AWS, OpenAI strike $38B deal, rivals fume
Amazon Web Services has entered into a $38 billion agreement with OpenAI that will provide computing infrastructure for artificial intelligence operations over the next seven years. The deal gives OpenAI access to hundreds of thousands of NVIDIA graphics processing units through AWS data centers, with potential expansion to tens of millions of central processing units for scaling agent-based applications.
OpenAI will begin using AWS computing resources right away, with full deployment scheduled by late 2026 and possible extension into 2027. The setup features NVIDIA GB200 and GB300 chips organized through Amazon EC2 UltraServers, creating low-latency networks designed to handle tasks ranging from ChatGPT responses to training future AI models. Sam Altman, who co-founded and leads OpenAI, said expanding frontier AI demands substantial and dependable computing power.
The collaboration builds on earlier work between the two companies. OpenAI foundation models became available through Amazon Bedrock this year, attracting thousands of enterprise clients such as Peloton, Thomson Reuters and Verana Health. These customers use the models for agent workflows, software development, scientific research and mathematics applications. Matt Garman, who runs AWS, said the infrastructure demonstrates why his company can support large-scale AI demands.
