Altman Claim on AI Water Usage Found Misleading
Fact-checkers disputed a recent comparison between ChatGPT query water consumption and California almond farming.
Updated on Sept. 21, 2026 in Artificial Intelligence
Live Poll
Do you trust the information provided by AI companies regarding their environmental impact?
On September 1, 2026, Sam Altman claimed that 38,000 ChatGPT queries use the same amount of water as producing one almond. PolitiFact subsequently rated the comparison as Mostly False, noting that estimates suggest 1,000 to 10,000 prompts actually equate to the water usage of one almond.
Why it matters
The dispute highlights the difficulty in quantifying AI environmental impact, which remains highly dependent on data center location, specific model architecture, and prompt length. Without transparent reporting from developers, verifying such sustainability claims remains a challenge for the industry.
While Altman claimed 38,000 queries per almond, the actual estimated footprint is 1,000 to 10,000 queries based on fresh water use of approximately 6 liters per almond. Current estimations place water usage at roughly 0.32 milliliters per individual ChatGPT query.
The players
Sam Altman
The CEO of OpenAI, a leading artificial intelligence research and deployment organization focused on generative models.
PolitiFact
A non-profit project that fact-checks statements made by prominent figures and public officials.
The details
Data centers rely on water for cooling computer systems, a process that prevents hardware from overheating during intensive computation. Agricultural water usage, such as that measured in 2019 California almond farming studies, accounts for total consumption including direct irrigation, rainfall, and water required to dilute pollutants. The discrepancy between AI and agricultural metrics arises from these vastly different baseline measurement methodologies.
Timeline
2019: A study measured water use for California almond farming.
2020: U.S. farms consumed nearly 30 trillion gallons of water for irrigation.
2021: A study estimated the fresh water requirement for one almond at approximately 6 liters.
2023: U.S. data centers consumed approximately 228 billion gallons of water.
September 1, 2026: Sam Altman made the claim regarding water usage on the Sources Podcast.
The Tech Race
The claim follows an industry trend of attempting to benchmark AI's environmental footprint against established agricultural and industrial sectors. This effort remains in its early stages, as there is currently no standardized industry metric for reporting the water-to-computation ratio.
The lack of official consumption data means users cannot currently evaluate the specific environmental cost of their own daily interactions with AI services. Future transparency from providers will be necessary to determine which platforms are more efficient for specific workflows.
The takeaway
The gap between Altman's claim and independent estimates illustrates the ongoing challenge of establishing credible sustainability benchmarks for generative AI. Readers should watch for future regulatory disclosures from data center operators to see if verifiable consumption data becomes available.
Further reading
For more on the environmental impact of large-scale infrastructure, browse our Artificial Intelligence section.
Source note: This article includes information reported by Houston Chronicle.
Live Poll
Do you trust the information provided by AI companies regarding their environmental impact?
