Unions and others have been sounding the alarm over artificial intelligence centres’ huge energy and water use — and their potential to hold up housing development and the electrification of other industrial sectors.
Last year, the Labour government explained how it will take forward the recommendations set out in tech entrepreneur and AI policy advisor Matt Clifford’s “game-changing” AI opportunities action plan.
It promised to “turbocharge” and “unleash” artificial intelligence to deliver a decade of national renewal, with prime minister Keir Starmer “throwing the full weight of Whitehall behind this industry”.
Although Britain is the third largest AI market in the world, the then technology secretary — and current business and trade secretary — Peter Kyle said it risks falling behind the advances in AI made in the USA and China.
“Backing AI to the hilt” can, says the government, put more money in working people’s pockets and boost productivity, resulting in gains of up to an average £47 billion to the UK each year over a decade.
The plan aims to stop “blockers” controlling the public discourse and getting in the way of growth of the sector by creating dedicated AI Growth Zones “that speed up planning permission and give them the energy connections they need to power up AI”.
Environmental and societal harms
But critics say the government has not properly considered the environmental or societal harms of the dozens of data centres being developed across the country.
The UNISON public services union says AI has the potential to support environmental goals through climate modelling, pollution detection, and making energy grids and buildings more efficient, for example. But it points out that “most of these beneficial uses rely on predictive AI, not the energy-intensive generative AI (like ChatGPT, Google Gemini and Copilot) that’s driving the current boom”.
According to Carbon Brief, a UK-based website covering the latest developments in climate science, climate policy and energy policy, a government analysis released by the DSIT science, innovation and technology department concluded that emissions from data centres would be minimal.
The government analysis reported that: “DSIT’s AI environmental impacts model estimates that UK’s greenhouse gas emissions from AI compute could range from 0.025 to 0.142 MtCO2 [a unit used to express the impact of greenhouse gas emissions] by 2035, equivalent to below 0.05% of the UK’s projected total emissions.”
Emissions much higher
But Carbon Brief’s recent analysis suggests emissions from data centres could be “hundreds of times higher than government estimates” — as much as 20% of the UK’s projected total emissions in 2035. Future data centres could result in emissions equivalent to roughly a third of UK households.
The Global Action Plan (GAP) climate and nature campaign points to 2025 International Energy Agency figures showing that the energy usage of just one typical AI data centre equates to the energy usage of around 100,000 households.
It says the developers’ own figures show that just 10 of the largest data centres currently in planning or construction will cause annual emissions of more than 2.7 million tonnes, effectively wiping out the carbon savings expected in 2025 from the switch to electric cars (2.9 million tonnes).
According to CACC campaign against climate change campaign coordinator Claire James: “The building of data centres can directly compete with new housing for electricity.”
Extra pressure
And in its 2025 Gridlocked report, the London Assembly says the rise of large data centres in West London has put extra pressure on the electricity network, competing with new housing for limited supply. This has resulted in slower development and higher costs for builders and local authorities.
GAP campaigns manager Owen Espley told Labour Research the government’s AI adoption drive could also put at risk the electrification of transport and heavy industry.
“Does a heavy industry that needs to decarbonise to stay competitive get the grid connection, or a bus garage that wants to electrify and move our buses away from fossil fuels and onto electric, or the local data centre?” he asks.
“The fact that big tech companies have such enormously deep pockets means that consistently they’re able to outbid any other economic sector for access to energy.” This puts at risk the ambitions or aspirations for other economic sectors to decarbonise, he added.
GAP also says that more data centres may mean more fossil fuels. As the pace of AI data centre demand outstrips the supply of renewables and of electricity grid connections: “The industry is openly admitting it wants gas generation on site to hasten development. This will undermine progress on emissions reductions.”
The Fuel Poverty Action campaign group says AI’s vast data-processing centres are keeping energy bills high. In March 2026, a coalition of 28 civil society organisations and unions in Ireland demanded a moratorium on new data centre connections, with increased household bills one of their concerns.
Water resources
AI data centres also use huge amounts of water for cooling, “placing significant pressures on local water resources and often drawing from drinking water supplies”, says the UNISON public services union. And they are being built in areas already classed as “seriously water stressed”.
The water demands of large-scale data centres “now rival those of towns and small cities”, says GAP. “A single mid-sized data centre can use as much water each year as several thousand UK households, while a hyperscale campus can require enough water each day to meet the needs of around 10,000 people, placing it among the largest industrial users in many local water catchments.”
As awareness about the environmental impacts of AI data centres has grown, so has resistance to their development.
A “Stop Dirty Data Centres” day of action in February 2026 brought together national and local campaign groups including Action to Protect Rural Scotland (APRS), Biofuel Watch, Corporate Europe Observatory, Foxglove (“an independent non-profit organisation that fights to make tech fair for everyone”), Friends of the Earth, Iver Heath and North Ockenden residents’ associations, and the London Mining Network.
In March, Edinburgh City councillors voted for a moratorium on new data centres after hearing a speech and receiving a written deputation from APRS.
Moratorium
At global level, the United Nations UN special rapporteur on water has called for a moratorium on the development of data centres based on concerns about “their water and energy consumption and the risks they pose to climate change, the sustainability of aquatic ecosystems, the human rights of impoverished populations and the survival of vulnerable productive sectors”.
Meanwhile legal action by GAP and Foxglove against the government and data centre developer Greystoke, has established a precedent that data centre developers’ claims on measures to protect the environment must be “concrete and enforceable”.
Greystoke is seeking to build a 90-megawatt data centre at Woodlands in Buckinghamshire. The local council initially rejected the proposal, but the government “called in” the plans and overturned the decision. GAP and Foxglove argued the government’s decision was unlawful because it failed to properly consider the energy use and environmental impact of the planned facility.
In January 2026, the government admitted it had made a “serious logical error” in giving the go-ahead and conceded the case. Greystoke continued to contest the case until accepting in April that measures to mitigate environmental impact would need to be made binding via a contract with the council.
Foxglove co-executive director Rosa Curling said the case was a “step in the right direction”, but “far from the end of the story”.
A motion to the 2026 PCS public and commercial services union Scotland Nation Conference explains that the government reclassified AI data centres as “Critical National Infrastructure” to bypass local planning resistance.
“There needs to be a government framework on data centres,” Espley told Labour Research.
“We need to know how many data centres we really need, where and what for; how much environmental harm we think is a reasonable trade-off, in terms of water and energy use and reliance on fossil fuels; how any harms — that can’t be avoided — will be mitigated; and how it is going to fit with our carbon pathways and just transition. We also need a lot more transparency over the planning system.”
At present, data centres are in the same planning category as warehouses, so some application processes have only considered the energy used by the canteen and office, for example, rather than the equipment inside them “which can consume as much electricity as hundreds of thousands of households”.
Consequences
Espley added: “I think people are waking up to the consequences of quite a new and emergent threat and it’s important for organisations like ours, trade unions and others to educate our members and the public about the issues so that we can force good quality debate.”
And, he said: “If their businesses are adopting AI or generative AI is used, union reps can ask questions about what that does to the environmental impact of the business as a whole. How has it been assessed? Is it being reported on?”
AI is really worrying in terms of the impact it’s going to have on jobs and skills and “so many other aspects of working life” and “union members will also be really concerned about its impact on whether we have a liveable planet,” he said.
James says unions also have a role in critically examining “jobwashing” — the exaggerated claims AI data centre developers are making about the number of jobs they will create.
The March 2026 APRS paper Data centres and employment includes an analysis of a long-term data set of Virginia in the USA which found it took $33 million of investment into data centres to create one job.
This is around 400 times the cost of a job from IT investment that did not go into data centres.
Ethical implications
Motions to the PCS Scotland national conference noted that although the Scottish government’s approach to AI has been more cautious and risk-aware compared to the UK government, concerns about the social, environmental and ethical implications of widespread AI adoption remain.
The conference noted “approvingly” the PCS model agreement on AI but instructed the union’s Scottish executive committee to go beyond this.
It should work within the union, and negotiate with employers, to ensure there are impact assessments on the environmental consequences of using generative AI, within and outside the UK and devolved governments.
The carbon emissions and energy uses of external IT services should also be considered in departmental carbon footprint evaluations.
This is an edited version of an article that appears in the June 2026 issue of Labour Research.
