In a recent interview published by The Indian Express at a live event in India, OpenAI CEO Sam Altman addressed growing concerns about artificial intelligence and its environmental footprint. Speaking during a session that can be viewed on YouTube, Altman pushed back strongly on claims about AI’s water and energy usage, calling some of the most widely shared figures “totally fake.”
Altman, who was in India for a major AI summit, dismissed viral claims that tools like ChatGPT consume extreme amounts of water per query. “You see these things on the internet where, ‘Don’t use ChatGPT, it’s 17 gallons of water for each query’ or whatever,” he said. “This is completely untrue, totally insane, no connection to reality.” He acknowledged that water usage was once a valid issue when data centers relied on evaporative cooling, but said that practice is no longer standard.
While he rejected the water claims outright, Altman conceded that broader concerns about energy consumption are fair. He clarified that the issue is not energy per individual query, but the total amount of energy used as AI adoption accelerates globally. In his view, this shift makes it more urgent to transition toward nuclear, wind, and solar power sources.
The discussion also touched on a commonly cited comparison suggesting that a single ChatGPT query uses the equivalent of 1.5 iPhone battery charges. Referencing a prior conversation with Bill Gates, the interviewer asked Altman whether that figure was accurate. His response was direct: “There’s no way it’s anything close to that much.”
Altman also criticized what he sees as misleading comparisons between the energy required to train AI models and the energy required for a human to answer a single question. He argued that people often overlook the immense resources required to develop human intelligence. “It also takes a lot of energy to train a human,” he said, pointing to decades of food consumption, education, and even the cumulative evolution of billions of people.
According to Altman, the more relevant comparison is the energy required for an already trained AI model to answer a question versus a human doing the same task. Measured that way, he suggested, AI may already be competitive in terms of energy efficiency.
