The environmental impact of AI

What is AI? 

Artificial Intelligence (AI) is technology that supports computer systems to reason, learn and problem-solve without step-by-step instructions (a bit like a human).

Companies have been developing AI for decades. Today "generative" AI, the newest form, appears in everyday tools such as predictive text, streaming recommendations and voice assistants – often working unnoticed in the background.

Generative AI is good at understanding normal language and learning from feedback. That’s why it can generate images and videos or produce essays and reports. But it also uses a lot of resources and power.

And while AI technologies might be used to deliver things like healthcare services more efficiently, they’re also being used to automate weapons and drive new oil and gas extraction – which can be devastating for people and planet.

What are AI data centres?

AI relies on data centres. A data centre is basically a building filled with computer servers. They can be used for a range of purposes, from data storage (if you have ever saved a file in "the cloud", it'll be in a data centre somewhere) to running real-time processes. And they can vary hugely in size and set-up, with big impacts on the amount of resources they use. 

How many AI data centres are there in the UK?

Newer forms of generative AI usually use more computing power – and more powerful data centres. This has led to an explosion in overall data centre numbers, as well as an increasing number of "hyperscale" (really big) centres. And that explosion shows no signs of slowing.

• In the year 2000, there were only about 50 data centres across the UK.
• By 2024, there were 266 UK data centres.
• As of early 2026, there are at least 450 UK data centres in existence and over 100 more proposed to planning committees across the UK.

How much water do AI data centres use?

The data centres that power AI need to keep the computers they contain cool. While some centres use waterless or "closed" systems that limit water use (and significantly more electricity in the process), others use massive amounts of water.

A data centre might use anywhere from 2 million to 19 million litres of water a day for evaporative cooling – equal to 14,000 and 132,000 UK households.

This usage is really worrying at a time when we're seeing more frequent heatwaves in the UK and worldwide. Half of England was in drought in 2026. Nearly 3,000 people died from heat-related issues across the UK in just 2 months. Yet areas that are already suffering water stress or are projected to be water stressed by 2040 are set to host over 80% of proposed water-intensive data centre developments.

Data centres don’t just make the effects of heatwaves worse. When it’s hot, just like people, data centres need even more water and energy to keep cool. Data centres then compete with people for the extra electricity for air conditioning. This can lead to interruptions in electricity supply or "brown outs".

In July 2022, two separate data centres serving an NHS trust failed during a heatwave. This affected 3 hospitals and massively disrupted patient care, while the trust had to spend £1.4 million to fix the problem.

How much energy do AI data centres use?

Data centres often rely on masses of energy to run computer and air-con. The buildings and the computer components use up huge amounts of metals and rare minerals.

A single large data centre – one using over 150 MW of electricity – needs about the same amount annually as half a million average UK homes. Data centres: 

• accounted for approximately 2.5% of UK electricity consumption in 2024, with demand projected to increase 6-fold by 2030
• used 1.5% of all electricity consumed around the world in 2024
• could use more electricity globally than the whole of Japan does each year by 2030

And none of these figures include the energy needed to build data centres themselves or produce the tech inside them.

More big AI data centres are planned

Another 315 data centres were registered with electricity regulator Ofgem earlier this year. On average, each of these proposed new data centres could use as much electricity as over a million homes. And because the number of data centres is increasing faster than new renewables, this means using more fossil fuels like gas.

No one (not even the government) knows how many data centres we might have across the UK in 5 or 10 years’ time or how much energy a new data centre in 2036 might use. Projections suggest that data centres could be using 11% of the UK's electricity in 2035. But we do know homes are already competing with data centres for access to power in areas like West London. And that shifting our heating and transport systems away from gas, petrol and diesel to tackle climate breakdown will mean the UK needs to produce twice as much electricity by 2050, if it's to cope with new data centres too.

If demand from new data centres keeps propping up fossil fuel generation and hogging new renewable power, they’ll keep standing in the way of cutting emissions and building new green homes.

Do AI data centres cause pollution?

Communities living near data centres have reported a range of problems, including air, noise and heat pollution.

• Poor air quality and noise pollution are often linked to on-site fossil fuel generation – as is planned for the Wapsey's Wood data centre in Buckinghamshire – or to testing and operation of the backup generators these centres rely on to keep the computers on during power cuts.
• Analysis from the US has estimated that air pollution from on-site gas generation at a single data centre in Northern Virginia could result in £42 to £79 million in annual health-related damages.
• In the 2026 heatwave, research has found that data centres can increase temperatures next door by 2 to 9 degrees.
• The Climate Change Committee calculates that a 2 degree rise in global temperatures would be expected to lead to 2 to 3 times more people dying from heat-related issues across the UK each year.

The potential human and environmental impact of data centres goes beyond the neighbourhood of each individual centre. More data centres means more demand for construction materials to build them and minerals to create the computer hardware they house. More than 70% of the world’s cobalt – essential for AI hardware like GPUs and batteries – is sourced from the Democratic Republic of Congo. The cobalt supply chain is strongly linked to child labour and unsafe working conditions, as well as local environmental harm.

And as AI tools use increasingly high-powered chips, the data centre boom could undermine or reverse efforts to cut electronic waste and move to a more circular economy. The UK already produces over a million tonnes of e-waste each year. Much of this is exported, often illegally, to West Africa and elsewhere, with harmful knock-on impacts for local people and the environment.

What's bad about AI?

AI can be used to try and limit pollution or increase renewable energy use – and AI companies often talk about this potential benefit. But AI tools are also being used by the fossil fuel industry to find and extract more fossil fuels than was previously possible. The Enabled Emissions campaign has calculated that AI tools provided by Microsoft to fossil fuel companies Exxon and Chevron led to 57.3 million tons of carbon emissions in 2020.

AI can also spread misinformation and disinformation about subjects like climate change. At a time when increasing numbers of people get their news through social media and many turn to chatbots to ask who to vote for, this could have big implications for democracy. Just like people, AI tools can get facts wrong or confidently make things up, and because AI systems still need to be "trained" on existing human-written texts, they can mirror the assumptions, biases and priorities of their creators.

Even more worryingly, AI can be used to intentionally mislead people – because it makes creating convincing false information easy, fast and free. In the UK, AI-generated images have been linked to the 2024 Southport riots, while a white nationalist rapper has been shown to be an AI-constructed fake paid for by a far right group. In the US, an AI-generated video showing a protestor "harassing" a US national guardsman during a protest was viewed over 40 million times. And in Australia earlier this year, research uncovered hundreds of AI-generated posts claiming to show real life politicians.

Are communities resisting AI data centres?

Opposing this, there's a growing global movement resisting dirty data centres and demanding that communities, the climate and nature are prioritised ahead of big tech.

In the UK, hundreds of people attended a public meeting in opposition to a massive data centre planned for the countryside near Torrington in Devon. Protesters took to the streets to oppose plans to replace a community asset with a data centre instead of affordable housing in East London. And the No Leeds Data Centre campaign has attracted tens of thousands of signatures to a petition against 3 hyperscale datacentres. In the US, 42 states saw coordinated protests against data centres on a single day in July. Indigenous communities and campaigners in Brazil, Chile and Uruguay have challenged data centres that threaten to take water and land needed by local people.

Community action is having an impact. The Danish energy operator has announced a temporary ban on new data centres. Cities, from Edinburgh in the UK to Oakville in Canada, are considering and enacting local bans. Many US states have already done the same. And reports in the US suggest that public protests blocked or delayed at least 75 individual data centre projects between January and March 2026.

It’s not just protests and petitions either. An increasing number of communities are taking legal action against data centres that threaten to pollute communities or compete with them for limited water and energy supplies. Research by Friends of the Earth Ireland has found that the rapid expansion of data centres added an average €360 (about £309) to household electricity bills, so together with Client Earth and Friends of the Irish Environment they’re taking the energy regulator to court for putting data centres before people and climate. In the UK, Foxglove and Global Action Plan stopped a planned Buckinghamshire data centre from going ahead without binding environmental commitments being agreed. And in summer 2026, campaigners in France successfully challenged the approval of a data centre in court, on the grounds it posed a risk to people and the environment.

How can you defend your rights?

Growing resistance movements around the world show real concerns about how AI is being developed and used. People clearly want to be part of the conversation.

Right now, technology companies are running the show, while the government plays catch-up and communities are left behind. We need to flip this situation and put communities at the forefront of decision-making.

That means thinking about why we’re building all these data centres, what needs they’re serving and asking how they're making people’s lives better. It means thinking about what the local consequences might be from each new data centre, and what tens or hundreds of new data centres could mean for our water and energy supplies. And it means considering who gets the benefits or profits from data centres – and who feels the negative impacts.