Hajra Naz
OpenAI CEO Sam Altman addressed recent claims that ChatGPT uses massive amounts of energy and water, calling some of the reports outdated or outright “fake.”
While acknowledging that the AI industry consumes significant energy overall, Altman emphasized that comparing a single chatbot query to human energy use is misleading.
Training AI vs. Training Humans
Altman highlighted that training an AI model shouldn’t be compared directly to a human performing one query.
“It takes a lot of energy to train a human,” Altman told The Indian Express at a major AI summit.
He explained that humans spend roughly 20 years of life consuming food and resources before they become capable of independent thinking.
“It takes, like, the very widespread evolution of the 100 billion people that have ever lived… to learn science or survive,” Altman added, illustrating the long-term energy investment in humans versus AI.
Debunking Water Usage Myths
Altman also addressed circulating claims that OpenAI uses excessive water to power ChatGPT:
“Water is totally fake. It used to be true—we did evaporative cooling in data centers—but now that we don’t, these claims online, like ‘17 gallons of water per query,’ are completely wrong.”
Energy per Query
Altman shared practical figures for energy consumption per ChatGPT query:
-
About 0.34 watt-hours, roughly equivalent to:
-
What an oven uses in just over a second
-
What a high-efficiency lightbulb consumes in a few minutes
-
He emphasized that while the AI industry as a whole consumes more energy due to growth, a single query is relatively minimal.
Advocating Alternative Energy
Altman is a strong proponent of alternative energy sources for AI, including solar, wind, and nuclear power.
Unlike some tech leaders, including Elon Musk, he is skeptical of space-based data centers as a realistic solution for energy savings in the next decade.
Outside of OpenAI, Altman has invested heavily in nuclear energy:
-
Former chairman of Oklo, a nuclear startup
-
Backer of Helion, aiming to build the world’s first fusion power plant in Washington State
Data Centers and U.S. Energy Use
In the U.S., energy consumption from data centers is becoming a major topic.
-
Last month, President Donald Trump said he was working with tech companies to ensure that residents don’t face higher energy bills from nearby data centers.
-
Consulting firm McKinsey & Company estimates that by 2050, data centers could account for 14% of total U.S. power demand.
Altman’s overall message: ChatGPT queries themselves aren’t energy hogs, but the AI industry should continue focusing on sustainable energy solutions as usage grows.
