Particle.news
Download on the App Store

OpenAI Strikes $38 Billion, Seven-Year AWS Deal to Power ChatGPT on Nvidia GPUs

The pact formalizes OpenAI’s shift to a multi‑cloud strategy after loosening Microsoft exclusivity.

Overview

  • OpenAI began using AWS immediately, with all purchased capacity targeted to be online before the end of 2026 and an option to expand usage starting in 2027.
  • AWS will provide Amazon EC2 UltraServers configured with hundreds of thousands of Nvidia GB200 and GB300 GPUs clustered for low‑latency performance.
  • The agreement underscores OpenAI’s reliance on Nvidia hardware rather than AWS’s Trainium chips for training and inference workloads.
  • Amazon shares jumped roughly 4–5% to record highs after the announcement, with analysts saying the win strengthens AWS’s AI posture and likely boosts its backlog.
  • Coverage indicates AWS is expected to add bespoke data‑center capacity for OpenAI as the industry confronts financing, supply, and power constraints tied to rapid AI buildouts.