Majority of US’s new AI datacenters to be built on drought-hit land

Guardian analysis finds facilities to be built in some of the driest areas as outcry grows over water needed to power AI

A record-shattering drought has racked much of the US. But the artificial intelligence industry is pushing ahead regardless, with the majority of planned datacenters set to be built in drought-ridden locations, a Guardian analysis has found.

About two-thirds of upcoming datacenters, which typically require a large amount of water to operate, are set to be built in places that have been among the driest in the country over the past year.

Of 809 planned datacenters, 517 are in locations that have been in drought conditions throughout the past year, according to data from Cleanview and the federal government, which grades drought across four levels of severity. A similar proportion of existing datacenters are already situated in drought-affected areas.

More than 60% of the contiguous US is currently at varying stages of drought , the largest expanse for spring in modern records , with a particularly severe lack of rain and snow in the south-east and west desiccating croplands and raising fears of a disastrous wildfire season.

Scientists have determined that the climate crisis, caused by the burning of fossil fuels, is worsening the duration and intensity of droughts in the US.

But a stampede of new datacenters are adding extra demands via their hefty energy and water requirements. Large datacenters, some the size of small towns, can require up to 5m gallons of water a day, equivalent to the water use of up to 50,000 people, in order to provide cooling to arrays of humming networked computers.

Overall, the multiplying datacenters across the US are set to demand as much as 73 billion gallons of water a year by 2028, up from about 17 billion gallons in 2023. Each 100-word AI prompt uses up roughly one 500ml bottle of water due to the cooling needs of datacenters, researchers have estimated .