The explosion of generative AI has caused a global boom in the construction of energy-hungry data centres with associated fossil fuel power expansion, as well as concerns about the strain they may place on clean energy resources, grid capacity and water use.
The recording above is from our webinar on 17 February 2026 - watch it as an introduction to the issues and read on to learn more.
Oliver Hayes is Head of Policy & Campaigns at Global Action Plan, which recently won a legal case against the government approving a hyperscale data centre in Buckinghamshire without properly considering the energy use and environmental impact.
Anne Alexander is a researcher and journalist covering the social, ecological and political impacts of AI systems, currently working on a forthcoming book on the AI Arms Race
James Meadway is an economist, host of the weekly economics podcast, Macrodose.
Campaign dates and resources
Global Action Plan map of planned datacentres
Stop Dirty Data Centres days of action 27 and 28 February
E-action - Ask your MP to say no to Drax's data centre bid
Pull the Plug campaign - London march 28 February
Find out more
If you'd like to follow up issues raised in the webinar, below are some useful links to further information, and answers to questions in the discussion and Zoom chat.
Useful summary: As a quick overview, this from the Grantham Institute is the most succinct we've found on AI and climate/environmental issues.
Video of another recent webinar: this webinar, hosted by Fuel Poverty Action, provides more information on AI issues including Drax's data centre proposals with a speaker from the Stop Burning Trees coalition.
AI as an economic bubble: There is little evidence that the demand for AI exists which could justify the massive investments we are seeing, a trillion-dollar game of corporate chicken.
This AI 'arms race' has serious climate and environmental consequences. The electricity of a typical hyperscale data centre is around that of a town of 30-40,000 people, running 24 hours a day. They are putting strain on electricity grids, but direct fossil fuel power is being built as a quicker alternative to grid connections.
Critical minerals are also needed for data centre construction, exacerbating geopolitical tensions. The flip side of this will be the increased generation of e-waste.
The greatest climate harm is caused by AI enabling fossil fuel extraction: While most of the discussion focuses on the carbon and water footprint of data centres, this research by Global Witness found that the emissions enabled by just two of Microsoft’s contracts, with Exxon and Chevron, in 2020, amounted to 57.3 million tons of carbon. This is over three times Microsoft’s own emissions in 2023, including all data centres.
US data centre boom: Driven by data centre expansion, the US is at the forefront of a global push for gas. The gas projects in development in the US will, if all completed, cause 12.1bn tonnes in carbon dioxide emissions over their lifetimes, which is double the current annual emissions coming from all sources in the US. Examples are Colossus 1 and 2, powering Musk's xAI Grok, in Memphis Tennessee and neighbouring Southaven, Mississippi (activists captured video footage of the pollution from gas turbines). Air pollution impacts are significant and the location of data centres often follows existing patterns of environmental racism.
Water demand: Data centres are heavy consumers of water for cooling and energy generation. They are often built in regions where water resources are limited, competing directly with communities for water use, for example Mexico, where climate change has caused severe drought (Al Jazeera video)
UK government push for AI: Last year the government announced that its plan to "turbocharge" AI would lead to "a decade of national renewal", "mainlining AI into the veins of this nation" with AI growth zones and £14 billion government investment. A £31 billion "tech prosperity deal" was agreed with the US, which has since stalled with US seeking concessions from the UK on the digital services tax, online safety and food safety regulations.
Planning fast-tracked: Data centres have already been designated as 'critical infrastructure' (national assets that are essential for the functioning of society) in 2024. New regulations have just allowed developers of large data centres to choose to have their application decided by national government as a Nationally Significant Infrastructure Project, instead of being decided by local authorities. A consultation on a National Policy Statement about how these decisions will be made is expected soon, and it is important that campaigners engage with it.
UK climate targets at risk: Labour has set a target of building at least 6GW of new data centre capacity by 2030 – trebling the country’s current capacity, which would require enough power to supply a city six times the size of Birmingham. They have not explained how this could be compatible with the UK's climate targets, or the goal for the UK electricity to be supplied by clean energy by 2030. Developers’ own figures indicate that just ten of the largest data centres in planning or construction would cause annual emissions equivalent to 2.7 million tonnes of CO2. This would effectively wipe out the carbon savings expected in 2025 from the switch to electric cars (2.9m tonnes)
Nvidia chief executive Jensen Huang said that the UK will need to burn gas to power AI, and Conservative and Reform politicians have demanded that climate targets be abandoned to enable AI expansion.
Hidden impacts and inadequate planning consideration: Global Action Plan and Foxglove succeeded in getting the government to withdraw its approval for a data centre in Buckinghamshire because they had not considered the climate or environmental impacts.
A planned data centre in Northumberland submitted an estimate of water consumption from its cooling system, which recycles water - but not the water used for the energy generation to run this more energy-intensive system, making the true figure for water consumption around 50 time higher. Most if not all data centres will have back up diesel generators on site, and these will need to be regularly tested. The energy and air pollution impacts of this may not be properly considered.
Local campaigns have been successful: For example an Edinburgh data centre was refused planning permission after local campaigning.
Burning trees for AI: Drax power station has submitted a bit for an AI data centre - find out more
What will be the impact on jobs? It is hard to predict and the impact of AI will fall much harder on some sectors than others - but it is clear that data centres themselves will not be a major source of local job generation. Nor is there much evidence so far for the claimed AI economic transformation (with one report claiming 95% of businesses adopting AI hadn't found a significant benefit).
Can AI be used to reduce emissions? A recent review found most claims of climate benefit relate to ‘traditional’ AI, which has a much lower environmental impact than consumer generative AI tools - and even for these, there was a lack of published academic research to back them up.
Can data centres be made more efficient by using the waste heat, for example for homes? There are two problems with this. One is that it can lock in a reason to keep the data centres running all the time, rather than looking to reduce energy use. The other is that converting energy multiple times is an inefficient use of energy compared to efficient home heating. In the UK, we don’t have the kind of district heating that makes this sort of project work, although there are a couple of pilots.