AI
Altman downplays model distillation threat, says OpenAI's scale beats margins
On a recent podcast, Sam Altman discussed AI security, cheaper models, and why he isn't worried about competitors reverse-engineering OpenAI's technology.
The New Stack
Sam Altman recently sat down with Patrick O'Shaughnessy on the Invest Like the Best podcast to discuss OpenAI's strategy in the face of emerging competition and security challenges. The conversation revealed his pragmatic stance on rivals building cheaper alternatives and underscored the company's confidence in sustained dominance through sheer usage volume.
When third-party distillation doesn't keep him up at night#
The topic of model distillation — where companies extract capabilities from powerful AI systems to create cheaper versions — came up during the discussion. While Altman acknowledged the practice exists, he characterized it as a manageable concern rather than an existential threat to OpenAI.
Altman explained that OpenAI has consistently expected lower-cost models to proliferate in the market. "I have always assumed that there are going to be great cheap models in the world, and we better be the greatest and the cheapest," he said. Even with competing offerings like Moonshot AI's Kimi K3 gaining traction amid claims they were built using distillation techniques, Altman contended that OpenAI's models deliver superior performance for the price. His view: external distillation of OpenAI's models ranks below many other priorities. "I would rather people not distill from us, for sure. But this is not in my top ten list of worries."
Usage scale over profit margins#
Altman pushed back on the premise that OpenAI must maintain extraordinarily high margins to fund its expensive training operations. Instead, he pointed to the company's unprecedented usage levels as the true competitive moat.
"We have so much usage of our models that we do not need to be a gigantically high-margin business to be able to afford model training," Altman stated. The sheer volume of queries processed by OpenAI's systems generates sufficient revenue to finance research and infrastructure, even at lower per-unit prices. This economic model suggests the company can undercut rivals on cost while remaining profitable — a position unavailable to competitors with lower usage volumes.
He also acknowledged space in the AI ecosystem for alternative approaches, including open-source models where users deploy their own weights rather than relying on OpenAI's infrastructure.
