OpenAI CEO Sam Altman toured the company’s Stargate data center in Saline, Michigan, and spoke about surging demand for AI compute, the future of work, and the risks of a global AI race. David Faber reports for CNBC in an exclusive interview conducted at the site.
A $45 billion bet on AI demand
The Stargate facility in Saline is one of the largest AI infrastructure projects in the world. Altman said the site alone costs $16 billion to develop and build, with at least another $30 billion going into the hardware inside. OpenAI is the primary customer of the facility.
When asked whether that level of investment would generate adequate returns, Altman expressed confidence. He pointed to rapidly growing revenue at OpenAI and across the industry, and said demand for compute is rising faster than the cost of using AI is falling.
“We still don’t think the world has appreciated how much AI every person and every business is going to want,” Altman said. He described a near-future scenario in which AI systems run continuously in the background, monitoring a user’s information and context to provide constant assistance rather than responding to individual requests.
Coding models as the turning point
Altman identified improvements in coding models as the single biggest driver of recent demand growth. He said these tools transformed how companies build products and increased the speed and efficiency of software development. That shift, he noted, became clearly visible in late 2024 and accelerated in early 2025.
He added that scientists and knowledge workers beyond software developers are now adopting the models at a higher rate, suggesting the growth is broadening across industries.
Jobs: a more optimistic view, with caveats
On employment, Altman said he has updated his outlook in a more positive direction, while acknowledging he still does not have a definitive answer. He offered two observations to support a less alarming view:
- Companies that have adopted AI the most are also hiring the most.
- Companies announcing layoffs attributed to AI tend to be those using AI the least.
He also said he underestimated how uneven AI capabilities would turn out to be. The models excel at specific, well-defined tasks but struggle with long-term, complex supervision work. This “jaggedness,” as he put it, means skilled human workers remain essential for combining AI tools with judgment and oversight.
Altman acknowledged that public anxiety about AI’s impact on jobs is legitimate. He criticized messaging from parts of the AI industry that frames a future of mass unemployment as inevitable, calling it “a terrible message.” He said the industry has failed not so much to explain the benefits of AI, but to explain how people will stay in control of their own futures.
The global race and its limits
On competition with China, Altman drew a distinction between economic rivalry and global safety risks. He said competing for economic advantage is acceptable, but that issues such as preventing loss of control over AI systems, cybersecurity, and biosecurity should not be treated as a race. He compared the need for international coordination to the role of the International Atomic Energy Agency in governing nuclear technology.
Altman also commented on competition among AI companies after news emerged that rival Anthropic had filed documents for a potential public offering. He said OpenAI is not focused on the timing of its own IPO, describing it as a financing event rather than a strategic priority. He confirmed an IPO would happen at some point.
He rejected the idea that AI will be a winner-takes-all market, saying the world will demand multiple providers for infrastructure this critical.
Stay up to date
AI for content creation: the latest tools, tips and trends. Every two weeks in your inbox:
