Tech giants like Microsoft, Google, and Meta have ramped up their water usage to cool their data centres, which have been used more since the artificial intelligence (AI) boom.
It has sparked environmental concerns among academics, with experts predicting that by 2027, AI demand could lead to water withdrawal of between 4.2bn and 6.6bn cubic metres of water – nearly half of the UK’s annual consumption.
According to academics from the University of California, Riverside, the issue has “largely remained under the radar” so far and it is a “critical time to uncover and address AI models’ secret water footprint”.
They stated that training OpenAI’s GPT-3 can directly evaporate 700,000 litres of water.
In 2022, Microsoft, Google, and Meta saw their water consumption surge by 34 per cent, 22 per cent, and three per cent, respectively.
Apple and Amazon were not mentioned in the study. In Apple’s latest environmental progress report it said it has reduced cooling water usage at its data centres by 60m gallons annually.
The concerns are growing as tech giants race to launch generative AI-powered products, which rely on energy-intensive large language models such as Chat GPT that need colossal server farms cooled by water.
Another assessment shows that the hundreds of millions of questions asked on Open AI’s platform require the equivalent energy of 33,000 US households.
And Alex de Vries, a PhD candidate at Vrije Universiteit Amsterdam, has warned in a recent paper that AI’s energy consumption will continue to rise alongside its increasing demand, irrespective of efficiency gains made to offset it.
OpenAI said: “We recognise training large models can be water-intensive, and that is one of the reasons we are constantly working to improve efficiencies.
“We also believe that large language models can be helpful in accelerating scientific collaboration and discovery of climate solutions.”
Microsoft said that “currently, AI compute accounts for only a fraction of the electricity used by data centres, which collectively use about 1 per cent of global electricity supply. How much this increases and how AI growth affects the global race to net zero will depend on many factors.”
Google declined to comment.