OpenAI CEO Sam Altman recently challenged claims that a single ChatGPT query uses six hours of shower water. He addressed the viral claim on the Sources Podcast with Alex Heath, amid rising scrutiny over AI data centers and their environmental impact.
Altman said modern cooling systems have reduced water use, making the shower comparison misleading. He offered a striking comparison, saying nearly 38,000 ChatGPT queries use water comparable to producing one California almond.
He admitted the figure came from memory and might not be exact, while presenting it as a broad estimate. “For every 38,000 ChatGPT queries, that is the same amount of water used for one almond,” Altman said.
The comments target a viral narrative that portrays every AI query as an unusually heavy drain on freshwater supplies. Altman argued that the comparison ignores advances in data center cooling technology.
According to Sam Altman, older facilities relied more heavily on evaporative cooling systems that consumed significant water. Modern large data centers, however, can use water at levels similar to large office buildings, he said.
Independent data still leaves room for debate around the exact figure. CalMatters reported that public data on data center water use remains limited, making Altman’s 38,000-query comparison difficult to verify.
The discussion also highlights a larger issue surrounding AI infrastructure. Growing ChatGPT usage can increase total resource demand even when individual queries consume relatively little water.
Altman’s almond comparison therefore shifts attention from dramatic individual claims toward the broader water footprint of AI infrastructure. The debate now centers on transparency, measurement, and sustainable cooling as AI expansion continues.
Also Read: Google’s $15 Billion India Data Center: Why Water, Wildlife Concerns are Growing