Every time AI tools are prompted, it correlates to two cups of water being used to cool the data center from which the information is derived. Big tech, particularly Amazon, Google and Microsoft, is tapping into water sources in some of the driest regions across the globe, according to a recent investigation by SourceMaterial and The Guardian.
The investigation revealed water-scarce regions around the world where data centers exist or are planned. The tech giants’ thirst for water may be insatiable as the demand for AI increases.
Experts have been sounding the alarm for years, and some tech companies are exploring ways to develop greener data centers.
“Extrapolate AI use globally, and it’s clear how water systems are taxed,” said Andrew Collier, an executive with Veolia Water Technologies & Solutions. Mr. Collier said that Microsoft’s environmental report revealed that its water consumption jumped 34% from 2021 to 2022, reaching nearly 1.7 billion gallons. Google’s consumption went up nearly 22% percent over the same time span, to 5.56 billion gallons.
AI adoption is increasing, as 9 in 10 companies will adopt generative AI by the end of this year, according to a study by Gartner. That means more demand for AI, and more demand for data centers, and water.
“Unchecked water usage by power-hungry data centers could exacerbate the already dire water crisis worldwide,” Mr. Collier said. Veolia is actually using AI to come up with solutions to the problem, through its GreenUp initiative, to optimize water and energy consumption, as well as “waste sorting and recycling and strengthening predictive maintenance.”
Data center cooling towers and humidifiers are sucking up water by the millions of gallons and taxing municipal water supplies, Veolia said.
According to the SourceMaterial investigation, in the U.S., Arizona, Nebraska and Wyoming were identified as places where data centers are located in water scarce zones. Globally, Spain, Israel, Kuwait, the United Arab Emirates, India, China, South Africa, Chile, and several other countries were identified.
Amazon is proposing three data centers in Aragon, Spain, next to existing Amazon data centers, and SourceMaterial estimates that when the new data centers are up and running, they will use enough water to irrigate more than 500 acres of corn for a year, one of the region’s main crops.
“Data center locations are often industry secrets. But by using local news reports and industry sources Baxtel and Data Center Map, Sourcematerial compiled a map of 632 centers– either active or under development– operated by Amazon, Microsoft and Google. It shows that those companies plan to increase the number of data centers they operate worldwide by 78%. With cloud computing already straining current infrastructure, advances in artificial intelligence are set to drive further demand as companies seek capacity to train models like ChatGpt. Overall, there are 11,800 operational data centers worldwide, with 5,426 located in the United States, according to estimates from industry intelligence firm Cloudscene. Between them, the three tech giants are planning construction on every continent except Antarctica,” shares Sourcematerial.
Just why those areas are desirable, comes down to a sort of Goldilocks scenario, picking locations that are just right. “It’s no coincidence they are building in dry areas,” said Lorena Jaume-Palasi, founder of The Ethical Tech Society, as “centers have to be built inland, where low humidity reduces the risk of metal corrosion, while seawater also causes corrosion if used for cooling.”
The fact that 70% of the earth’s surface is covered by water is misleading when it comes to water scarcity, because freshwater, the kind that we drink, bathe in, and use to irrigate fields, is incredibly rare, according to the World Wildlife Federation. “Only 3% of the world’s water is freshwater, and two-thirds of that is tucked away in frozen glaciers or otherwise unavailable for our use. At the current consumption rate, this situation will only get worse.”
Many companies are exploring alternatives to using massive amounts of water for cooling, including dry cooling and liquid cooling inside a server, and immersion cooling, or floating data centers that are cooled by the water they sit upon.
