AI Briefing
KO

Annual Water Usage of US Data Centers Triples to 17 Billion Gallons

·2026.08.25 05:09

Key point

Driven by generative AI demand, water usage by US data centers has tripled since 2014 to reach 17 billion gallons.

Details

According to a report by the Congressional Research Service (CRS), water usage by US data centers in 2023 reached 17 billion gallons, an increase of approximately 3 times compared to 2014. This figure reflects data up to the point just before the full-scale expansion of generative AI infrastructure, and usage is estimated to have increased further since then.

Water Usage Structure and AI Impact

More than 80% of total water usage is indirect consumption occurring during electricity production, rather than direct cooling of the data centers themselves. AI systems require high-density hardware with higher power consumption than traditional servers, which simultaneously increases cooling demand and water usage associated with electricity generation.

The Cooling Method Dilemma

Avoiding evaporative cooling within buildings does not result in a "waterless" data center. Switching to air-cooled systems with high power consumption eliminates the burden of water usage for cooling, but water usage during the electricity generation phase increases, meaning the overall water footprint may not decrease.

Lack of Data Transparency and Regulation

Currently, the US federal government does not systematically track water usage by data centers, and detailed figures are often kept confidential due to municipal-level contracts. This has led to social conflicts, including lawsuits in water-scarce regions like California over disputes between data center construction and agricultural water use. Bills mandating water usage reporting and encouraging reuse have been introduced in Congress, but most remain in the early stages of adoption.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.