Overview
- OpenAI will begin running workloads on AWS immediately, with full contracted capacity slated to be in place before the end of 2026.
- AWS plans to deploy hundreds of thousands of Nvidia accelerators, including GB200 and GB300, in clusters built for training and inference.
- The agreement carries an initial seven-year term with options to expand the collaboration after the initial period.
- AWS infrastructure, including EC2 UltraServer and services that host foundation models, is positioned to speed responses and deepen reasoning in future GPT and ChatGPT iterations.
- The move joins a broader industry shift as OpenAI lines up capacity across multiple providers, and Amazon shares rose roughly 5% following the announcement.