California considers data center water disclosure amid debate on consumption
OpenAI CEO Sam Altman's claims about low water consumption by AI queries are being disputed by experts who cite a lack of transparency from tech companies. Amidst rising water demands from data centers, California lawmakers are advancing bills that would mandate water usage disclosure for data center operators and developers. These proposals aim to provide more clarity on the industry's environmental footprint, especially as local communities increasingly clash over new data center projects.
OpenAI CEO Sam Altman has asserted that artificial intelligence poses no significant water problem, claiming it takes 38,000 ChatGPT queries to consume as much water as a single almond in California, and comparing a data center's water footprint to that of an ordinary office building. These claims come as Governor Gavin Newsom considers signing legislation that would compel data centers to disclose their water consumption.
However, experts like Shaolei Ren, a professor at UC Riverside, and Michael Kiparsky, director of the Wheeler Water Institute at UC Berkeley, contend that Altman's comparisons are nearly impossible to verify due to the limited public data released by tech companies. They highlight the numerous variables, such as location, temperature, cooling technology, and computational intensity, that complicate any simple water usage statistic.
In response to this data gap, Assemblymember Diane Papan (D-San Mateo) has authored two bills that have passed the Legislature and await the Governor's signature. One bill would require data center operators to disclose water sources and consumption for business permits, while the other would prevent local governments from approving new projects without developers revealing water usage plans and covering infrastructure upgrade costs. Governor Newsom had previously vetoed a similar measure, seeking more clarity on its business impacts, but pressure from communities clashing over proposed data center projects in California is mounting.
Federal research indicates a sharp increase in data center water consumption, with one study finding direct usage grew from 5.6 billion gallons in 2014 to 17 billion gallons in 2023. Ren's research further suggests that data centers significantly strain water systems during peak demand, drawing from treated municipal supplies and adding substantial daily water demand comparable to major cities. While dry cooling systems could reduce water use, they often lead to higher energy consumption, creating a complex trade-off.