OpenAI has signed a landmark multi-year partnership with Amazon Web Services (AWS), valued at approximately $38 billion, marking a major shift in the company’s cloud infrastructure strategy. The deal, announced November 3, follows changes to OpenAI’s long-standing arrangement with Microsoft that previously made Azure its exclusive cloud provider.
Under the new agreement, AWS will supply the computing infrastructure to power OpenAI’s large-scale AI operations, including both model training and inference for products like ChatGPT and Sora. The partnership will see OpenAI renting thousands of NVIDIA GB200 and GB300 GPUs through Amazon EC2 UltraServers, giving the company expanded access to high-performance computing resources for its next-generation models.
The collaboration is set to run for seven years, with AWS confirming that OpenAI is expected to use its full planned GPU capacity by the end of 2026 and could expand its usage further in 2027. The $38 billion figure reflects the scale of compute demand behind OpenAI’s rapid model development and the rising costs of sustaining large language models at global scale.
AWS CEO Matt Garman described the deal as a reflection of OpenAI’s growing infrastructure needs and AWS’s position in meeting them. “As OpenAI continues to push the boundaries of what’s possible, AWS’s best-in-class infrastructure will serve as a backbone for their AI ambitions,” Garman said. “The breadth and immediate availability of optimized compute demonstrates why AWS is uniquely positioned to support OpenAI’s vast AI workloads.”
This move comes after Microsoft modified its exclusive agreement earlier in 2025, allowing OpenAI to deploy non-API products — such as ChatGPT and its video generation model Sora — on alternative cloud platforms. Microsoft still retains its right of first refusal on infrastructure deals, but the latest expansion effectively opens the door for OpenAI to diversify its compute partnerships.
The announcement also follows OpenAI’s recent release of its first open-weight model, gpt-oss, which AWS now hosts via its Amazon Bedrock service. According to Amazon, gpt-oss has quickly become one of the platform’s most requested public models, used by enterprise clients including Comscore, Peloton, and Thomson Reuters.
The AWS partnership signals a broader trend in the AI industry: the move toward multi-cloud strategies as companies scale up model development and deployment. For OpenAI, the deal provides redundancy, flexibility, and cost control at a time when the demand for compute power continues to surge. For AWS, the partnership strengthens its position as a leading provider of AI infrastructure, competing directly with Microsoft Azure and Google Cloud in the high-performance computing market.
