Sam Altman pulled back the curtain on a less glamorous side of AI: building it is chaotic, fragile—and insanely expensive.
Speaking about OpenAI’s infrastructure push, Altman described how even a single severe weather event temporarily knocked out part of a key Texas data centre tied to its Stargate project with Oracle and SoftBank. Add in supply chain headaches and relentless deadlines, and the reality is clear: scaling AI isn’t just hard—it’s unpredictable.
Now, with OpenAI eyeing a future IPO after a massive valuation surge, the company is making a noticeable shift. Instead of aggressively building its own global network of mega data centres, it’s dialling things back—scrapping some ambitious plans and leaning more heavily on cloud infrastructure.
The reason? Wall Street is no longer impressed by unchecked spending. Investors want proof that OpenAI can turn its enormous costs into sustainable revenue.
That creates a tension at the heart of the AI race. As competitors like Google and Anthropic keep pushing forward, OpenAI is being forced to balance ambition with discipline—right when compute demand, chips, and energy needs are exploding.
In other words: the company that helped ignite the AI boom is now learning that winning it might require slowing down.
Author: Mohammed Najem