During a recent episode of the “Sources” podcast with Alex Heath, OpenAI CEO Sam Altman addressed concerns about the water usage associated with ChatGPT. He pointed out that the amount of water required for 38,000 ChatGPT queries is comparable to that needed to produce a single almond, which he described as somewhat trivial.
Altman acknowledged the growing criticism that tech companies are facing online regarding their water consumption. He reflected on the notion circulated that running a single ChatGPT query consumes as much water as a six-hour shower, which he dismissed as exaggerated.
He mentioned that while he didn’t have precise calculations available, the water used for 38,000 queries is about the same as that used to grow just one almond in California. He seemed puzzled about the discrepancy in perspective, noting how people consuming many almonds generally don’t feel guilty about their water impact.
He also commented on past cooling methods used in data centers, explaining that modern facilities don’t rely on evaporative cooling anymore. Instead, he indicated they use a similar amount of water as typical office buildings do for their daily functions, like running sinks and toilets.
Altman characterized the criticism regarding data center water usage as a strong meme that doesn’t really stand up to scrutiny. Almonds, which are notorious for their high water use, typically require about one gallon per almond to produce, according to the Almond Board of California.
Alongside almonds, other crops such as pistachios, walnuts, and peaches also use significant amounts of water. On average, Americans consume around 2.3 pounds of almonds each year.
Previously, Altman estimated that a single query on ChatGPT requires about 0.3 milliliters of water, which is roughly 1/15 of a teaspoon. Given that a gallon contains around 3,785 milliliters, he suggested that about 3,800 queries equate to one almond’s water usage.
However, it’s worth noting that Altman’s estimate might be outdated, possibly reflecting an era when data centers still used evaporative cooling methods. As ChatGPT grows more popular, reportedly handling 2.5 billion user prompts daily, the water usage conversation continues to be relevant.
In the tech realm, OpenAI and other companies are also venturing into developing AI agents, which demand significantly more energy than chatbots do. These AI agents tackle complex software issues across various sectors, including finance and retail.
Meanwhile, communities nationwide are pushing back against the rise of large, energy-intensive data centers, citing concerns over noise, light pollution, and fears of job loss due to AI advancements.
In a recent statement, President Trump urged the nation to welcome data centers, suggesting that towns resistant to them risk becoming “backwards and poor.”
This spring saw frustrations in Fayetteville, Georgia, where residents were alarmed to learn that a newly established data center consumed 30 million gallons of water, leading to low water pressure during a drought. In Morgan County, complaints arose from residents about a nearby Meta data center allegedly polluting their water supply, although it appeared that only a limited number of homes were affected, which might have been tied to construction activities rather than the data center itself.
