When OpenAI announced last week the end of its drawn-out corporate restructuring, one of the freedoms the company had managed to negotiate for itself was the ability to more easily sign cloud contracts with Microsoft’s competitors. With the new agreement in place, the company waived its first right of refusal to be OpenAI’s compute provider. OpenAI is wasting no time taking advantage of that freedom.
On Monday, Amazon announced a new multi-year, $38 billion cloud partnership with OpenAI. “Starting immediately,” Amazon Web Services will provide the company with access to “thousands” of NVIDIA GB200 and GB300 GPUs for inference and training its next-generation models. Amazon expects to deploy all the capacity OpenAI has agreed to buy by the end of 2026, with the option to purchase additional capacity in 2027 and beyond. Amazon says the partnership “will help millions of users continue to get value from ChatGPT.”
Of course, the question is how OpenAI will pay for all of its cloud commitments. The Information recently reported the company was generating about $12 billion in annualized revenue. As part of just its restructuring agreement, the company agreed to spend $250 billion on Azure services from Microsoft. It also has a revenue-sharing agreement with the tech giant that will continue when and if OpenAI is able to develop artificial general intelligence.
