OpenAI Chief Executive Officer Sam Altman has pushed back against growing criticism over the environmental footprint of artificial intelligence, asserting that the water consumption of 38,000 ChatGPT queries is equivalent to the water required to cultivate a single California almond. Speaking on the Sources podcast with tech journalist Alex Heath, Altman acknowledged that the metric was recalled from memory and could vary slightly, but maintained it accurately reflects the minimal real-world water usage of modern AI workloads compared to common agricultural practices.
Altman argued that widely circulated concerns regarding data center water intensity rely on outdated cooling infrastructure and fail to reflect modern facility design. He noted that contemporary hyperscale data centers operate with closed-loop systems that require water volumes comparable to standard commercial office buildings for everyday plumbing. The OpenAI chief described popular perceptions of high per-query water consumption as empirically flawed and resistant to public correction, pointing to internal baseline estimates that place an average ChatGPT prompt’s direct water footprint at approximately 0.32 milliliters (0.000085 gallons).
However, environmental researchers and policy experts caution that precise calculations remain difficult due to a lack of standardized corporate disclosures regarding cooling mechanisms, regional water sources, and indirectly consumed power grid water. The debate has gained regulatory momentum in California, where state lawmakers are considering transparency measures to mandate detailed data center utility reporting. In response to heightened environmental scrutiny, major cloud providers have highlighted efficiency gains, with Amazon reporting a 2% decline in global data center water usage to 2.5 billion gallons in 2025, and Microsoft citing up to a 90% water-efficiency improvement over its legacy facilities.