OpenAI CEO Sam Altman has reignited the debate over artificial intelligence’s environmental footprint by comparing the water consumption of ChatGPT queries to almond farming. In a podcast interview, Altman claimed that 38,000 interactions with the chatbot use roughly the same amount of water as producing a single almond in California. The statement is the latest attempt by the tech executive to push back against criticism that AI data centers place an unsustainable burden on local water supplies.

What You Need to Know

Altman’s remarks come amid growing scrutiny of the water required to cool AI servers. He argues that modern data centers use no more water than a typical office building, largely because they have moved away from evaporative cooling. Critics, however, point to studies suggesting each ChatGPT query still consumes measurable water, and the cumulative effect across millions of users remains significant. The comparison to almonds, while attention-grabbing, does not resolve the broader question of whether AI expansion is outpacing sustainable resource management.

A New Metric for AI’s Water Use

Speaking on the Sources Podcast with Alex Heath, Altman framed the water issue as a “robust meme” that does not withstand scrutiny. He said that for every 38,000 ChatGPT queries, the total water used equals that needed to grow a single almond in California. Altman acknowledged he was recalling the figure from memory and that it might not be exact. He also claimed that large modern data centers use the equivalent amount of water as an office building, factoring in sinks and toilets.

You might wonder how that comparison holds up. Researchers estimate that one almond requires about 1.1 gallons of water. Meanwhile, Altman previously stated that a single ChatGPT response uses between 1 and 50 milliliters of water, depending on the cooling technique. That would put the almond-equivalent range somewhere between 85 and 5,500 queries, not 38,000. The discrepancy highlights the complexity of calculating water use across different data center designs and locations.

The Cooling Technology Shift

Data centers historically relied on evaporative cooling, which consumes large volumes of water as heat is dissipated through evaporation. Altman emphasized that this method is now rare. Modern facilities increasingly use closed-loop liquid cooling or air-side economizers that drastically reduce water draw. He compared the water footprint of a current-generation data center to that of an office building, though he did not specify which facility or region he was referencing.

The shift matters because water stress varies by geography. A data center in California, for example, draws from a drought-prone system, while one in Georgia may have more abundant supplies. Microsoft, which operates dozens of data centers globally, has invested in water-positive strategies but still reports significant consumption in arid regions.

Why This Matters

Altman’s almond comparison is unlikely to satisfy environmental advocates or regulators. The core issue is scale: as AI adoption accelerates, the cumulative water and energy demand from millions of queries compounds rapidly. If each query uses even a fraction of a teaspoon, the total across OpenAI’s user base becomes substantial. The debate also affects policy decisions. Some municipalities have already pushed back against new data center construction due to water concerns. And with training large models requiring even more resources, the industry faces pressure to disclose accurate water usage metrics.

For consumers, the takeaway is that every ChatGPT interaction carries a hidden environmental cost, however small. The comparison to almonds frames that cost in tangible terms but does not erase the need for transparency. The real question is whether AI companies can scale their operations without straining local water supplies, especially in regions already facing scarcity.

  • Altman’s claim: 38,000 ChatGPT queries equal water for one almond.
  • Independent estimate: 85 to 5,500 queries per almond based on past statements.
  • Cooling evolution: Modern data centers use closed-loop systems, reducing water use.
  • E Policy impact: Local governments may restrict new data centers over water concerns.

The Bigger Picture

SAM ALTMAN’s defense of ChatGPT’s water footprint fits a pattern of tech leaders downplaying resource use. The, And other executives have similarily argued that AI’s environmental impact is manageable. Yet the data remains contested. As AI becomes embedded in daily life, the conversation will shift from abstract comparisons to concrete measurement and regulation. For now, the almond analogy gives critics a vivid talking point but does little to resolve the underlying tension between innovation and sustainability.