WP-29: Terrain

The substation is on the ballot

On November 3, voters in ten states took up some forty local measures on data centers and sided against the projects in all but one. A siting is now a campaign, and most of it is decided before the application is filed.

By Brushworks research deskNovember 2026Working paper

Abstract. Large computing infrastructure in the United States is now sited through local politics as much as through permits. We read the 2025 and 2026 record (project delays, state bills, some forty local ballot measures and the national exit poll), identify the two grievances that organize the opposition, and describe a program for developers, utilities and public bodies that starts at site selection, earns its claims before making them, and measures every contact against a sealed control.

In brief

The finding
A data center siting is a local campaign with an electorate, a calendar and an organized opponent. The opposition crosses party lines, starts before the filing and turns on the power bill and on not being asked.
For a developer or operator
Start at site selection. Make the commitments before the communications: pay the full cost of power, name the tenant, cap the water, sign the benefit agreement. Expect the question to reach a ballot.
For a utility or a public body
The rate case and the siting are one argument in the household’s mind. Answer the bill first.

Local opposition is now the binding constraint on large computing infrastructure in the United States. It does not sort by party, it organizes before a project is announced, and it is built on two things: a household’s power bill and a community’s sense that it was not asked. This paper sets out what the record of 2025 and 2026 shows, why the usual response fails, and what a program built on behavior and a sealed control does instead.

What the record shows

One tracker counted at least 75 projects worth about $130 billion blocked or delayed in the first quarter of 2026, close to its total for all of 2025, and another 45 projects worth about $68 billion in the second quarter.1, 2 The same tracker found organized opposition groups in 49 states, more than 300 state bills filed in the first six weeks of the year, and moratorium proposals in 14 states from members of both parties.1 These are counts of disruption, not of permanent cancellation, and they come from one privately produced tally.

Projects blocked or delayed by local opposition in the first two quarters of 2026, in billions of dollars of announced investment, with the number of projects. Disruption, not permanent cancellation. Basis: record, one private tracker as published and as reported (notes 1 and 2).

Then the question went to voters directly. Reporting from election week counts 41 local measures in ten states, 21 of them in Ohio, with voters siding against development in every one but a town of fewer than a thousand people in North Dakota.3 A second count puts the number of restrictive measures approved at 44.4 In one Ohio city a charter amendment restricting construction passed with 64.7 percent, on unofficial returns.5

Local ballot measures on data center development in November 2026, by outcome. Counts differ by source: 41 in one tally, 44 restrictive measures approved in another. Basis: record, election-week reporting, unofficial returns (notes 3 and 4).

The national exit poll asked about it. Twenty-three percent of voters called AI data centers crucial to their vote for Congress and another 33 percent called them a major factor. Asked about a new facility in their own community, 76 percent were opposed.6 The opposition leaned toward one party and was not confined to it: a September poll found 70 percent of Democrats and 57 percent of Republicans in favor of limiting new construction.3

How voters weighed data centers in 2026. The first three rows are from the national exit poll of November 3; the last two from a September poll on limiting new construction. Basis: official exit poll and published polls, as reported (notes 3 and 6).

Why it forms, and when

Three findings matter more than the totals.

It is about the bill. A 2026 Federal Reserve Bank of Dallas working paper estimated that data centers had raised wholesale electricity prices by 6.9 percent as of last year.3 A voter does not need the paper. She has the envelope. Research on local energy siting found the same mechanism a decade ago: in Ontario, the governing party lost between four and ten percent of its vote share in places with wind turbines, a cost borne locally for a benefit spread widely.7

It is about not being asked. An analysis of 46 projects delayed or canceled between January 2024 and May 2026 called lack of transparency the single most reliable driver of fast-forming, durable opposition: nondisclosure agreements signed by local officials, tenants that were not named, water and power impacts that were not stated.8 The literature on place attachment says what practitioners learn the hard way: what gets labeled as reflexive objection is usually a defense of a place by people who were given no part in changing it.9

It is early. The same analysis describes opposition that organizes within days.8 By the hearing, the room has been decided. A developer who starts communicating at the application has entered a race in its last week.

Why the usual response fails

The usual response is a jobs figure, a tax figure and a mailer, sent to everyone, late. It fails for reasons this Review has set out before. Late persuasion on a salient question moves almost no one; the best available estimate of the average effect of campaign contact in a contested general election is zero.10 A message sent to everyone is written for no one. And a claim the audience can check and finds thin does more damage than silence.

Renaming the facility does not help. Neither does a message carried by a messenger whose sponsor is hidden. We do not run either, and the record suggests they do not work.

What a program does instead

Read the terrain before the site is chosen. Permitting history in the neighboring jurisdictions, the rate case calendar, the last three local elections precinct by precinct, who attended the last zoning fight. The first deliverable is sometimes a recommendation not to file here. That is the cheapest result a program can return.

Find the people who have not decided. In every one of these fights there is a committed opposition, a small committed support, and a large middle that pays a power bill, has heard something, and has not been spoken to by anyone. They are identified by what they do: ratepayer status, well ownership, proximity, attendance, local news readership. Not by party, and not by any identity field.

Earn the claim, then make it. The commitments that answer the grievance are specific: the operator pays the full cost of its power and the infrastructure that carries it; the tenant is named; water use is stated and capped; the benefit agreement is signed before the application. Two days after the election one of the largest developers pledged the first of these in public.4 A program can carry a commitment like that to the household it matters to. It cannot substitute for one.

Say it from a disclosed sponsor, through people the community knows. Every placement is from the sponsor’s own accounts. Local validators are real, briefed and disclosed.

Hold out a control. One tenth of every cohort receives nothing. Support, hearing turnout and petition signatures are measured against it weekly. If a cohort moves the wrong way, the readout says so on the first page and the contact stops.

The bill, the water and the agreement

A grievance this specific has a specific answer, and most of the answer is economic before it is communicative. The household that opposes a data center is usually making a prediction about its own costs. A program that wants to change the prediction has to change the facts it rests on, and then make the new facts easy to check.

The bill. The cost a large new load imposes on a grid is real: generation to serve it, transmission to carry it, and the reserve margin the system has to hold. The question a ratepayer asks is who pays. Several utilities and state commissions have moved toward large-load tariffs, which require a facility to pay for a minimum share of the capacity it reserves whether or not it uses it, so that the cost of a project that underperforms does not land on households. Where such a tariff exists, it is the most credible sentence a developer has. Where it does not, offering its substance by contract does the same work.

The water. Cooling demand varies by design and by climate, and a community that has lived through a dry summer will assume the worst figure it has read. The answer is a number, a cap and a meter: what the facility will draw at peak, the ceiling it agrees to, and where the reading is published.

The agreement. A community benefit agreement signed before the filing tells a town that it was asked. Signed after a contested hearing, the same document reads as a settlement. The research on procedural fairness explains why the order matters: people accept outcomes they dislike when they believe the process that produced them was fair, and they reject outcomes they might have accepted when it was not.11

A siting run as a campaign, from site selection to the hearing. The commitments come before the communications; the holdout is sealed before any contact. Basis: illustrative sequence, months relative to filing.

None of this is persuasion in the narrow sense. It is the substance a persuasion program carries. The program's job is to find the households for whom each commitment answers the question they are actually asking (the ratepayer for the bill, the well owner for the water, the neighbor for the agreement), to say it to them in a voice they already trust, and to measure, against people left alone, whether it changed what they believe about their own costs. When it does not, the result is not a reason to spend more. It is a reason to change the commitment, or the site.

Questions a board should ask before the filing

For a developer's or utility's board, the record in this paper reduces to a short list of questions to ask before a site is filed.

Has anyone read this place? Not the parcel, the place: its last three local elections, its last zoning fight, its rate case, who organized and who turned out.

What will the household pay, and can we say so first? If the project will raise local power or water costs, by how much, and what commitment would prevent it.

Who has been told? Have local officials been asked to sign nondisclosure agreements? Is the tenant named? Each secret is a grievance waiting to be discovered.

What will we measure? Not reach or impressions, but whether the people who live there believe something different about their own costs after hearing from us, against people we did not contact.

What would make us walk away? A board that cannot answer this has not yet decided whether it is running a campaign or a fait accompli.

Implications

For a developer or operator. Treat the siting as a contest with an electorate and a calendar. Start at site selection. Budget for commitments before communications. Expect the question to reach a ballot.

For a utility or a public body. The rate case and the siting are one argument in the household’s mind. Answer the bill first.

For anyone. There are sites where the terms on offer do not survive contact with the people who live there. A measured program finds that out in weeks, and for a fraction of the cost of finding it out at the hearing.

Notes and references

  1. Data Center Watch, Q1 2026 report (public summary), datacenterwatch.org/q1-2026.
  2. Data Center Watch second-quarter figures as reported by Bloomberg, September 21, 2026; the 2025 total as reported in June 2026 coverage of the first-quarter report.
  3. Al Jazeera, “How the US midterm elections reflected resistance to data centres,” November 7, 2026. Two measures were still being counted at publication.
  4. FOX 4, “How data centers lost in the midterm elections,” November 2026, including the developer’s statement of November 5.
  5. Unofficial county results, Defiance, Ohio, as reported in election-week coverage.
  6. National exit poll conducted by SSRS for a consortium of news organizations, November 3, 2026, as reported by Data Center Dynamics.
  7. Stokes, L. C. (2016). Electoral backlash against climate policy: a natural experiment on retrospective voting and local resistance to public policy. American Journal of Political Science, 60(4), 958–974.
  8. Analysis of 46 projects, January 2024 to May 2026, by an energy and climate advisory firm, as reported by Route Fifty, July 2026.
  9. Devine-Wright, P. (2009). Rethinking NIMBYism: the role of place attachment and place identity in explaining place-protective action. Journal of Community & Applied Social Psychology, 19(6), 426–441.
  10. Kalla, J. L., & Broockman, D. E. (2018). The minimal persuasive effects of campaign contact in general elections. American Political Science Review, 112(1), 148–166.
  11. Tyler, T. R. (2006). Why People Obey the Law. Princeton University Press.

Basis of figures. Project counts and values: one private tracker’s tally of blocked or delayed projects. Election results: news reporting from November 4 to 12, 2026, unofficial where noted; two counts of the local measures differ (41 and 44). Polls: as reported, with the pollster named. No Brushworks program figures appear in this paper.

Next paperThe Review

Start a conversation

Brief us.

Tell us the race, the calendar and the margin. Within a week we will tell you whether we can move it and what it would take. Candidate work and independent-expenditure work are taken by separate teams; say which you are. Governments and institutions: tell us the mandate and the jurisdiction, and we will tell you what registration it requires before we say anything else.

Send a briefbrief@brwrks.com

  • Candidate practice
  • Independent-expenditure practice
  • International practice

Brushworks Strategies Inc., a Wyoming corporation · Washington, New York, Ottawa, London, Munich, Dubai, Tokyo