OpenAI CEO Sam Altman busts myths about how much water AI data centers really use

OpenAI CEO Sam Altman has pushed back against growing concerns over the amount of water used by AI data centers, saying the problem is being exaggerated.

Altman said that roughly 38,000 ChatGPT queries use the same amount of water as producing one almond in California.

He was quick to add that he was recalling the number from memory and that it might not be exact.

“But it’s close,” Altman said during an interview on the Sources podcast.

Altman was responding to the growing criticism surrounding the water needed to cool the huge data centers being built to power AI.

The concern is not completely made up. Data centers can require large amounts of water for cooling, particularly in areas where traditional cooling systems are used.

But Altman argued that the way people talk about AI's water use has created a misleading picture.

He said modern data centers generally use about as much water as an office building, rather than being the enormous water consumers they are sometimes portrayed as.

Altman said people often look only at the water being used directly inside a data center, while a proper calculation should consider the full water footprint of the electricity and infrastructure involved.

His comments come as communities around the US are increasingly questioning new data center projects because of their demands for electricity, land and water.

The backlash has become a headache for AI companies, which are spending billions of dollars on new facilities to keep up with demand for their models.

Altman acknowledged that the industry still has to earn the public's trust.

But his argument is that the debate should be based on the actual amount of water being used rather than what he described as a widely repeated “meme” about AI data centers.