B6 · Responsible and Safe Use of AI

AI and the Environment

AI affects the environment in two directions. Training and running AI systems uses electricity, water for cooling and materials for chips, which adds to carbon emissions and resource use. At the same time, AI can help protect the environment by improving forecasts, saving energy and monitoring nature.

6 min Beginner Jump to the activity

What you need: Nothing. This page starts from the beginning, with no programming and no formulas.

Definition

Asking a chatbot a question feels weightless. Somewhere, a building full of computers is using electricity and water to answer it. How much?

AI affects the environment in two directions. Training and running AI systems uses electricity, water for cooling and materials for chips, which adds to carbon emissions and resource use. At the same time, AI can help protect the environment by improving forecasts, saving energy and monitoring nature.

In simple words

A tractor can plough a field much faster than oxen, and it also burns diesel. Whether it is good for the farm depends on how it is used and how much fuel it takes. AI is a powerful engine too. It can help save energy and forests, and it also consumes a great deal of electricity and water itself.

Why it matters

AI's energy use is growing fast, in a world trying to cut emissions. Understanding both its costs and its benefits helps people use it wisely and ask companies to build it efficiently.

Key terms

Data centre
A building full of computers that store data and run online services, including AI.
Carbon footprint
The greenhouse gases released by an activity.
Training
The long, energy-heavy process of building a model from data.
Inference
Using a trained model to answer a request. It happens billions of times a day.
Renewable energy
Electricity from sources such as sunlight and wind that do not run out.
E-waste
Discarded electronic equipment, which contains harmful and valuable materials.

Explained in detail

Where the energy goes

AI runs in data centres: large buildings full of servers and specialised chips, running day and night. Energy is used in two stages. Training a large model keeps thousands of chips busy for weeks or months. Then, every time anyone uses the model, it runs again to produce an answer; this is called inference. Because popular tools answer billions of requests, inference now accounts for most of AI's energy use. Generating an image or a video takes much more computing than producing a short text answer.

How big is it?

The International Energy Agency estimated that data centres of all kinds used about 415 terawatt-hours of electricity in 2024, around one and a half per cent of the world's total, and projected that this could more than double by 2030, driven mainly by AI. That is still a small share of all electricity, but it is growing faster than almost any other use, and it is concentrated in particular places where it can strain local power grids. Where the electricity comes from coal, the emissions are high; where it comes from sun and wind, they are much lower.

Water and materials

Chips get hot, and many data centres cool them by evaporating water. Researchers have estimated that a large AI model can consume roughly half a litre of fresh water for every few dozen questions, depending on where and when it runs. In dry regions, this competes with water for farms and homes. AI chips also need mined metals and complex factories, and old servers become electronic waste that must be recycled safely.

The other side: AI for the planet

  • Better forecasts of floods, cyclones and heat waves, giving people time to prepare.
  • Smarter electricity grids that balance supply from sun and wind with demand.
  • Cutting energy waste in buildings and factories; in 2016 one company reported cutting the energy used to cool its data centres by up to 40 per cent with AI.
  • Monitoring forests, oceans and wildlife from satellite images and sound.
  • Helping scientists design better batteries, solar materials and crops.

Using AI with care

Companies can reduce the impact by building smaller, more efficient models, running data centres on renewable energy and reporting their energy and water use openly. Individuals can help too: use AI when it truly helps, prefer simple search for simple facts, avoid generating dozens of images or videos for fun, and keep devices longer before replacing them. The aim is not to stop using AI, but to use it where its benefits outweigh its costs.

How it works

AI's costs to the environmentAI's benefits to the environment
Electricity for training and answering requestsFlood, cyclone and heat-wave forecasts
Water to cool data centresSmarter grids for solar and wind power
Mining and manufacturing for chipsDetecting deforestation and illegal fishing
Electronic wasteDesigning better batteries and materials

Try it yourself

The journey of one question

You ask a chat assistant to suggest a name for your pet.

Cost or benefit to the environment?

Decide how each use of AI affects the planet.

  • A model that warns villages of floods days ahead

  • Training a very large model on coal-powered electricity

  • AI that reduces a factory's energy waste

  • Satellite monitoring that detects illegal logging

  • Generating fifty images for fun and deleting them

Common mistakes

  • Thinking digital means weightless. AI runs on real electricity, water and metal.
  • Thinking AI is only harmful to the environment. It also helps protect it.
  • Using the heaviest tool for the smallest task.

Check yourself

Question 1

1. Why do data centres use water?

Question 2

2. What is inference?

Question 3

3. Which is an environmental BENEFIT of AI?

Summary

  • AI uses significant electricity, water and materials, and its use is growing fast.
  • It also helps the planet through forecasting, efficiency and monitoring.
  • Efficient models, clean energy and thoughtful use reduce the cost.

Finished reading?

Sources: International Energy Agency: Energy and AI · UNESCO: Recommendation on the Ethics of Artificial Intelligence