A brownfield site in Newton Heath
In Newton Heath, on the north side of Manchester, Latos Data Centres has applied to build a data centre off Bower Street. It would be two buildings with about 28,000 square feet of floorspace on a three-acre brownfield site. The developer calls it a “neural edge” data centre. I have worked in technology for most of my life, and I don’t know what that means. It is open for objections now, and I object.
My objection isn’t to computing as such. The problem is that nobody can tell us what this building is for. Data centre developers rarely tell the public what workloads they will run. Neither do the councils that grant them planning permission or the government bodies that give them subsidies. In 2025, someone used a freedom of information request to ask the Department for Science, Innovation and Technology which companies use Britain’s largest data centres, those of more than 15 megawatts, and what they use them for. The department said it did not hold that information. This is despite the government having designated data centres as critical national infrastructure. The state is promoting these projects, and in some cases providing land, money and political will, without knowing what goes on inside them. Freedom of information requests help a little, but you can only ask about things you already suspect exist.
How can a community decide whether a development is needed, useful or harmful when nobody, not even the people approving it, has the facts?
“Magnets for investment”?
In September, Labour MP Ian Lavery asked Prime Minister Andy Burnham at Prime Minister’s Questions for a national moratorium on data centres. Burnham refused, arguing that a data centre “can be the magnet that clusters in other investment over time”.
Anyone who has been inside a data centre knows better. There’s nobody there. It’s machines and electricity, with an occasional visit from someone to plug or unplug something. Construction is quick and workmanlike: pour the concrete, put up a shell, bring in the power and cooling, fit out the racks. Once it’s running, the local economy would hardly notice it, apart from the noise, the traffic and a big empty box. The think tank Verdant estimates that planned sites will create only about a quarter of the permanent jobs the industry claims. Comrades at Anti*Capitalist Resistance have pointed out that the few jobs these sites do create are mostly low-paid.
Built in the wrong places
It is expensive to move electricity to a data centre and cheap to move data out of one. So data centres belong where the power is, not where the people are. Yet they keep being proposed in cities. We don’t generate electricity in Newton Heath, and it isn’t a major telecoms hub.
There is one exception that makes sense, though only for capital. In finance, milliseconds make millions, which explains the data centre planned at the old Truman Brewery on Brick Lane, in a residential area on the edge of the City. Tower Hamlets council refused permission, and ministers overruled it this summer. The data centre will draw up to 5.2 megawatts in a borough with a severe housing shortage, and the wider scheme it belongs to will provide only six homes for social rent.
Meanwhile, Scotland has so much renewable power coming in from the North Sea that turbines are regularly disconnected because the grid can’t take it all. The same happens around the Wash in eastern England. If we need more computing capacity, those are the places to build it.
Bigger than you think
Phil pointed out that Brick Lane is small by industry standards. At Drax power station in North Yorkshire, Drax, Brookfield and its AI business Radiant plan more than a gigawatt of AI data centre capacity, about 200 times Brick Lane, next to a plant that burns imported wood pellets. In Lanarkshire, DataVita has promoted its AI campus near Airdrie as powered by more than a gigawatt of renewables. A Guardian investigation found no evidence it has that power, or even the grid connection it needs. In Memphis, xAI’s Colossus complex was built around gas turbines. SpaceX now owns it and leases it out: Anthropic has agreed to pay up to $1.25 billion a month for its computing power. Burning natural gas is arguably as bad for the climate as burning coal once methane leaks are counted.
The national picture is even starker. Britain’s electricity demand peaks at about 45 gigawatts on a cold winter evening. According to Ofgem, data centre projects in the queue for grid connections would need about 73 gigawatts, more than one and a half times what the whole country uses at peak. Consultancies expect only 10–20% of them ever to be built, which tells you how speculative this rush is. It comes just as we are trying to electrify transport and home heating. Data centres on the grid also have to install diesel generators as backup, and the weight of all that equipment is one reason they sprawl across such large areas instead of building upwards.
Manchester is no exception. Kao Data’s 40-megawatt site at Kenwood Point in Reddish, Stockport, is due to open this year. A campaigner from Greater Manchester Opposition to Data Centres described further sites planned across the city region.
A bubble, and who pays for it
There’s growing concern that slow and uneven capacity growth and lss-maaking business plans create risks. If AI demand grows more slowly than expected, some data-centre projects or regions could end up with more capacity than they can profitably use. Building continues anyway. The hardware inside lasts four to six years before it is obsolete. In March, OpenAI closed a record $122 billion funding round. The Wall Street Journal reports that it expects to spend that within about three years on computing contracts, and its own chief financial officer has reportedly worried about whether it can pay for them.
When the bubble bursts, working people will pay. British industry already pays about four times as much for electricity as American industry, and the government is under pressure to subsidise power for data centres while households pay more. In AI growth zones, councils can keep all the growth in business rates and receive targeted government funding, so the public sector carries much of the cost of attracting US companies such as Microsoft, Google, OpenAI and Palantir. Campaigners warn that public lending and pension funds, including the Greater Manchester Pension Fund, are being drawn into financing the build-out. It is our pensions that are being bet on this.
Water is another problem. We’ve had hosepipe bans because there isn’t enough water for people. Where will the water for these machines come from, and who will be told to go without?
Generative AI is where the waste is
As Hannah pointed out from online, not all AI is the same. A model trained to read radiographs, check turbine blades or spot upside-down milk bottles on a production line does a specific job, and it often needs only modest computing power, perhaps on a hospital’s own servers. Most of the energy is going into generative AI: huge language models trained on everything that can be scraped, much of it copyrighted, much of it slop, with many companies training on the same data at the same time.
In the 1990s.users set the model running, went home, and hoped it had finished by morning. Because computing power was scarce, people had to be careful about what we fed it. Now the approach is to throw everything in and boil the ocean.
Businesses that adopt AI mostly do what they already did, a bit faster. AI is built from what people have already written and said, so it struggles to produce anything new. What it does transform is the flow of subsidy and speculative investment, and anyone who could find out whether it adds real value seems determined not to look.
What AI reproduces
Several people in the discussion argued that the problem is not only environmental. Liz noted that transparency is a central demand in trade union critiques of AI: workers find it introduced into their workplace without explanation, then worry about who will be blamed when it goes wrong. She also asked whether we are already drowning in poor-quality information, and whether more of it is liberating anyone.
A researcher in the room described new university guidance asking for £5,000 to £10,000 in every research budget for AI tokens. That money could go to participants and fieldwork, and once it becomes standard practice it gives the tech companies a guaranteed income. They also pointed to the injustice built into how these systems are made. Writing and languages, including Indigenous languages, are taken without acknowledgement and sold back to us by the token. Low-paid workers in Kenya and elsewhere do the mind-numbing labelling that keeps the systems running.
Ian argued that AI doesn’t make universities worse so much as amplify what was already worst about them: a system where students absorb what they’re told, repeat it in an exam and forget it. AI and data centres plug into existing relationships of alienation and power and intensify them. They also reproduce racism: facial recognition systems have repeatedly been shown to misidentify Black people.
Stephen, speaking from a background in automation and control engineering, asked the question underneath all this. Why, at this stage of capitalism, is so much money pouring into these projects, and what return do investors really expect? He pointed back to Harry Braverman’s Labor and Monopoly Capital and Mike Cooley’s Architect or Bee?, and to earlier debates about nuclear power, as reminders that the left has tools for analysing technology as social power and not only as kit.
People are already saying no
Resistance is growing. On 16 September, the Scottish Parliament voted 80 to 26 that no planning decisions should be made on hyperscale data centres until the Scottish Government publishes national planning guidance. In the United States, voters sided with restricting data centre development in 40 of the 41 local ballot measures on the issue in this month’s elections. They backed outright bans in California, Florida and Ohio and moratoriums in Maine, and rejected every rezoning put to them that would have let a data centre go ahead. Voters are demanding that developers carry the risk instead of the public, and courts have sometimes backed residents’ demands for referendums and disclosure. Nobody put “no computers” on a ballot. People were voting against opaque concrete developments with no proven need and unclear costs.
Anti*Capitalist Resistance supports the Green Party’s call for a pause, and goes further: on the information available, we see no need for more data centres. We back a moratorium. Our international organisation, the Fourth International, has launched a campaign against data centres dedicated to AI. Those sites show most clearly how speculative this investment is.
The wrong kind of opposition
There is a danger here. In the US, part of the movement against data centres is led by Steve Bannon and a section of the far right that is hostile to the tech billionaires. They frame it as nostalgia for a better past and resentment of shadowy financial elites. Anti-science populism can easily take over popular anger at these projects.
The left needs a different argument. Jobs are part of it. London used to be where people from Manchester went to find work. Now it has the highest unemployment of any UK region: 6.6%, against a national rate of under 5%. Office work there was first offshored and is now increasingly automated, and young people are struggling to find jobs in a way not seen since the early 1980s. It is easy to blame AI for this, not the politicians and the system behind it.
Data centres are a symptom of how capitalism concentrates power, hides its workings from the people affected, and calls waste “growth”. If I pay people to build giant ice-cream statues, GDP goes up. That doesn’t make the statues useful.
A socialist approach starts from use value. Phil compared it to the private car, which sits idle for about 98% of its life: an absurd use of resources that we’ve been taught to see as normal. A rational society would put computing to work where it is genuinely useful, such as medical imaging or modelling stresses on wind turbines. It would use automation to abolish boring, exhausting work, like the night shifts I once spent counting newspapers, and share out the work that remains through a shorter working week. It would not build gigawatt-scale data centres so that chatbots can write students’ essays for them.
What next
Hannah asked the most urgent question of the evening: what are we going to do? Groups are already forming. There is Greater Manchester Opposition to Data Centres, sessions organised in Reddish, and a discussion at Trafford trades council. These efforts need to join up and plan demonstrations locally, not in London, because the data centres are being built here. They need to object to planning applications such as Newton Heath’s, demand disclosure of what every site is for, resist public subsidy and pension fund investment, and push councils and the government to put that money back into communities.
Above all, the opposition needs to be democratic and anti-capitalist, so that the anger at these grey boxes turns into a fight for control over the resources and technology that shape our lives.
Manchester Anti*Capitalist Resistance holds public meetings every two months, along with smaller and online meetings. ACR’s quarterly magazine, Ecosocialist Futures, is out now.
Thank you for a thoughtful article. A local campaign in Cardiff has just forced escalation of the decision on a data centre from the local authority to Welsh Government, although that just changes the focus of the struggle.
Thanks Duncan very informative