OpenAI CEO says AI may need “significant fraction” of Earth’s electricity

OpenAI CEO Sam Altman publicly stated that a “significant fraction” of Earth’s electricity should be dedicated to running AI systems, making the admission during AMD’s AI conference last week. The comment reveals the massive energy demands of artificial intelligence companies and raises serious concerns about the environmental impact of AI’s exponential growth.

What he said: Altman made the startling admission when asked by AMD CEO Lisa Su whether there would ever be enough GPUs to meet AI demand.

• “Theoretically, at some points, you can see that a significant fraction of the power on Earth should be spent running AI compute,” Altman said. “And maybe we’re going to get there.”
• The exchange came after Su mentioned ChatGPT’s recent outages, which likely stemmed from computing power shortages.

The energy reality: ChatGPT already consumes massive amounts of electricity, despite Altman’s attempts to downplay individual usage.

• Altman claimed each ChatGPT query uses “about 0.34 watt-hours, about what an oven would use in a little over one second.”
• However, a University of California, Riverside and Washington Post study found ChatGPT uses nearly 40 million kilowatts of energy daily—enough to power the Empire State Building for 18 months or charge eight million smartphones.
• These figures don’t include other large language models or AI systems, meaning the actual environmental impact is even greater.

Why this matters: The admission highlights how AI companies are prioritizing computational power over environmental concerns as they race toward artificial general intelligence.

• Conventional electricity generation relies heavily on fossil fuel combustion, which contributes to climate change.
• Adding energy-intensive AI systems to an already strained power grid creates compounding environmental problems.
• Altman’s casual dismissal of these concerns suggests the industry may not be adequately addressing sustainability challenges.

The bigger picture: OpenAI has previously acknowledged resource constraints, with Altman admitting earlier this year that the company had run out of GPUs—the specialized chips needed to power large language models.

• The June ChatGPT outages likely resulted from insufficient computing power rather than technical glitches.
• Altman recently committed to purchasing GPUs from AMD, whose CEO he was addressing when making these comments.
• In a recent blog post, Altman claimed the world is approaching a “gentle singularity,” where AI meets or surpasses human capabilities.