Speaking at an event hosted by The Indian Express this week, OpenAI CEO Sam Altman addressed concerns about AI’s environmental impact.
First, Altman, who was in India for a major AI summit, said concerns about AI’s water usage were “absolutely fake,” but acknowledged that it was a real problem at the time, “when we were doing evaporative cooling in data centers.”
“I don’t do that now, but I see things on the Internet saying things like, ‘Don’t use ChatGPT. It takes 17 gallons of water per query,'” Altman said. “This is completely untrue, completely insane and has no connection to reality.”
He added that it’s “fair” to be concerned about “overall energy consumption rather than per-query because there’s so much AI being used around the world right now.” In his view, this means the world needs to “move very quickly to nuclear, wind and solar power.”
There is no legal obligation for tech companies to disclose their energy and water usage, so scientists are trying to find out on their own. Data centers have also been implicated in rising electricity prices.
Citing an earlier conversation with Bill Gates, the interviewer asked if it was accurate that one ChatGPT query is currently equivalent to 1.5 iPhone battery charges, to which Altman replied, “There’s no way there’s anything close to that.”
Altman also complained that many discussions about ChatGPT’s energy usage are “unfair,” especially when they focus on “how much energy it takes to train an AI model compared to what it costs a human to run one inference query.”
tech crunch event
boston, massachusetts
|
June 9, 2026
“But training people also requires a lot of energy,” Altman says. “It takes 20 years of life and all the food you’ve eaten during that time to become intelligent. Not only that, but creating humans required the extensive evolution of the 100 billion humans who have lived so far, learning how to avoid being eaten by predators, learning how to understand science, etc.”
So, in his view, a fair comparison is, “When you ask ChatGPT a question, how much energy does it take for that model to be trained to answer that question compared to a human? And perhaps on an energy efficiency basis, measured that way, AI is already catching up.”
The full interview can be viewed below. Conversation about water and energy usage starts around 26:35.
