OpenAI struck a $38 billion, seven-year cloud deal with AWS, a move that immediately expands the ChatGPT maker’s access to computing power and sent Amazon shares up about 5% on November 3, 2025. The pact underscores how central infrastructure has become to the AI race—and how cloud giants are vying to power the next wave of frontier models.
Announced on November 3, 2025, the agreement allows OpenAI to run and scale core AI workloads on Amazon Web Services right away. OpenAI will access hundreds of thousands of Nvidia GPUs to train and serve its models, with both companies targeting full planned capacity by the end of 2026 and room to expand into 2027 and beyond. The $38 billion commitment spans seven years, marking one of the largest cloud infrastructure deals in AI to date.
The tie-up follows recent changes to OpenAI’s structure and partnerships that increased its operational flexibility and ended Microsoft’s exclusive role as cloud provider, clearing the way for OpenAI to diversify where it sources compute. For AWS, the win puts Amazon squarely in the mix to handle some of the industry’s most demanding AI workloads.
The deal is a validation for AWS at a moment when investors are scrutinizing how cloud platforms capture surging AI demand. It also gives OpenAI another channel for reliable, large-scale compute as it builds more capable systems and serves hundreds of millions of users. Reuters reported that Amazon plans to deploy clusters featuring Nvidia’s latest accelerators to support training and inference at scale, a setup tailored for high-intensity AI operations.
The broader backdrop is a capital-intensive AI buildout. While OpenAI’s revenues are rising, the company—and its partners—are committing vast sums to data centers and chips to meet escalating compute needs. That spending spree has stirred debate about sustainability even as demand from businesses and consumers continues to grow.
Amazon shares rose about 5% on the announcement, briefly touching a record high as investors welcomed the boost to AWS’s growth story. Microsoft shares dipped intraday before stabilizing, reflecting expectations that major AI customers will use multiple clouds.
OpenAI will begin using AWS immediately, with most capacity slated to come online over the next year and a half. All planned AWS capacity is targeted for deployment by end-2026, with the option to expand further in 2027, setting up another ramp in available compute as OpenAI trains its next generation of models.