Questioning how much water AI is actually using
This article questions the significant water consumption of AI-powered data centers, using Google's facilities in The Dalles, Oregon, as a primary case study. It highlights concerns about the immense amount of freshwater used for cooling and evaporation, the lack of corporate transparency regarding usage data, and the potential environmental impact. Community advocacy groups like Columbia Riverkeeper are raising awareness and calling for greater accountability and regulation regarding data center water usage.
Mary Hahm, a freelance journalist based in the Columbia River Gorge, explores the significant environmental impact of AI-powered data centers, specifically focusing on their substantial water consumption. Prompted by her own extensive AI use, Hahm investigates Google's data centers in The Dalles, Oregon, as a case study, noting the visible steam plumes from the massive facilities.
The article highlights that large data centers can consume up to 5 million gallons of water daily, with projections indicating AI-supporting data centers could collectively use 720 billion gallons annually across the U.S. by 2028. Google's operations in The Dalles reportedly withdrew 461.1 million gallons in 2024, consuming 361.4 million gallons through evaporation, a significant loss of freshwater from the Columbia River system. Hahm criticizes the lack of transparency from companies like Google regarding precise water usage data, often citing it as a trade secret, which exacerbates public concern.
Hahm emphasizes the critical need for corporate transparency and robust community advocacy to hold tech companies accountable for their environmental footprint. She points out that organizations such as Columbia Riverkeeper are actively challenging data center development along the Columbia River. While Google defends its water stewardship projects, claiming to replenish a portion of its freshwater consumption, Hahm questions the actual effectiveness and calls for more detailed, openly published data. The article concludes with an optimistic view on future technological advancements like direct-to-chip cooling, which could reduce water demand, but stresses the immediate necessity for increased regulation and caution to prevent irreversible ecological damage.