OpenAI's own CFO says there will not be enough compute in 2026, even after buying ahead.

Sarah Friar told investors that if you want more computing power next year, good luck. Demand is rising like a wall while supply is capped by land, power, and three-year build times.

In the same run of interviews where she reset the hiring bar, OpenAI's finance chief Sarah Friar disclosed something the market applauding AI has not fully absorbed. Even after aggressive advance buying, OpenAI will not have enough computing power in 2026. Her phrasing was blunt: for anyone hoping to buy more compute next year, good luck, because she does not know where to find it. Demand, she said, is rising almost like a vertical wall. Supply is not.

The reason the ceiling is hard is physical, not financial. A large data center takes roughly three years to bring online, and the ones breaking ground today will not produce usable capacity until late 2027 at the earliest. OpenAI is already pre-purchasing capacity for 2028. Money can be raised faster than power lines, permits, and buildings can be delivered, which means the constraint is not a budget line a company can simply expand. It is a queue with a multi-year wait.

For any executive planning a large AI rollout in 2026, this reframes the whole conversation. The bottleneck is not whether your team is fluent or whether the models are capable. It is whether the underlying capacity exists to run what you deploy at the scale you promised the board. A plan that assumes cheap, abundant compute on demand is planning against a supply picture that even OpenAI, one of the largest buyers on earth, describes as scarce.

The strategic read is quieter than the headline. When the raw resource is capped, efficiency stops being a virtue and becomes the whole game. The firms that price this constraint into their timelines and scope, and that design their work to burn as little of the scarce resource as possible, will out-deliver the ones still assuming they can buy their way out. Scarcity does not reward the biggest spender. It rewards the most disciplined design.