AWS, OpenAI strike $38B computing deal
Under the multi‑year pact, OpenAI has secured access to AWS computing capacity comprising hundreds of thousands of Nvidia GPUs, with the ability to scale further, and has reportedly begun using AWS immediately.
OpenAI has signed a seven‑year, $38 billion agreement to run core artificial‑intelligence workloads on Amazon Web Services, a tie‑up that has immediately buoyed investor sentiment.
Under the multi‑year pact, OpenAI has secured access to AWS computing capacity comprising hundreds of thousands of Nvidia GPUs, with the ability to scale further, and has reportedly begun using AWS immediately.
The companies have said the targeted capacity is scheduled to be in place by the end of 2026, with scope to expand into 2027 and beyond.
Shares in Amazon rose about 4% on 3 November 2025 and closed at a record, lifting the Nasdaq while the Dow has ended lower on the day. The move followed Friday’s post‑earnings rally and reflected optimism that AWS has secured a flagship AI customer.
Inside the deal: chips, clusters and timeline
AWS said it has been provisioning EC2 UltraServers linked across low‑latency networks, clustering Nvidia GB200 and GB300‑class accelerators to support both model training and inference at scale for services such as ChatGPT.
OpenAI’s chief executive and co‑founder Sam Altman said “scaling frontier AI requires massive, reliable compute,” adding that the AWS partnership has strengthened the compute ecosystem the company needs.
AWS chief executive Matt Garman said AWS’s infrastructure will serve as a backbone for OpenAI’s ambitions.
The announcement has come days after OpenAI completed a restructuring that has ended Microsoft’s exclusive rights to supply cloud resources, a change that has allowed the AI company to diversify providers. The deal has also signalled fresh competitive momentum for AWS amid scrutiny of cloud AI market share.