OpenAI has signed a seven-year agreement worth $38 billion to purchase cloud services from Amazon.com Inc (NASDAQ:AMZN)'s Amazon Web Services (AWS), marking its first major infrastructure push since last week’s restructuring expanded the ChatGPT maker’s operational and financial flexibility.
Announced Monday, the pact will give OpenAI access to “hundreds of thousands” of Nvidia Corp (NASDAQ:NVDA, ETR:NVD) GPUs to train and run its models, the companies said. Amazon plans to deploy large volumes of Nvidia’s GB200 and forthcoming GB300 accelerators in dedicated clusters designed to both power ChatGPT’s responses and train OpenAI’s next generation of systems.
The deal highlights the AI sector’s voracious demand for computing power as firms race to build increasingly capable models. CEO Sam Altman said, “Scaling frontier AI requires massive, reliable compute,” describing the AWS partnership as strengthening “the broad compute ecosystem that will power this next era and bring advanced AI to everyone.”
Amazon shares hit record high
Shares of Amazon hit a record on Monday, adding nearly $140 billion in market value as the stock rose about 5% after a near-10% jump on Friday. Microsoft shares dipped briefly on the news. “This is a hugely significant deal and a strong endorsement of AWS compute capabilities to deliver the scale needed to support OpenAI,” said PP Foresight analyst Paolo Pescatore.
Under the arrangement, OpenAI will begin using AWS immediately, with all planned capacity expected online by the end of 2026 and room to expand in 2027 and beyond.
The move is also a vote of confidence in AWS amid investor concerns that Amazon’s cloud unit had fallen behind Microsoft and Google in AI. Those worries eased somewhat after AWS reported strong growth in the September quarter.
Broader strategy
OpenAI’s infrastructure strategy has broadened as it seeks to reduce single-vendor reliance. The company already offers open-weight models on Amazon Bedrock for AWS customers and has previously tapped Alphabet’s Google for cloud services. It has reportedly agreed a separate multiyear arrangement with Oracle and, as part of last week’s restructuring, committed to purchase additional capacity from Microsoft’s Azure.
Altman has outlined an aggressive build-out, saying OpenAI is committed to investing toward 30 gigawatts of computing resources—an ambition he has framed as ultimately adding about 1 gigawatt of capacity per week over time. Such scale has prompted questions on Wall Street about funding, given OpenAI’s rapid revenue growth alongside mounting losses, and speculation that the company is laying groundwork for a potential initial public offering.
For now, the AWS pact cements another large pipeline of compute for OpenAI while sharpening competition among the major cloud providers. As the industry chases “frontier” systems, the winners will likely be those able to secure—and reliably deliver—massive, cost-effective compute at speed.