The AI Boom Has a Dirty Secret, and Your Town Is Paying for It

The Hum Behind the Revolt: How AI Data Centres Ignited a Transatlantic NIMBY Politics

I

On the outskirts of Hertfordshire, a large white box is growing. Around it are green fields, a river, and further off, housing estates. Inside the box are tens of thousands of chips performing trillions of calculations per second. The switching of the chips is silent, but every watt of power that feeds them ultimately escapes as heat. Thousands of fans spin day and night to keep the heat from destroying rows of server racks. Transformers in the substation emit a low hum. Security lights burn through the night, cutting the darkness into pieces.

What nearby residents hear is this.

In May 2026, Gallup published a poll. When asked about building a data centre in their own area, 71% of respondents were opposed, with 48% strongly opposed. For a nuclear power plant, opposition was 53%, with 34% strongly opposed. Since Gallup began surveying attitudes toward nuclear plants in 2001, opposition had never exceeded 63%. On its first appearance in the survey, the data centre cleared that line with ease.

The potential hazards of a nuclear plant are physically far more serious than those of a data centre. But a meltdown is abstract. It is distant and unlikely. The hum of a data centre is immediate. You can hear it, see it, and feel it in your bills.

Over the past two years, this unpopularity has only deepened. Some polls show half of Americans support a complete moratorium on data centre construction.

II

Bernie Sanders has called for a pause on all new data centre construction. He told the BBC: “People look around and they are worried—worried about how fast this technology is moving, worried about how little control they have over it, worried about how data centres will affect electricity prices and water use in their own communities.”

But the opposition is not only on the left. Some town councils have been voted out over their data centre policies. Some states, red and blue, are proposing bans. In New Jersey and Virginia, Democratic governors won partly because of a backlash over rising electricity prices. Virginia is home to the world’s largest concentration of data centres. Its new governor pushed legislation mandating new energy tariff standards for data centres and requiring disclosure of energy and water use.

The most symbolic turn came in Texas. A state once known for being friendly to business, data centres, and energy companies, its Republican Senate candidate and current Attorney General, Ken Paxton, now publicly attacks data centres. He opposes tax breaks for them and has proposed that if the technology “empowers AI chatbots that harm children’s safety,” the operators of those facilities should bear criminal liability.

Texas once fought fiercely to win data centres, securing most of a multibillion-dollar relocation investment by Elon Musk’s SpaceX AI giant. Paxton’s shift is a dizzying political U-turn.

In March 2026, the chief executives of the seven largest tech companies appeared with Trump to sign a “Ratepayer Protection Pledge.” They promised to “build, bring in, or buy” their own energy and ensure costs would not be passed on to “ordinary households.” The pledge is non-binding. To many, it looks more like a public relations document.

III

The argument has crossed the Atlantic.

In Slough, Europe’s largest data centre cluster, more than 500 people signed a petition saying the town “should not be an experiment.” They question whether the town benefits and are calling for a pause on data centre development there. Residents complain that the heat from data centres makes already hot summers worse. One site originally earmarked for 1,000 homes was sold to a data centre company as part of the local council’s debt repayment.

In Brick Lane, in east London, a data centre project likely to serve high-frequency traders in the City of London has met opposition. Some would rather see more social housing and local businesses in east London. Baroness Pola Uddin, a former Labour peer, said: “I understand the need to build data centres, but the health and wellbeing of local residents must be properly considered—people who live, breathe, and go to school here. We must consider choosing more industrial areas.”

Protesters held signs: “Water is for people, not for AI.” Another read: “Hey ChatGPT, what’s the fastest way to destroy Brick Lane?”

A data centre project in Buckinghamshire, repeatedly rejected by the local council because it was on green belt land, was approved after the Labour government came to power. Protesters crowdfunded a legal challenge arguing that the environmental impact had not been considered. The government admitted an error and withdrew its approval. The developer eventually conceded it needed a binding clean energy obligation agreement with the council. The winning campaign group, Foxglove, vowed to refocus on challenging other data centre projects across the UK.

Microsoft’s data centre project in Leeds has collected more than 42,000 objections, and 250 people took to the streets.

The Labour government has designated data centres as “Critical National Infrastructure.” It argues they are a key part of public services and essential to attracting investment and keeping Britain competitive in the global economy. The political challenge is whether local people are willing to put up with them.

IV

An operating data centre typically requires only dozens to hundreds of permanent staff. Even the largest new facilities directly employ only a few hundred people. The tax revenue it contributes may be sharply reduced by exemptions. The electricity it consumes may push up the load on the entire regional grid. The cooling water it uses competes directly with residents’ water in dry seasons.

In its submission to a House of Commons committee, Water UK criticised the government’s AI growth strategy for “saying nothing about water.” “There seems to be an assumption that the country’s water resources will always be sufficient to meet its economic needs. Nothing could be further from the truth.” The body noted that more than three-quarters of England’s data centres are concentrated in the south and east, precisely the areas under the greatest water stress, while the water needs of these facilities are “explicitly excluded” from the government’s water resources planning.

The UK government aims to triple existing data centre capacity to 6 gigawatts by 2030. Progress may reach only about 3.3 GW, leaving a “compute gap” of up to about 5 GW. Water constraints could make that target even more remote.

Meanwhile, AI companies approach trillion-dollar valuations while predicting mass unemployment. When many people question whether the technology will wipe humanity off the face of the Earth, asking them to accept noise, security lights that never go off, and pressure on local energy and water supplies, and telling them it is for the national interest, is difficult.

V

In the United States, with control of both chambers of Congress hanging in the balance, the scale of this highly localised opposition could become a decisive factor in the midterms.

In the UK, Prime Minister Andy Burnham has promised “growth in every postcode.” Data centres may help deliver that promise. Yet the government’s leading role is also a clear example of centralised state power, at odds with Burnham’s desire to devolve more power to local government. The Green Party’s call for a moratorium on data centre construction poses a challenge to the Labour government from the left. Since work began on the National Policy Statement, the prime minister has changed, and the political temperature around the issue has shifted.

Some in Westminster believe there is an “America brain” phenomenon at work: that because it is an issue in the US, people assume it must be one in the UK too. They point out that while the UK is struggling to reach 6 GW, the US ambition for 2030 is at least 20 times that, with some single hyperscale projects exceeding the UK’s entire planned capacity. The impact on local electricity prices is not the same, though partly because UK energy prices are already higher.

For the UK, there is also an important argument about data sovereignty: avoiding complete dependence on the US or other countries for the information technology infrastructure that underpins large parts of a service economy. One minister told the BBC: “This time, on AI, we have to get it right, but only if we have a certain level of data infrastructure.” The UK’s ambition is seen as relatively modest: sustained AI adoption in the NHS, national security, and financial services by 2030.

VI

In the US, the energy system operates at the local level, and the impact on electricity prices is immediate. In the UK, energy prices are already higher, and the marginal impact of data centres on bills is relatively smaller. But the UK’s land and water constraints are more rigid, and local government’s role in planning approval makes opposition more likely to run through legal and planning processes.

The shared dilemma is structural: the AI industry is expanding far faster than the supporting infrastructure and political consultation mechanisms can be built. Grid connection may take a decade. A large reservoir may take fifteen years to plan and build. An AI data centre can be built in eighteen months to a few years. This mismatch of timescales means that even if technical solutions exist, such as closed-loop water cooling or self-built renewables, the political trust deficit will be hard to bridge in the short term.

The tech industry argues that if you account for water lost to pipe leaks, its water demand is not that large. In a dry season, residents see a data centre using water while their lawns turn brown. When AI executives conjure visions of cash flows being completely disrupted, small-town residents hear notice of another rise in their electricity bill.

VII

Data centres concentrate economic benefits globally. They disperse costs locally.

Tech companies enjoy the computing dividends and valuation growth that data centres bring, while transferring noise, water consumption, grid pressure, and environmental costs to local communities. When that transfer accumulates to a certain point, political backlash becomes inevitable.

The petitions in Slough, the protests in Brick Lane, the objections in Leeds all have their own local triggers. But they share the same logic: when the benefits of a technology cannot be perceived by the local community, and its costs can be precisely measured, political backlash is almost unavoidable.

Continuing to respond to local concerns with the language of “national interest” or “human welfare” is likely, as has already happened in the US, to be rejected by voters at the ballot box. A more effective path may be to localise the benefits of data centres through binding community benefit agreements, transparent disclosure of resource use, and giving local government bargaining power in approvals.

Otherwise, the pace of the technological revolution may slow in the mud of local politics. Doomsday warnings are distant. The electricity bill is not.