The AI industry has spent years telling us the future is weightless and invisible. Software. Intelligence. Models. Clouds.
But AI has an inconvenient physical footprint: sheds. Lots of very big sheds.
And increasingly, the people who may have to live next to them aren’t impressed.
From New York to Norfolk, communities are organising against proposed datacentres. Planning hearings are becoming battlegrounds. Petitions are attracting thousands of signatures. Politicians are discovering that state-of-the-art AI infrastructure doesn’t necessarily produce the same warm glow in a village hall as it does in a Washington or Whitehall strategy document.
How this is resolved may determine whether the next phase of the AI boom gets built at all.
Build. Faster.
The UK’s answer so far is essentially: we don’t care, and we’re going to build them anyway.
The government wants AI Growth Zones with fast-tracked planning and priority access to the electricity grid. It is also considering treating major datacentres as Nationally Significant Infrastructure Projects.
Then there’s Stargate UK, part of the giant AI infrastructure push involving OpenAI and Nvidia.
The logic is familiar: AI means growth. Growth means infrastructure. Infrastructure means datacentres. Therefore: build datacentres.
The trouble is that infrastructure has a habit of becoming extremely real once you put it somewhere people actually live.
The Village Hall mobilises
The backlash is already here.
In Norwich, a petition against the proposed Apex datacentre scheme reportedly attracted 5,000 signatures. In the Scottish Borders, campaigners have branded a proposed £2 billion development a “monster”. In Great Torrington, around 230 people turned out to oppose the £13 billion Xlinks megaproject. And in Larbert, Police Scotland has warned that “robust security” will be required because “a great deal of public opposition is likely.”
Scotland, meanwhile, is reportedly moving towards a moratorium, with one planning figure saying they have “never seen this level of objection.”
This is no longer simply a handful of residents objecting to a large building. It is becoming a political constituency.
And it has a remarkably simple argument: you get the servers. We get the electricity bill, the noise, the water shortages, the view… the list goes on.
Across the Atlantic, the revolt is bigger
America is showing what happens when that constituency scales.
A single day of action reportedly produced 142 protests across 42 states. More than 500 jurisdictions have moved to block datacentre construction, with more than 150 bans reportedly introduced in one month.
New York’s Assembly has passed a statewide datacentre moratorium, while officials approving projects are facing recall campaigns. Datacentre opposition is even becoming a bipartisan rallying cry ahead of the US midterms.
That is the part the industry should pay attention to.
American politics is famously polarised. Republicans and Democrats agree on remarkably little. But “not in my county” appears to be developing into one of the few genuinely cross-party positions.
A June Reuters/Ipsos poll found that only 14% of Americans supported an AI datacentre being built in their own community. Meanwhile, Data Center Watch estimates that at least 75 projects, worth roughly $130 billion, were blocked or delayed in the first quarter of 2026 alone.
The AI boom is colliding with something much older than artificial intelligence: local democracy.
The inconvenient jobs problem
This is where the industry’s category story starts to wobble. Datacentres are being sold as engines of economic growth. They allegedly bring investment, infrastructure, technological leadership and jobs.
But a Guardian analysis suggests UK datacentres could create only around a quarter of the jobs the technology sector predicted. That’s an awkward proposition to sell to a community.
The local resident doesn’t necessarily experience a datacentre as “AI infrastructure”. They experience it as a huge industrial building.
They see the electricity demand. They hear the cooling systems. They worry about water. They watch agricultural or undeveloped land disappear. They can’t see past its huge slab sides.
And then someone admits the development will create a relatively small number of local permanent jobs.
The macroeconomic story may still be compelling. The local story is considerably harder.
Who owns the category?
And this is where we get to our signature question: who owns the category?
Hyperscalers and infrastructure developers are making a classic category move.
Datacentres aren’t buildings. They’re critical national infrastructure.
They’re not industrial sheds. They’re engines of the AI economy.
They’re not a local planning issue. They’re strategic national competitiveness.
It’s clever. But there is a problem with this category design: the audience has a vote.
The paradigm shifting category story that works in Davos may not work in the village hall.
To a minister, a datacentre can represent billions in investment and sovereign AI capability. To someone living beside it, it can be a nightmare.
Both descriptions can be true. But only one of them determines whether the planning application gets approved.
Is this NIMBYism?
The obvious response is to call objection NIMBYism.
And sometimes it probably is. Who’d want a 30-acre industrial facility outside their previously tranquil back door? Every infrastructure project attracts people who want the benefits but not the building.
But dismissing the entire movement as NIMBYism misses something important. The physical costs of AI have been largely absent from the industry’s story.
We’ve spent years talking about AI as if it were infinitely scalable because the product is digital.
It isn’t. Every frontier model ultimately requires land, electricity, cooling, cables, transformers and buildings.
AI may be virtual to the person using ChatGPT. It is emphatically not virtual to the person living next to the substation.
That doesn’t mean every proposed datacentre should be rejected. It means communities are entitled to ask a much more basic question: what’s in it for us?
What would community-first AI look like?
Maybe there is a better way forward. If the industry wants these projects to succeed, the answer cannot simply be better PR. It needs better category design.
What would it look like if every datacentre came with a genuinely local proposition?
Cheap power or energy infrastructure for the community. Investment in local grids. Water guarantees. Skills programmes that lead to actual jobs. Local tax revenues. Community ownership or revenue-sharing. Transparent energy and water accounting.
And, crucially, an honest admission that there are trade-offs.
Because “AI is the future” is not an effective planning argument. Neither is “national competitiveness.”
A community has every right to ask why this piece of land, this amount of electricity, this amount of water and this level of disruption are necessary – and what it receives in return.
The British opportunity – or trap?
There is another issue lurking underneath all this.
What if America begins to baulk at the prospect of a vast datacentre infrastructure while Britain’s government continues to fast-track development, what happens?
Britain could become the West’s default server farm based on financial capital, deep connectivity, a supportive government and a willingness to build.
But there is also a category risk. Do we really want Britain to become the country that builds everyone else’s AI infrastructure?
Sovereignty is compromised if the models belong to American companies, the capital comes from elsewhere, the hardware comes from elsewhere – and Britain supplies the land, electricity and planning permission. And generates a well of deep resentment.
We may end up winning the datacentre race while losing the argument about who benefits from it.