Discussions of the AI boom’s massive energy requirements have been making headlines for years, but the impact extends beyond the electric grid with respect to communities and the environment. Water usage in AI computing processes has been sidelined as a quieter conversation, but it is a tremendous bottleneck.
The AI boom’s water footprint continues to expand in regions from the East Coast to the Great Lakes, all the way to California. A mid-sized data center consumes as much water as a small town, with larger ones requiring up to 5M gallons of water every day – enough for a city of 50,000 people. In fact, if water use trends continue, U.S. data centers could require volumes comparable to New York City’s average daily supply of 1M gallons per day through 2030.
Direct water use to cool data center server rooms and chips is not the only issue at hand. Around 50% of the electricity used by U.S. data centers can be attributed to fossil fuel power plants, which use significant volumes of water to heat up steam to operate turbines. Not to mention, the millions of microchips that process information each consume thousands of gallons of water in the manufacturing process.