OpenAI CEO Sam Altman has pushed back against concerns about ChatGPTâs water consumption, claiming that approximately 38,000 queries use as much water as is required to grow one California almond.
Altman made the comparison while discussing growing opposition to artificial intelligence data centers on the premiere episode of journalist Alex Heathâs Sources podcast.
The OpenAI chief acknowledged that he was quoting the number from memory and that it might not be exact. However, he maintained that widespread claims about ChatGPT consuming huge quantities of water are exaggerated.
How much water does a ChatGPT query use?
Altman previously estimated that an average ChatGPT query consumes approximately 0.000085 gallons of water, equivalent to about 0.32 milliliters.
Using that estimate, 38,000 prompts would represent approximately 12.2 liters of water. Altman compared this with the water required to grow a single California almond.
However, the comparison is not straightforward. A frequently cited 2019 study estimated that producing one California almond requires approximately 3.56 liters of water. Applying that figure to Altmanâs per-query estimate would produce a ratio closer to 11,000 ChatGPT queries per almond.
This difference does not necessarily prove Altmanâs broader argument wrong, but it shows that such comparisons depend heavily on which agricultural and data-center estimates are used.
Altman also rejected viral claims suggesting that asking ChatGPT one question consumes water comparable to running a shower for several hours.
âThat has been a robust meme and difficult to disprove, but I donât think holds up to any scrutiny,â Altman said, according to a report from Business Insider.
Why AI data-center water use remains controversial
Artificial intelligence systems require large data centers filled with powerful computing equipment. Those machines generate heat and must be cooled continuously, sometimes through systems that consume water through evaporation.
Altman argued that many modern data centers no longer operate like older, water-intensive facilities. He said a large contemporary facility may use roughly as much water during normal operations as a conventional office building.
A 2024 review of Virginiaâs data-center industry found that water use varied considerably between facilities. It concluded that current statewide consumption was sustainable, while recommending that local governments assess the expected water requirements of individual projects.
Concerns have nevertheless intensified as technology companies rapidly expand the infrastructure required to train and operate AI models. The issue is particularly sensitive in drought-prone regions where data centers may compete with residents, farms and businesses for limited water resources.
Altman said the industry faces what he considers a reputation problem rather than simply a water problem. He argued that AI companies must demonstrate enough value to communities to justify the infrastructure supporting their services.
The discussion comes as interest in OpenAIâs environmental footprint continues to grow alongside ChatGPT usage. It follows wider scrutiny of the technology sector, including recent questions about the reliability of ChatGPT and other major AI platforms as demand increases.
Altmanâs 38,000-query comparison offers a striking response to critics, but the differing estimates show why the environmental cost of an individual AI request remains difficult to calculate precisely.
Make Swikblog your go-to source on Google for reliable updates, smart insights, and daily trends.
