Gov. Gavin Newsom is considering legislation that would require data centers to reveal their water consumption as OpenAI chief executive Sam Altman disputes the significance of artificial intelligence's water demands. Altman stated on a podcast that 38,000 ChatGPT queries consume the same volume of water as the production of one almond in California. He added that a large modern data center uses water at levels comparable to an office building for sinks and restrooms.
State lawmakers recently approved two bills sponsored by Assemblymember Diane Papan that aim to increase transparency around data center operations. One measure would obligate operators to disclose water sources and consumption when applying for business permits. The second would prevent local governments from approving new facilities unless developers provide water plans and cover costs for any necessary supply upgrades.
Newsom rejected a similar proposal last year, citing concerns about impacts on businesses and technology users. Technology companies and business organizations have opposed the current bills as well. According to LAist, the measures reflect growing pressure amid local resistance to new data centers across the state.
Experts note that limited public data makes it difficult to confirm claims about data center water use. Shaolei Ren, a professor at UC Riverside, pointed to variables such as facility location, outside temperatures, cooling methods, and prompt complexity as factors that influence consumption. Michael Kiparsky of UC Berkeley emphasized that no reliable published figures exist for California.
Studies indicate rising water demands from data centers overall. Facilities used roughly 17 billion gallons directly in 2023, compared with 5.6 billion gallons in 2014. Projections show hyperscale centers could require 150 billion gallons between 2025 and 2030, matching the annual household consumption of 4.6 million U.S. residences.
Ren's research, available as a preprint, warns that peak demand on the hottest days could add between 697 million and 1.45 billion gallons of required daily capacity. This volume approaches New York City's average daily water supply, and data centers often draw treated potable water from public systems rather than irrigation sources. The analysis suggests water use could be halved through dry cooling systems, though such options increase energy requirements.
This report is based on coverage by LAist: https://laist.com/news/is-sam-altman-right-that-almonds-use-more-water-than-chatgpt-queries