Trump's Voters Want AI — Just Not in Their Backyard

By Daniel Rao

Photo: Inside a data centre — the kind of building voters increasingly do not want nearby. Credit: Christopher Bowns, via Wikimedia Commons (CC BY-SA 2.0).

American voters have just delivered the artificial intelligence industry a verdict it cannot afford to ignore, and the verdict is gloriously, maddeningly contradictory. Build the future, they say. Just do not build it next to my house. According to CNBC's All-America Economic Survey, published on October 24, 63% of registered voters oppose an AI data centre within 10 miles of their home, while only 13% would support one. That is not a close call. That is not a margin-of-error wobble. That is a country looking at the physical footprint of the AI boom and saying, in effect: we like the idea, we distrust the building, and we are fairly sure we do not want to live next to it.

The contradiction sharpens when you ask the same voters about AI itself rather than its architecture. A majority, CNBC found, support stopping the development of advanced AI until Congress passes safety regulations — a striking position for a country that is also being told, daily, that falling behind on AI would be a national catastrophe. And yet the same survey shows voters, including the president's own supporters, want development to continue while Congress works on the rules. Hold those two thoughts at once and you have the defining political tension of the AI era: Americans want the technology, want the guardrails, and above all want somebody else's neighbourhood to host the machinery.

We checked CNBC's survey findings against the Reuters/Ipsos and Quinnipiac polls from earlier in October, and the pattern is consistent across all three: opposition to nearby data centres is broad, rising, and increasingly bipartisan. This is not one survey's quirk. It is a trend, and it arrives ten days before midterm elections that will decide control of Congress. Candidates from both parties are already running against AI expansion in places where a data centre proposal has become a local fight. The industry spent three years arguing about models. The argument has now moved, decisively, to land, power and water.

The split: build AI, yes; build it here, no

The 10-mile question is the one that matters, because it turns an abstraction into a neighbour. Ask voters about "AI" and you get philosophy. Ask them about a building with its own substation, its own water demand and its own hum, ten miles from their front door, and you get politics. Sixty-three percent opposed, 13% in favour: those numbers describe a public that has made up its mind about the physical side of the boom long before the policy debate has caught up.

There is a useful comparison sitting in the same polling season. The survey's finding that a majority back pausing advanced AI development until Congress passes safety rules sounds, at first hearing, like a public turning against the technology. But read alongside the rest of CNBC's data, it looks more like a public demanding a referee before the game gets rougher. Voters are not, on this evidence, Luddites. They are conditionalists: proceed, but with rules, and not in my backyard. That combination — permissive on technology, restrictive on siting — is the worst possible combination for an industry whose business model requires building very large buildings very quickly in a great many places.

The texture of the opposition matters too. This is not an abstract anxiety about robots. It is concrete: electricity bills, water use, noise, traffic, the sense that a community bears the costs while distant shareholders collect the gains. Brad Carson, a former Democratic congressman from Oklahoma and co-founder of Public First Action PAC, an AI-industry-backed group pushing regulation and safeguards, put it plainly: "Many people are annoyed by these data centers. People are also strongly insistent upon reasonable guardrails of AI, and that's going to be hard to beat." Annoyed is a mild word for a political force. It is also, historically, the word that precedes a movement.

The MAGA split in numbers

If you want to understand why this survey rattled Washington, look at the president's own coalition. Among supporters of the Make America Great Again movement, 55% want companies to keep developing advanced AI while Congress works on regulations, against 27% who favour a pause. That is a pro-development majority, and it fits the administration's posture of AI dominance first, rules later. But ask those same supporters about a data centre within 10 miles of home and the picture flips: only 29% would support one, against 40% opposed. The gap between "build AI" and "build it here" is 26 points wide inside MAGA itself.

That split is the whole story of the next year in miniature. The administration's AI policy is built on speed: build the infrastructure, win the race, sort out the details later. Its voters' preferences are built on place: speed is fine, but the substation goes somewhere else. A movement that agrees with its leader on the destination and disagrees with him on the address is a movement that will produce some very interesting primary campaigns, because every data centre proposal creates a local constituency of opponents who vote in every election, while the benefits are diffuse, national and largely invisible.

It also explains why the president himself has started talking about data centres in the language of local deals rather than national destiny. When your own voters support the technology by a wide margin and oppose its local footprint by a wider one, the rational political move is to change the local bargain, not the national message. Which brings us to the dividend.

Democrats lead the opposition, but this is not partisan

Democrats are, on these numbers, the most opposed to nearby facilities: 80% oppose a data centre within 10 miles of home, with just 4% in support. Among all Republicans and Republican-leaners, 24% would support a nearby facility — a minority, but a noticeably larger one than among Democrats. If the story ended there, it would be a familiar partisan tale. It does not end there, because the other polls in the sequence show the opposition spreading across party lines rather than settling into them.

Reuters/Ipsos, polling on October 7, found 56% of voters back limits on building data centres — including 69% of Democrats, 55% of independents and 43% of Republicans. That same poll found 84% saw AI as a threat to American workers, and 80% of Democrats and 61% of Republicans supported stricter AI regulation. A Quinnipiac relay from October 2 found 72% would oppose a new data centre in their community, up from 65% in March. Tamilla Triantoro, associate professor at the Quinnipiac University School of Business, warned: "Securing local support for the physical infrastructure behind AI may become one of the industry's most immediate challenges." Three pollsters, three different months, one direction of travel: up.

The trust backdrop makes that climb easier to understand. The same month produced the finding that 73% of Americans don't trust AI companies, a number that reframes every siting fight: when a company you do not trust asks your town to host its largest building, the answer is rarely yes. Trust is the industry's missing permit, and no zoning board can grant it. The data centre debate is, at bottom, a trust debate wearing a planning permission costume.

What Trump noticed: the dividend post

Donald Trump, to his credit as a political reader of rooms, noticed the pattern before most of the industry did. On October 21, in a Truth Social post reported by Reuters, he argued that companies building data centres should pay a "dividend" to communities to make developments popular. His verbatim reasoning is worth quoting in full, because it is unusually candid about how local opinion actually changes: "I have watched the process closely, and data centers that were highly unpopular in a certain area, have become extremely popular when a financial deal is struck." The translation is unflattering and probably accurate: opposition is, in part, a negotiation.

Trump added that new and large data centres are "mandated by me to make their own electricity, and give or sell any excess back to the local community, or grid." That second clause is doing quiet work. The single most potent objection to data centres — that they will raise everybody else's electricity bills — is answered, in Trump's formulation, by making the facility its own power island, with surplus flowing back to neighbours. He had made the same point in February, saying major tech companies must build their own power plants for data centres to protect consumers from rising bills. The policy, such as it is, amounts to: you may build, but you may not bill your neighbours for the privilege.

Whether a "dividend" is a bribe, a benefit-sharing agreement or simply the price of consent depends on your view of local politics, but the direction is unmistakable. The White House has moved from celebrating data centres as monuments to American greatness to treating them as developments that must buy local goodwill. That shift did not happen because the industry asked for it. It happened because voters, in poll after poll, told the president's own pollsters that the monuments were unwelcome next door. The politics of the buildout now runs through the parish council, not the press release.

Update, Oct. 24: how the poll was run — and what the industry says back

The numbers above come from CNBC's All-America Economic Survey, conducted October 14–18 by Hart Research Associates and Public Opinion Strategies among 1,000 registered voters, with an overall margin of error of 3.1 percentage points; the data-centre question went to about half the sample, so its margin is wider, at 4.38 points.

Read the fine print and the story holds: the same survey found 52% of all respondents would rather pause advanced AI development until Congress passes safety regulations than press ahead to keep pace with China. The industry's answer, via Data Center Coalition representative Dan DeOrio, is that data centres create jobs, attract investment and generate tax revenue — and that operators intend to pay for the energy they use. That promise, not the technology, is what the next round of local fights will be about.

The money question: bills, water, power

Strip away the rhetoric and the data centre fight is a household-budget fight. Electricity is the largest and most visible cost: a facility that draws as much power as a small city can, if its costs are socialised through the grid, push up bills for everyone else on that grid. Water is the second cost, and in drought-prone regions it may be the more emotive one: cooling systems for AI hardware consume water at industrial scale, and communities that have watched reservoirs fall do not warm to new industrial thirst. Jobs, the industry's traditional answer, are the weakest card of all: a modern data centre employs relatively few people once construction ends, and the jobs it does create are often specialised and filled from outside.

That asymmetry — concentrated local costs, diffuse national benefits — is why Senator Elizabeth Warren's proposal has found an audience. In a September/October CNBC interview, Warren called for a national moratorium on new AI data centres until developers demonstrate that the full costs, covering grid, water and jobs, are accounted for. Her summary was characteristically blunt: "You do a national moratorium." Whether a moratorium is workable policy is a separate question — the industry would call it a gift to America's rivals, and it has a point — but as politics it is potent, because it names the exact bargain voters feel they are not getting: proof, before construction, that the deal is fair.

The industry's counter-argument deserves a fair hearing. Data centres are, genuinely, the factories of the AI age, and factories have always been unwelcome neighbours that societies eventually accept because the alternative — importing the product and exporting the jobs — is worse. The companies also point, with some justification, to the tax revenue, the grid investment and the construction employment that a facility brings. But "eventually accept" is doing a lot of lifting in that sentence, and the polling suggests the public is not yet at the acceptance stage. It is at the "show me the money, the megawatts and the water accounting first" stage, and it is not bluffing.

Why November 3 makes this unavoidable

The November 3 midterms decide control of Congress, and that single fact converts a local planning dispute into a national campaign issue. Candidates from both parties now run against AI expansion in places where a data centre proposal has divided a community, because opposing a data centre costs a candidate nothing with the national party and gains them everything with the local electorate. The dynamic is self-reinforcing: each local campaign legitimises the next, each poll showing 63% or 72% opposition emboldens the next challenger, and each company's decision to press ahead regardless confirms the suspicion that local voices do not count.

The parallel from Britain is instructive. The backlash against data centre expansion is not an American invention: communities, councils and campaigners in the UK have fought similar battles over power, water and landscape, and the industry has learned there, as here, that a planning consent is not the same thing as a social licence. The news that Microsoft walked out of Nscale's Essex data centre after a power row showed that even the largest companies will abandon a site when the local equation stops working. American developers should read that episode as a preview, not an aberration: when power and politics collide, politics usually wins the site.

There is also a longer-term political danger for the industry that goes beyond any single election. If data centres become, in the public mind, something imposed on communities rather than chosen by them, the regulatory response will eventually be national, blunt and bipartisan — the kind of regulation that pleases nobody and constrains everybody. The smarter companies already understand this, which is why some have begun offering community benefit funds, local power deals and water-recycling commitments before they are forced to. The dividend idea, whatever its provenance, is the industry's least-bad option: pay for consent now, or pay for regulation later.

Our take: the backyard is the ballot box

Here is our judgment, plainly stated. The AI industry has spent three years winning the argument about the future and losing the argument about the neighbourhood, and the neighbourhood is where elections are held. Voters are not anti-technology; the CNBC numbers show a public that wants development to continue under rules it trusts. But voters are, emphatically, anti-imposition, and a data centre that arrives with its power demands, its water use and its distant ownership is experienced as an imposition however many jobs its press release promises. The industry's colossal error has been to treat local consent as a permitting hurdle rather than a political fact.

Trump's dividend post, for all its transactional bluntness, recognises that fact more honestly than most corporate statements do. People change their minds when the deal changes. That is not cynicism; it is how communities have always weighed big infrastructure, from railways to wind farms. The companies that survive the next decade of siting fights will be the ones that offer real, verifiable local value — their own power, water accounting the public can audit, and money that stays in the community — before breaking ground, not after the protests start. The ones that do not will discover, site by site and election by election, that the most powerful force in AI politics is not a model or a chip. It is a voter who has just learned what is being built ten miles from home. And that voter's answer, right now, is no.

The buildout-politics context is broader than one survey, and it includes the strange new skyline politics of Trump's AI towers, where the symbolism of AI power is literally being built into the landscape. But symbolism does not pay a water bill, and voters know it. The question for the industry is no longer whether Americans want AI. It is whether Americans will let AI move in next door. On the evidence of October's polls, the answer is: not until the deal improves, and not without being asked properly first.

Frequently asked questions

What did the CNBC survey actually find?

CNBC's All-America Economic Survey, published October 24, found 63% of registered voters oppose an AI data centre within 10 miles of home, with only 13% in support. A majority also support stopping advanced AI development until Congress passes safety regulations, though CNBC did not publish a single headline percentage for that question in the material available to us.

Do Trump's supporters oppose data centres?

It is split. Among MAGA supporters, 55% want companies to keep developing advanced AI while Congress works on regulations, versus 27% who favour a pause — but only 29% would support a data centre within 10 miles of home, against 40% opposed. Support for the technology and support for the building are two different things.

Is this just a Democratic position?

No. Democrats oppose nearby facilities most strongly (80% oppose, 4% support), but Reuters/Ipsos found 43% of Republicans back limits on building data centres, and Quinnipiac found 72% of all respondents would oppose a new facility in their community. This is a cross-party trend, not a partisan talking point.

What is Trump's "dividend" idea?

Trump proposed that companies building data centres pay a "dividend" to host communities, arguing that unpopular developments become popular when a financial deal is struck. He also said new large data centres are mandated to make their own electricity and give or sell excess power back to the local community or grid.

Could this affect the midterms?

Yes. The November 3 midterms decide control of Congress, and candidates from both parties are already campaigning against AI expansion in affected areas. Local siting fights are becoming national campaign material, which raises the political cost of every new proposal.

Sources: CNBC All-America Economic Survey (published October 24, 2026) via CNBC relays; Reuters' October 21, 2026 report of Donald Trump's Truth Social post on data centre dividends; Reuters/Ipsos poll (October 7, 2026); Quinnipiac relay (October 2, 2026); CNBC interview material with Senator Elizabeth Warren (September/October 2026).

Daniel Rao is a technology journalist covering cloud computing, artificial intelligence and the business of the internet.