Sam Altman defends AI energy use with human training comparison

×

As OpenAI races to secure more infrastructure through deals with companies including Oracle and Nvidia, Sam Altman is also pushing back on criticism of AI’s energy appetite. Speaking at an AI summit in India earlier this year, Altman told The Indian Express that training AI should be weighed against the energy required to produce human expertise.

His argument came in response to questions about the rising energy demands of both training and inference. The cost goes beyond electricity for data centers: it also includes water for cooling and the materials and components needed to build the hardware running these systems.

Altman’s defense was blunt. He argued that humans are deeply inefficient, and that the fair comparison is the total energy needed to create a human expert versus a machine expert. In that framing, AI is treated as a direct substitute for human intelligence, with the added advantage that machine systems can be trained much faster.

Recommended reading

Mira Murati’s lab unveils Inkling, a 975B open model

There is also a narrower technical case behind the remark: AI energy efficiency could improve over time as hardware and model design advance.

Critics, however, were sharply dismissive. As TechRadar notes, detractors said the comparison ignores the fact that the human brain operates on roughly 20 watts of power. They also argued that the framing itself is ethically troubling and dehumanizing, especially when used to justify the growing resource demands of large-scale AI systems.

The exchange captures a widening fault line in the AI boom: companies are investing heavily to build bigger systems, while scrutiny is intensifying over the environmental and ethical costs of keeping them running.