OpenAI to run its advanced AI workloads on AWS's infrastructure under a multi-year tie-up

The agreement, valued at USD 38 billion, covers the coming years and reflects OpenAI's commitment to expand its compute capacity by leveraging AWS's price, performance, scale and security.

Under the terms, OpenAI will gain immediate access to AWS compute power that includes hundreds of thousands of NVIDIA GPUs, and the flexibility to scale further to tens of millions of CPUs. Deployment is targeted to be completed by the end of 2026, with expansion into 2027 and beyond possible.

AWS emphasises that it has unusual experience of running large-scale AI infrastructure securely and reliably, citing clusters with more than 500,000 chips. The infrastructure being built for OpenAI will use the latest NVIDIA GB200 and GB300 GPUs, clustered through AWS's EC2 UltraServers on a low-latency network to support both real-time inference and training of next-gen models.

OpenAI CEO Sam Altman said, "Scaling frontier AI requires massive, reliable compute. Our partnership with AWS strengthens the broad compute ecosystem that will power this next era and bring advanced AI to everyone."

Meanwhile, AWS CEO Matt Garman said: "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."

The partnership builds on the existing collaboration between the two companies. Earlier in the year, OpenAI's open-weight foundation models became available on AWS's Amazon Bedrock service, making these models accessible to millions of AWS customers. Customers using those models include Bystreet, Comscore, Peloton, Thomson Reuters, Triomics and Verana Health.

Overall, this strategic partnership positions OpenAI to leverage AWS's high-performance, secure and scalable compute infrastructure, while providing AWS a landmark deal that reaffirms its leadership in supporting frontier AI models. (ANI)