OpenAI on May 19 introduced Guaranteed Capacity, a new offering that lets enterprise customers lock in long-term access to its compute with discounted token pricing in exchange for one-, two-, or three-year commitments. CEO Sam Altman framed the move as a response to growing demand for predictable access: "Customers are increasingly asking us for certainty on capacity. As models get better, we expect that the world will be capacity-constrained for some time."
The longer the commitment, the deeper the discount, with capacity drawable across OpenAI's full portfolio of models. The program is available only until OpenAI sells out of its current allocation, with the company saying it plans to open another window in the future. It is pitched at enterprises running production AI agents and customer-facing applications where losing compute access would be a material operational risk.
The launch underscores OpenAI's shift from pure model vendor to infrastructure provider — the company has now secured roughly 10 GW of contracted US AI infrastructure and over $1 trillion in multi-decade supply commitments across cloud, chip and data-center partners. Altman called the deal a "big win-win" because the commitments also help OpenAI plan its build-out.