The True Cost of AI: What Every Teacher Should Know

Every time you ask an AI chatbot a question, something happens that you cannot see. Somewhere, in a large building full of servers, computers wake up, process your request, and generate a response. These are the data centres that you keep hearing about, obviously they use electricity, the servers get hot and water is used to cool them down. As a teacher – I think it’s important you understand the controversy surrounding this.

Of course, none of that process is visible in the chat window. None of it is mentioned in the marketing. And the scale of it, according to a new United Nations report published in June 2026, is significantly larger than most people realise.

For teachers thinking about climate education, this needs to be understood – it matters. AI is now part of daily life for most of our pupils (and us, although we may not realise it!) The environmental cost of AI is one of the most immediate and personal climate connections available to us – and almost none of our pupils are aware of it.

What the UN report actually says

The United Nations University Institute for Water, Environment and Health published its report on the environmental costs of AI in June 2026, timed for World Environment Day. The numbers are striking.

The scale – by 2030

AI could consume 3% of the world’s electricity – double its current share.

AI’s carbon emissions could equal the entire carbon footprint of the United Kingdom.

AI could use more water for cooling than the annual global drinking water requirement of the entire human population.

Global data centre electricity use could reach 945 terawatt-hours – roughly equivalent to the annual electricity consumption of Japan.

Source: UNU-INWEH, Environmental Cost of AI’s Energy Use, June 2026

To put some of those numbers in classroom terms:

Training GPT-4 required approximately 600 million litres of water – enough to fill 237 Olympic swimming pools, or supply the annual drinking water needs of around 81,000 people.
A single AI chat session uses significantly more energy than a standard Google search – estimates vary, but the order of magnitude difference is consistent across studies.
If data centres were a country, their electricity consumption would already make them one of the largest energy users in the world.

One finding in the report is particularly important for classroom discussion: making AI more efficient does not necessarily reduce its total environmental impact. The report invokes the Jevons paradox, the economic principle that when something becomes more efficient and cheaper to use, people use more of it, and total consumption rises. More efficient AI means more people using more AI more often. The footprint grows.

Worth noting: The report also found that the benefits and burdens of AI’s expansion are highly unequal. The strategic advantages flow to wealthy nations and large corporations. The environmental costs, that is water stress, land use, energy demand, are often borne by communities that have the least to gain.

Why this matters for climate education

AI sits at an interesting intersection for climate teachers. It is something pupils use daily and feel ownership over, and it is also something they rarely examine critically. And the environmental cost of it is almost entirely invisible in normal use.

That invisibility is instructive in itself. We talk a lot in climate education about the hidden carbon in our consumption – the emissions embedded in the things we buy and use that we cannot see at point of purchase. AI is perhaps the purest example of this, tha chat window is instant but the data centre consuming the electricity and water is thousands of miles away.

There is also a justice dimension. As with so much of the digital economy, the infrastructure that makes AI possible, namely the the mines for rare earth minerals, the manufacturing of chips, the data centres built where land and energy are cheapest, ends to be located in places with less regulatory power to push back on its environmental impacts. The benefits of AI accrue disproportionately to wealthy users in wealthy countries. The costs are distributed elsewhere.

And there is the question of what AI is actually being used for. Using an AI chatbot to generate a shopping list uses the same infrastructure as using it to model drug interactions or analyse climate data. The energy cost does not reflect the value of the use. That asymmetry is worth exploring with older pupils.

AI and teachers: What to do with it in school

This is not an argument for banning AI or telling pupils their ChatGPT habit is destroying the planet. That kind of messaging tends to produce guilt and defensiveness rather than critical thinking.

It is an argument for making the invisible visible, e.g. for helping pupils understand that the tools they use have a material cost that is not included in the interface, and that understanding that cost is part of being a thoughtful digital citizen.

Some practical approaches:

Look at how AI companies describe their environmental impact. What do they say? What do they not say? Who funds the research they cite?
The AI prompt, the data centre, the energy grid, the water source, the community near the data centre — map the chain. Where does the cost land? Who decided it would land there?
Do more efficient technologies mean less environmental impact? Use AI as a case study for why the answer is often no.
What could individuals do? What could companies do? What requires policy? What is the difference between individual responsibility and systemic change?
We have produced a free classroom activity on the true cost of AI, suitable for upper primary and lower secondary PSHE. Download it below.

A note on honesty

I am aware that I am writing this on a platform that uses AI, and that you may be reading it having used AI to help with your planning today. That is not a contradiction that needs resolving before you can engage with this topic. It is part of the topic, and part of climate education and living like a role model.

We use things without fully knowing their cost. That is true of almost everything in a modern economy. The point of education is not to produce guilt about that, but awareness, critical thinking, and the capacity to make more informed choices and to push for systemic change.

The true cost of AI is not a reason to stop using it. Howver, it is a reason to think about how we use it, who bears the cost, and what kind of infrastructure we want to build for the future.

That is a very good conversation to have with young people.

————

The free Ground Level lesson plan on the true cost of AI is available to download from the notyetzero Ground Level page. For climate education support, training, or curriculum development, get in touch.