AI feels clean and effortless. You type a question, and an answer appears instantly. But behind every AI tool sits a huge physical system: buildings full of computers using electricity, water and materials.
Here’s what the evidence shows, in simple, everyday language.
1. AI uses a lot of electricity
AI runs in data centres. These are giant buildings full of powerful computers that work all day, every day.
The International Energy Agency predicts that the world’s data centres could use twice as much electricity by 2030 as they do now, largely because of AI (IEA, 2025).
Training a single large AI model has been shown to use enough electricity to power around 120 homes for a year (Climate Impact Partners, 2025).
But what about one question to ChatGPT?
Recent analyses suggest:
-
One chat prompt ≈ running an LED bulb for about 2–3 minutes.
-
A longer chat = roughly the energy of boiling a kettle once.
A single question is tiny, but millions of questions every minute quickly add up (Epoch, 2025; De Vries, 2023)..
2. AI uses a huge amount of water to stay cool
Powerful computers get extremely hot. Many data centres use large amounts of clean water to cool them down.
Some facilities use millions of litres of water per day, similar to the daily use of a town with 10,000–50,000 people (EESI, 2025).
Research has shown that training big AI models can consume hundreds of thousands to millions of litres of freshwater once cooling is included (Li et al., 2023).
A simple way to imagine this…
If a town fills its swimming pool each week, a busy AI data centre might use that amount of water in a day or two.
3. AI requires physical machines, not “the cloud”
AI isn’t floating in the air. It relies on thousands of specialised chips made from metals like cobalt and lithium, which must be mined.
The United Nations Environment Programme notes that making a single small computer can require hundreds of kilograms of raw materials, and that mining for the minerals used in AI chips can damage ecosystems (UNEP, 2024).
When these chips become outdated, they turn into e-waste, which is difficult to recycle responsibly.
4. AI encourages more digital use, often without us realising
A normal Google search uses a tiny amount of energy. But a generative AI answer — like a ChatGPT paragraph — uses around 4–10 times more energy (CSS-IRL, 2024; Solve, 2025).
Why? Because instead of just finding information, AI also creates it.
Think of it like this:
-
A Google search = a sip of energy
-
An AI-generated answer = a full glass
One full glass is fine.
Billions of glasses a day? That’s significant.
So… Is AI “bad” for the planet?
Not automatically. AI can help the environment too — by predicting extreme weather, spotting pollution, monitoring forests, and helping buildings use less energy (UNEP, 2024; UNRIC, 2025).
The issue is scale.
The more we rely on AI for everything, the more electricity, water, materials and waste it produces.
What can we do to reduce AI’s environmental impact?
Even small choices make a difference.
1. Use AI intentionally
If you can write something quickly yourself, do that!
Avoid regenerating outputs repeatedly.
2. Write clearer prompts
Clear questions require less computer power.
3. Reuse outputs
Edit what you already have instead of generating lots of new versions.
4. Choose companies that use renewable energy
Some AI providers are moving to wind, solar and greener cooling systems.
5. Teach students digital responsibility
Help them understand that AI has a footprint and should be used thoughtfully.
AI is powerful and useful, but it is not impact-free.
Every chat, every image, every generation uses electricity, water and materials somewhere in the world. By using AI more thoughtfully, we can help make sure it becomes part of a greener future rather than a growing environmental burden.
References
Barker, C. (2024) Artificial intelligence and the environment: Taking a responsible approach. Jisc National Centre for AI blog, 18 September. Available at:
https://nationalcentreforai.jiscinvolve.org/wp/2024/09/18/artificial-intelligence-and-the-environment-taking-a-responsible-approach/
Barker, C. (2025) Artificial intelligence and the environment: Putting the numbers into perspective. Jisc National Centre for AI blog, 2 May. Available at:
https://nationalcentreforai.jiscinvolve.org/wp/2025/05/02/artificial-intelligence-and-the-environment-putting-the-numbers-into-perspective/
Climate Impact Partners (2025) The carbon footprint of AI. Available at:
https://www.climateimpact.com/news-insights/insights/carbon-footprint-of-ai/
CSS-IRL (2024) AI Environmental Impact Report. Available at:
https://cssirl.ie/ai-environmental-impact-report/
De Vries, A. (2023) ‘The growing energy footprint of artificial intelligence’, Joule, 7(10), pp. 2191–2194. Available at:
https://www.sciencedirect.com/science/article/pii/S2542435123003653
Environmental and Energy Study Institute (EESI) (2025) Data centers and water consumption. Available at:
https://www.eesi.org/articles/view/data-centers-and-water-consumption
Epoch (2025) How much energy does ChatGPT use? Available at:
https://epoch.ai/p/public/energy-chatgpt
Li, P., Yang, J., Islam, M.A. and Ren, S. (2023) ‘Making AI less “thirsty”: Uncovering and addressing the secret water footprint of AI models’, arXiv preprint arXiv:2304.03271. Available at:
https://arxiv.org/abs/2304.03271
Solve (2025) The environmental impact of AI: Complete guide 2025. Available at:
https://www.solve.co.uk/blog/environmental-impact-ai
United Nations Environment Programme (UNEP) (2024) AI has an environmental problem. Here’s what the world can do about that. Available at:
https://www.unep.org/news-and-stories/story/ai-has-environmental-problem-heres-what-world-can-do-about
United Nations Regional Information Centre (UNRIC) (2025) Artificial intelligence: How much energy does AI use? Available at:
https://unric.org/en/artificial-intelligence-how-much-energy-does-ai-use
