OpenAI is confronting a familiar scaling challenge: the cost of the underlying infrastructure is rising faster than anticipated. In a CNBC interview, CEO Sam Altman acknowledged that the company’s need for more compute and memory is “definitely a headwind,” underscoring the financial pressure that accompanies efforts to remain at the cutting edge of generative AI.
At the same time, Altman shared a concrete advance from the company’s engineering work. He said the newest OpenAI model delivers a 54% increase in token efficiency specifically for agentic coding tasks. That improvement points to materially greater work done per unit of compute for that use case, offering a partial counterweight to the growing infrastructure bill.
Partnership and customers
Altman moved to quell reports of strain with one of OpenAI’s largest commercial partners, confirming that Microsoft will remain one of the company’s biggest and most important customers. That affirmation provides clarity about a major revenue relationship amid questions about how OpenAI will fund its heavy infrastructure needs.
Government role and regulation
Pressed about speculation that the U.S. government might take an equity stake in OpenAI - including reports of a rumored 5% ownership - Altman was emphatic in rejecting those claims. He described such reports as inaccurate. While he dismissed ownership rumors, he said OpenAI has implemented “many changes” following extended discussions with Washington and expressed hope for a smoother, more collaborative regulatory relationship going forward.
IPO timeline and international competition
Asked directly whether an initial public offering could occur in 2026, Altman gave a succinct answer: "I don't know." He also sounded a note of geopolitical caution, observing that Chinese open-source AI models are becoming “very good,” a brief but clear recognition of rising global competition in AI development.
The comments portray a company balancing two realities: mounting compute and memory costs that complicate rapid scale-up, and engineering breakthroughs that materially improve model efficiency for specific tasks. How those opposing forces net out will shape OpenAI’s cost trajectory and its commercial positioning with large customers and regulators alike.