Sam Altman Says AI Proof Automation Costs Could Fall 100-Fold in a Year

OpenAI CEO Sam Altman described a striking moment in AI-driven scientific discovery: an internal proof-automation system reportedly spent a weekend solving some of researchers' hardest career problems one after another.
Altman said a proof involving the Navier–Stokes equations cost roughly $10 million to produce. He estimated that OpenAI might be able to reproduce it a year later for around one-hundredth of that cost.
The experience left him excited for two consecutive nights, but also unexpectedly sad, with a sense that something was ending. His reaction reflects a broader shift in scientific research, where work that once demanded years of human effort could increasingly be accelerated by AI systems.
If the cost of advanced reasoning continues to fall rapidly, the implications could extend beyond research to computing infrastructure, memory demand, and AI-powered commerce. However, Altman's cost figures are estimates, not a guarantee that all scientific discoveries will become 100 times cheaper each year.