Public opinion turns sharply against AI data centres
Opposition to a nearby data centre rose from 42% to 75% in a year across every party, turning the AI buildout into a midterm liability.
- Government & policy
- Culture & impact
- Compute & infrastructure
- Major
Over roughly a year, American public opinion on the data centres that power artificial intelligence moved from split to overwhelmingly hostile — a shift steep enough, and uniform enough across the political spectrum, that by late August 2026 it had become a subject of open concern inside the industry itself. The clearest single measure came on 20 August, when Heatmap News published polling showing 75% of Americans would oppose a data centre being built near where they live, with 61% saying they would strongly oppose it. A year earlier the same question, asked with the same wording, had split roughly evenly.
This entry treats that turn as a single trend rather than a single event: it is not pegged to one announcement so much as to the moment a year of polling, local fights and campaign advertising converged into a recognised national problem for AI infrastructure — and prompted, from parts of the industry and its allies, the counter-suggestion that the backlash was not organic at all.
Timeline
- September 2025 — Heatmap Pro first asks registered voters whether they would support or oppose a nearby data centre: 43% support, 42% oppose.
- February–March 2026 — opposition passes half in Heatmap’s tracking (51%); a University of Pennsylvania survey in the same window records 49% opposed.
- March 2026 — Virginia’s Court of Appeals halts the 37-building “Digital Gateway” campus after a community challenge, and Oracle and OpenAI scrap a planned Abilene, Texas site.
- May 2026 — Heatmap’s number reaches 70% opposed.
- 4 June 2026 — three House Energy and Commerce Republicans ask the FBI to investigate whether foreign adversaries are behind the opposition; the same week, an OpenAI-linked super PAC admits running false-flag “doomer” accounts.
- June–July 2026 — the Annenberg Public Policy Center survey fields, finding 61% opposed.
- 10 August 2026 — Senator Bernie Sanders warns three AI chief executives to pause development and, with Representative Alexandria Ocasio-Cortez, pushes a moratorium on new AI data centres.
- 19 August 2026 — a National Republican Senatorial Committee memo warns that data-center opposition is costing Republican candidates.
- 20 August 2026 — Heatmap publishes the 75% figure.
The polling collapse
The Heatmap Pro poll, conducted by Embold Research among 2,045 registered voters between 8 and 13 August with a margin of error of ±2.3 points, is notable less for any single number than for the trajectory: opposition rose from 42% to 75% in twelve months, a 33-point swing, while strong support fell from 13% to 4%. Heatmap’s Robinson Meyer stressed that the wording had not changed across the year, and that the movement was near-uniform across demographics — by the August reading Republicans were 43 points net opposed, independents 65, Democrats 75, and rural voters 63. Heatmap also reported that more than 530 counties and municipalities had by then restricted or banned the facilities.
Independent surveys pointed the same way. The Annenberg Public Policy Center found 61% of adults opposed to new local data centres in its June–July fielding, up 12 points from February–March — with 69% of Democrats, 54% of Republicans and 53% of independents opposed, and support down from 21% to 14%. Annenberg’s Matt Levendusky framed the result as optimism about AI that “disappears” when the question becomes a facility nearby. Echelon Insights polling reported by Newsweek put opposition at 62%, holding at 58% even after respondents were read the industry’s economic-benefit case, and found that more Americans would accept a nuclear plant in their community (34%) than a data centre (27%).
The concerns driving the shift, across these surveys, were concrete rather than abstract: electricity bills, water use, and a sense that large facilities were being sited with little local say. Few respondents framed their objection as opposition to AI as such.
From backyard to ballot box
What made the trend a story in August was that it had stopped being a set of local zoning fights and started shaping campaigns. Reporting documented data-centre attack ads in gubernatorial and Senate races in Wisconsin, Texas, Pennsylvania and Ohio, run by both parties; governors including Pennsylvania’s Josh Shapiro and Texas’s Greg Abbott issued orders imposing cost, water and transparency conditions on developers; and the retired footballer Jason Kelce and the activist Erin Brockovich lent the cause celebrity and organising muscle.
The most telling document came from the Republican side. A National Republican Senatorial Committee memo, first reported on 19 August, warned that “anti-data center attacks have been effective” against the party’s candidates — singling out Ohio, where Republican Jon Husted trailed Democrat Sherrod Brown — and urged AI companies to make the public case for who benefits from the facilities and who pays. Semafor reported the wider party scrambling “to remake” its stance, with candidates calling for community “guardrails” and even the rollback of industry tax breaks. It was a striking reversal for a party that had championed the AI buildout, and a sign the issue now cut across the coalition that had promoted it.
The “psy-op” question
As the polling worsened, a competing explanation gained currency in parts of the industry and among its political allies: that the backlash was engineered. On 4 June, House Energy and Commerce chairmen Brett Guthrie, John Joyce and Bob Latta asked FBI Director Kash Patel and White House science advisers David Sacks and Michael Kratsios to investigate whether foreign adversaries — China in particular — were orchestrating the opposition, citing work by the Bitcoin Policy Institute and the advocacy group Power the Future.
The Committee on Energy and Commerce writes to express our concerns regarding evidence that strongly suggests foreign influence campaigns targeting artificial intelligence development in the U.S.
— Chairmen Guthrie, Joyce and Latta, letter to the FBI and PCAST
The investor Kevin O’Leary, whose proposed 40,000-acre Utah project drew heavy local resistance, likewise attributed opposition to Chinese influence campaigns before later agreeing to shrink the plan; a parallel op-ed genre argued the resistance was “manufactured” by environmental-philanthropy money. The claims sat awkwardly against the data: the opposition was, if anything, strongest among Republican and rural voters, and one of the few documented astroturf operations of the period ran the other way — the pro-AI super PAC that admitted commissioning false-flag “doomer” accounts. Sceptics of the psy-op framing noted that a movement spanning both parties, hundreds of localities and every region is difficult to reconcile with a foreign operation, and that treating constituents’ electricity bills as enemy propaganda was itself a political risk.
Why it mattered
For most of the buildout, community opposition had been a friction — a delayed permit here, a voided rezoning there — against a backdrop of broad public indifference. The August 2026 polling marked the point at which indifference became majority hostility, and hostility became electorally material enough that the industry’s own allies asked it to change its public conduct. It did not stop any specific project by itself, and it left open the question the polls could not settle: whether the collapse in support was a durable verdict on how the AI power boom was being conducted, or a spike that benefits messaging and time might reverse. Either way, it reset the political ground on which every subsequent data-centre fight — and the moratorium proposals already in Congress — would be waged.