Big Tech’s AI Push Requires a Big Bath

There’s plenty to worry about with artificial intelligence: Will it put people out of work? Is it cheapening our art? Will humans eventually be subservient to all-knowing AI overlords?

Questions like that tend to drown out other nontrivial repercussions, like how much water it consumes. Water, you may have noticed, isn’t an abundant resource everywhere. Nor is it cheap. 

Have a Cool Drink

Whether it’s a car, computer, or nuclear reactor, machines need to stay cool or risk overheating. That’s why when crypto really took off, digital miners set up shop in cold places or communities with cheap energy bills. Air cooling systems — i.e. mega fans — are one option, but that gets expensive, especially in these days of high energy costs. If you’re a Big Tech firm with a data center that contains hundreds if not thousands of complex computers endlessly processing, developing, and storing information, you’re going to need something superior and cheaper — water.

Microsoft, Meta, and Google have all significantly boosted their water usage to cool their data centers amid the generative AI boom, the Financial Times reported over the weekend. In 2022, the latest figures available, Microsoft increased its water consumption 34% to roughly 1.7 billion gallons, Google by 22%, and Meta by 3%. And those increases are only likely to continue:

• AI demand is likely to drive up water withdrawal to between 4.2 billion and 6.6 billion cubic meters by 2027, according to academics cited by the FT. That’s about half the amount of water consumed by the UK’s 68 million people each year.
• It’s not like Big Tech has its own special access. They get their water from the same place we do: municipal watersheds. Shortly before OpenAI finished training its latest AI model, ChatGPT-4, one of its data center clusters in Iowa allegedly consumed 11.5 million gallons of water, or 6% of all the water used in the district, according to a lawsuit filed by residents. 

You Look Parched: Water consumption levels could be even higher than what’s being estimated, and researchers are calling on Big Tech for greater transparency. Katie Crawford, a research professor at the University of Southern California, told the FT that without better information it’s difficult to gauge AI’s environmental and societal impacts, especially “when many parts of the planet are experiencing deep and extended droughts, and fresh drinking water is already a scarce resource.” But there’s no time for those concerns when the world desperately needs bland high school essays, soulless clickbait articles, and images of Pope Francis in a puffy Balenciaga coat.