OpenAI’s $38 billion AWS deal signals move to cut reliance on Microsoft

OpenAI, AWS

OpenAI has signed a massive $38 billion, multi-year deal with Amazon Web Services (AWS), marking a significant expansion of its cloud infrastructure strategy — and a quiet step away from total dependence on Microsoft’s Azure.

Under the new agreement, AWS will provide the core compute power for OpenAI’s next-generation AI workloads. The deal, spanning seven years, gives OpenAI access to hundreds of thousands of NVIDIA GPUs — including the latest GB200 and GB300 chips — with the option to scale to tens of millions of CPUs for massive agentic AI workloads. All infrastructure is expected to be operational by the end of 2026, with expansion into 2027 and beyond.

The move signals OpenAI’s intent to diversify its computing backbone, which has until now relied heavily on Microsoft’s Azure cloud platform. AWS’s proven record in running secure, large-scale AI infrastructure at over half a million chips makes it a natural partner as OpenAI pushes toward training more advanced and demanding models.

“Scaling frontier AI requires massive, reliable compute,” said Sam Altman, OpenAI CEO. “Our partnership with AWS strengthens the broad compute ecosystem that will power this next era and bring advanced AI to everyone.” AWS chief executive Matt Garman added that the collaboration highlights AWS’s unmatched capability to handle “vast AI workloads with immediate availability.”

This partnership builds on existing ties between the two companies. OpenAI’s open-weight foundation models are already available on Amazon Bedrock, serving thousands of enterprise customers including Peloton, Thomson Reuters, and Triomics.

For OpenAI, the AWS deal is both strategic and symbolic — expanding its cloud options while subtly rebalancing power in the AI infrastructure landscape long dominated by its alliance with Microsoft.