OpenAI CEO Sam Altman has pushed back against concerns over the water consumption associated with artificial intelligence, claiming that around 38,000 ChatGPT queries use roughly the same amount of water required to produce a single almond in California.
Altman made the comparison during an appearance on the Sources podcast with technology journalist Alex Heath, as AI companies face growing scrutiny over the environmental impact of the data centres required to train and operate increasingly capable models.
The OpenAI chief acknowledged during the discussion that he was recalling the calculation from memory and that the precise figure could be wrong. His comparison appears to draw on an estimate he published in 2025, when he said an average ChatGPT query consumed approximately 0.000085 gallons, or 0.32 millilitres, of water.
However, the 38,000-query comparison is disputed. Independent fact-checking has found that publicly available information is insufficient to verify OpenAI's per-query figure, while estimates of the amount of water required to produce an almond also vary depending on how water consumption is measured.
PolitiFact rated Altman's comparison "Mostly False", estimating that roughly 1,000 to 10,000 conversational AI prompts could correspond to the direct freshwater required for one California almond, depending on the assumptions used. Researchers have also noted that AI water consumption can vary significantly according to the model, data centre location, hardware, cooling system, prompt length and electricity source.
Altman also challenged claims that individual AI queries require large quantities of water, arguing that newer data centres have become less dependent on evaporative cooling. He said modern facilities can have direct water requirements comparable to conventional office buildings.
The debate extends beyond the amount of water consumed by an individual prompt. AI systems operate at enormous scale, meaning relatively small per-query requirements can accumulate as billions of interactions are processed. More computationally intensive reasoning models and AI agents can also require substantially different resources from simple chatbot requests.
Data centres can additionally consume water indirectly through electricity generation, while their local impact depends heavily on where facilities are built and the availability of water in those regions.
Altman's comments highlight a broader debate around measuring AI's environmental footprint. While technology companies are improving the efficiency of models, chips and cooling infrastructure, researchers continue to call for greater transparency around the energy and water requirements of large-scale AI systems.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.
Related Articles
