Sam Altman compares ChatGPT water usage to almond production
OpenAI's CEO frames AI's environmental footprint as negligible compared to agricultural and commercial water use.
OpenAI CEO Sam Altman has claimed that the water consumption required to process thousands of AI queries is comparable to the production of a single nut. The statement arrives as the tech industry faces intensifying scrutiny over the environmental costs of scaling large language models.
Speaking on the 'Sources' podcast, Altman asserted that 38,000 ChatGPT queries consume as much water as the production of one almond in California. In an effort to further downplay the resource intensity of AI infrastructure, Altman also claimed that a modern data center uses roughly the same amount of water as a large office building.
The Cooling Challenge
As AI models grow in complexity and scale, the hardware required to run them generates immense heat. Data centers rely on sophisticated cooling systems—often involving the evaporation of millions of gallons of water—to prevent servers from overheating. This physical requirement has turned water-use efficiency (WUE) into a primary metric for environmental critics and regulators who argue that the digital economy has a tangible, thirsty footprint.
Framing the Footprint
By comparing digital queries to agricultural production, Altman is attempting to shift the narrative regarding AI's resource consumption. Framing the impact as negligible relative to the water-intensive nature of farming aims to neutralize concerns about the sustainability of generative AI. However, this comparison often overlooks the localized impact of data centers. Unlike agricultural water use, which is spread across rural landscapes, data center consumption is concentrated in specific hubs, often placing sudden and severe pressure on the municipal water supplies of water-stressed regions.
The Transparency Gap
Industry observers remain skeptical of these comparisons due to a lack of granular transparency regarding how OpenAI and its partners calculate water usage. While the almond analogy suggests efficiency, it does not account for the total lifecycle of the infrastructure or the varying efficiency of different cooling technologies. As the race for AI supremacy accelerates, the industry is expected to face increasing pressure to provide audited, site-specific data on water consumption rather than broad analogies. Whether these claims hold up under independent environmental audits remains a key point of contention for the sector.