Skip to content

Leadership1 publisherNot yet confirmed elsewhere3 min readPublished

Data centre opposition hits 75%, and the friendly county has stopped existing

A new Heatmap Pro poll puts local opposition at three-quarters of Americans, with strong opposition above 60%. Siting is now a permitting programme, not a land deal.

The Board Room · Leadership desk

How we use AISend a correction

Photograph accompanying Data centre opposition hits 75%, and the friendly county has stopped existing
Photo: heatmap.news

What happened

  • A new Heatmap Pro poll by Embold Research finds 75% of Americans would oppose a data centre built near where they live.
  • The same question, worded identically, has been asked four times in 12 months and has moved 33 points against the facilities.
  • Candidates from both parties in the Wisconsin governor's race have moved to distance themselves from data centres.

Compiled by The Board RoomSomething wrong?How this is made

Why it matters

  • constraint The rural siting playbook loses its premise: there is no cohort in this poll whose reception is materially warmer, so cheap land no longer buys a quieter hearing.
  • cost With four in five opponents strongly opposed, sponsors carry land, grid and equipment commitments against approval dates that a mobilised minority can push out.
  • decision Boards approving 2026 capacity have to underwrite political durability alongside power and fibre, or accept an unpriced tail on every site.
  • precedent When opposing a project costs a candidate nothing in either party, tightening ordinances becomes the cheap move for local officials watching the same numbers.

Opposition polling matters to a developer only insofar as it predicts who turns up at the hearing. On that test, the intensity split is the number to price. Heatmap Pro reports 75% opposition overall and more than six in 10 strongly opposed [1][2], which means at least four out of five opponents sit in the category that files comments, organises neighbours and runs candidates [16]. Strong opposition was 55% in May, so the hard core has grown by at least five points in a matter of months [18]. Meanwhile strong support has fallen from 13% last August to 4% [8], a drop of roughly two-thirds [17]. There is almost no local constituency left for a sponsor to activate; anyone speaking for the project will have to be recruited or paid.

The pace deserves attention because the instrument did not move. Heatmap says it asked the same question four times in 12 months without changing a word, and recorded a 33 point swing [3], about 2.75 points a month [14]. That tracks the arithmetic of the readings themselves: 42% opposed last August against 75% now [5][1][15]. Unchanged wording removes the convenient dismissal that the question got harsher.

The more expensive finding is the absence of a soft spot on the map. Rural voters, the cohort that cheap land and spare interconnection capacity usually points you toward, are 63 points net negative, and urban and suburban voters are only a few points more supportive [10]. Republicans, the friendliest partisan group, are still 43 points net negative, with independents at 65 and Democrats at 75 [9]. Heatmap describes the movement as present across age, gender, income, party and the rural-urban divide [11]. A siting strategy built on finding a receptive county is now a search across a distribution with no favourable tail.

Politicians read the same numbers earlier than boards do. In the Wisconsin governor's race, candidates in both parties have moved to distance themselves from data centres [12]. When opposing a project costs an elected official nothing across the primary and the general, the discretionary parts of the approval process, the extra study, the deferred vote, the moratorium pending an ordinance rewrite, get cheaper to use. And the argument has left the trade press: Jason Kelce fronted an ad telling Americans to mail their urine to AI data centres over water use [13].

What follows is a budgeting question rather than a rhetorical one. If approval probability is materially below one and time to approval has a long tail, then land options, transformer and turbine slots, and reserved silicon are all being carried against a date the sponsor does not control. That cost belongs in the model as a line, with a named owner for local affairs, a contingency for hearings that run past schedule, and a written view on what the host community is offered before the first filing rather than after the first packed meeting.

One caveat the poll's own framing supports: this measures a stated response to a hypothetical nearby project, asked four times by one pollster, Embold Research, for Heatmap Pro [1][3]. It is not a record of votes cast on real applications. It is, though, the cheapest early read available on how the next hearing goes.

What to watch

  • Whether the next reading of the same unchanged question flattens near 75% or keeps climbing.
  • Whether any jurisdiction that has already approved a large campus reopens or conditions the permit.
  • Whether hyperscaler capex disclosures begin naming permitting delay, rather than power availability, as the binding constraint.
Loading claim ledger
Loading source directory links
Loading share composer
Loading topic controls
Loading related stories