OpenAI is dialling back its grand plans to construct massive data centres, a move that reflects growing investor concerns about the company’s spiralling costs as it prepares for a potential stock market flotation this year.
Chief executive Sam Altman recently pulled no punches about the headaches involved in running facilities of this enormity. Speaking at BlackRock’s U.S. Infrastructure Summit in Washington, D.C., he described how “anything at this scale, it’s just like so much stuff goes wrong.”
Altman recounted a severe weather event that knocked out operations at the Abilene, Texas site—the flagship for the $500 billion Stargate project involving OpenAI, Oracle and SoftBank—while grappling with supply chain woes and tight deadlines.
Fresh off a $110 billion funding round last month that valued the firm at $730 billion, with major contributions from Amazon and Nvidia, OpenAI faces pressure to prove its business model to public market investors.
Instead of building its own infrastructure, it is turning to cloud capacity from partners like Microsoft, Amazon and Oracle. As Daniel Newman, CEO of Futurum Group, put it, “OpenAI has come to the realization that the market doesn’t necessarily appreciate the reckless approach to growth and spending. The market wants to see OpenAI’s revenues rolling at a pace in which the spending can be justified.”
This shift comes against fierce competition from Anthropic and Google, as compute shortages hobble AI model development. OpenAI has now capped its compute spending at $600 billion by 2030, a far cry from earlier $1.4 trillion pledges that spooked markets and echoed dot-com excesses.
Experts like Virginia Tech’s Walid Saad warn that erecting gigawatt-scale centres takes years amid regulatory and logistical hurdles.
Newman sums up the high-stakes race ahead. “OpenAI is doing what it must do, which is gain access to compute at scale. This is the race.”
With profitability under the microscope, the path to a successful IPO will demand disciplined execution every step of the way.