
Artificial intelligence is becoming an increasingly large part of everyday life for many people.
It's not just about asking ChatGPT questions; it's being built into websites, software, and programmes that we all use regularly. Even if you think you're actively avoiding AI by not using ChatGPT, or generative filters, or even TikTok trends, you're probably still using it. This reliance on the technology is also causing horrifying damage to our planet, and one of the scarcest - and most vital - sources we have.
Every single interaction with AI uses water.
Yes, AI is wasting incredible amounts of water, the very same stuff we all rely on to survive, and the statistics are horrifying.
Advert
According to the Government Digital Sustainability Alliance (GDSA), AI is forecasted to increase the amount of water used globally from 1.1bn to 6.6bn cubic metres by 2027.

That amounts to more than half of the UK’s total water usage.
ChatGPT originates from GPT-3, a large language model (LLM) which was simplified to create the AI we know now. This LLM needed a shocking 700,000 litres of water during the ‘pre-training’ phase.
Advert
While water covers 75% of the Earth's surface, only a tiny 0.5% of that is available fresh water.
How does AI use water?
There are three ways in which AI uses so much water:
- Used at data centres to keep the required computers cool
- Used indirectly for electricity generation
- Used during the manufacturing process for the hardware and servers they run on.
In a dream world, this water could just be cleaned and recycled over again. Unfortunately, it cannot simply be reused for drinking or consumption after cooling, because it becomes contaminated with dust, minerals, or chemicals. It can be treated and released back into local bodies of water or used for toilet flushing.
Advert

Why are people confused about AI using water?
Many people do not realise that asking ChatGPT about their dating life or to solve an argument is using up precious natural resources.
On X, one user wrote: "Today I joined a tiktok live. I made a reference to AI wasting water. The girl was confused. I offered to explain. She refused and asked ChatGPT instead."
Advert
On TikTok, one woman tried to do an explainer video about why it uses so much water, and the comment section was still full of people not understanding it.
One wrote: "I can literally use ai on my phone rn, how does it need water?"
A Reddit user asked why they can't use dirty water for the cooling, and one person replied: "Dirty water leaves deposits on the pipes, etc., which means corrosion, blockages, damage to the delicate electronics and infrastructure. People are always suggesting 'just use seawater!' but you might as well just use chicken soup or beef stew instead, it's ridiculously corrosive, salty, filled with fish poops and would quickly wreck the server farms."
Now you know.
Advert

What can be done?
Surely we're not all doomed?!
Well, there is some hope on the horizon, as people are working to come up with viable alternatives that are better for the planet. These include placing new data centres in areas that are not experiencing water stress or at risk of drought or flooding.
There is also hope that rainwater can be used, or changing data centres over to closed cooling systems, which would mean they reuse the fresh water for cooling and dramatically reduce their usage.
It is also hoped that many may be able to switch to air cooling, instead of relying on water.
However, it's not just water we need to think about. According to the International Energy Agency, a single request made through ChatGPT uses 10 times more electricity than a Google Search.
It's even more unforgivable when you work out that 1.1 billion people worldwide lack access to fresh water. Not only that, but a total of 2.7 billion people live with water scarcity for at least one month of the year.
The best possible thing you can do would be to reduce, or ideally stop, your reliance on ChatGPT, as it will be much better for the planet.
Topics: Artificial intelligence, Environment, Technology, World News, ChatGPT, Climate Change