Last week, OpenAI finished its lengthy corporate reorganization, securing a key freedom: the ability to sign cloud contracts with companies other than Microsoft. As part of this new agreement, Microsoft gave up its exclusive right to serve as OpenAI’s primary computing provider. OpenAI quickly put this new freedom to use.
On Monday, Amazon announced a significant multi-year cloud partnership with OpenAI worth $38 billion. Immediately, Amazon Web Services (AWS) will start giving OpenAI access to “thousands” of powerful NVIDIA GB200 and GB300 graphics processing units (GPUs). OpenAI will use these to run its AI models and train its next-generation systems. Amazon expects to have all the agreed-upon capacity ready for OpenAI by the end of 2026. They also included an option for OpenAI to buy more capacity in 2027 and beyond. Amazon stated this partnership will help “millions of users continue to get value from ChatGPT.”
Naturally, a big question arises: how will OpenAI manage to pay for all these massive cloud commitments? The Information recently reported that OpenAI generates about $12 billion in revenue each year.
Yet, as part of its earlier restructuring deal with Microsoft, the company committed to spending $250 billion on Azure services. OpenAI also has an ongoing agreement with Microsoft to share revenue, especially if and when OpenAI develops artificial general intelligence. This new deal with Amazon adds another substantial financial obligation to OpenAI’s books, highlighting the immense costs of developing and running advanced AI.
