OpenAI has come to terms on a massive 7-year, $38 billion deal with Amazon Web Services to secure cloud computing capabilities needed to power its suite of advanced AI tools such as ChatGPT and Sora.
The deal which was announced on Monday gives OpenAI access to hundreds of thousands of Nvidia chips housed in Amazon’s global data centers, with full capacity slated to come online by the end of 2026.
AWS said the agreement will allow OpenAI to scale rapidly while tapping the “price, performance, scale, and security” of its cloud network.
“Scaling frontier AI requires massive, reliable compute,” OpenAI CEO Sam Altman said in a statement.
“Our partnership with AWS strengthens the broad compute ecosystem that will power this next era and bring advanced AI to everyone.”
The $38 billion pact marks the first time the ChatGPT maker has turned to Amazon for infrastructure, breaking years of exclusive reliance on Microsoft’s Azure cloud.
It comes just one week after OpenAI restructured its ownership to gain more freedom in financing and operations — a shift that removed Microsoft’s right of first refusal to supply cloud services.
AWS Chief Executive Matt Garman called the deal proof that Amazon’s infrastructure can handle “the vast AI workloads” of frontier model builders like OpenAI.
Amazon’s shares jumped roughly 5% Monday after the announcement, hitting an all-time high.
Under the terms, OpenAI will use Amazon’s UltraServer clusters — racks of Nvidia GB200 and GB300 processors — to train and run its models, process ChatGPT queries, and expand its so-called “agentic AI” systems, where software can complete tasks autonomously.
The partnership will also let OpenAI tap into millions of CPUs for specialized workloads, giving it a way to handle soaring user demand as AI adoption widens.
Amazon said all planned capacity will be online by the end of next year, with expansion continuing through 2027 and beyond.
The partnership represents Amazon’s bid to reclaim ground from faster-growing rivals Microsoft and Google, whose cloud divisions have seen stronger revenue growth on the back of surging AI demand.
While AWS remains the world’s largest cloud provider, analysts said it has lagged behind in attracting high-profile AI customers.
The company is already investing heavily to change that. Amazon last week reported 20% quarterly growth in cloud revenue — its fastest since 2022 — and opened an $11 billion data-center campus in Indiana dedicated to training and running models from OpenAI rival Anthropic.
Amazon has invested $8 billion in Anthropic, which uses its proprietary Trainium chips but also struck a major deal last month to tap as many as one million of Google’s TPU chips.
Until recently, Amazon had been locked out of the OpenAI ecosystem. The startup’s exclusive cloud agreement with Microsoft barred it from buying capacity from other providers.
That arrangement was rewritten last month as part of OpenAI’s sweeping reorganization, giving Altman latitude to pursue deals across the industry.
OpenAI has now committed nearly $600 billion in new cloud contracts across Amazon, Microsoft, Oracle, and Google — a staggering figure for a company generating roughly $13 billion in annual revenue.
The deals are intended to solve what Altman has described as “severe computing shortages” that have constrained model training and product launches.
