US datacenters tripled their water footprint in 10 years
US datacenters used around 17 billion gallons of water in 2023, three times the amount consumed in 2014, according to a Congressional Research Service (CRS) report published by the Library of Congress. The surge coincides with a boom in datacenter construction driven by the rise of generative AI since OpenAI's ChatGPT launched in late 2022, with new facilities housing power-hungry AI hardware that requires greater cooling capacity. The report warns the true figure is likely higher now, and highlights a lack of systematic federal oversight of the industry's water use, since data collection is largely left to state-level bodies.
The CRS found that direct water use at datacentres for cooling accounts for only about 2 percent of total US water consumption, while indirect use through electricity generation makes up more than 80 percent of a datacentre's overall water footprint. Around 97 percent of the water supplied to facilities comes from municipal drinking water systems, though water service agreements with operators can restrict disclosure of usage figures. The report notes cooling technology choices involve trade-offs between water and energy use, with evaporation-based systems more energy-efficient but thirstier than air-cooled alternatives, and points to ongoing transparency concerns, including a recent lawsuit alleging Amazon Web Services misrepresented its water use in Northern Virginia.
- US datacentre water use tripled to 17 billion gallons between 2014 and 2023
- AI boom driving construction of power-hungry, water-intensive facilities
- Federal oversight of datacentre water use remains limited and fragmented
New here? Start with this
Datacentres are the physical hubs of computers that store data and run the software behind websites, apps and online services. Running them generates heat, so operators use water-based cooling systems to keep the equipment from overheating, alongside water used indirectly to generate the electricity they consume. A US government research body, the Congressional Research Service, has been examining how much water this all uses and how that has changed over time.
The rise of generative AI tools, following the launch of ChatGPT in late 2022, has driven a wave of new datacentre construction, since AI computing hardware uses more power and needs more cooling than older systems. This matters because water is a shared resource, and much of the supply to these facilities comes from the same municipal systems that serve homes and businesses, raising questions about competing demands, particularly in drought-prone areas.
Oversight of datacentre water use in the US is patchy, with data collection largely handled by individual states rather than the federal government, and companies are not always required to disclose how much they use. Large technology firms including Amazon operate major datacentre clusters, such as in Northern Virginia, a region that has become a focal point for scrutiny of the industry's water and energy demands.
AI Americas Art Celebrity Cricket Culture Entertainment Sport Technology World