OpenAI said it has signed a $38 billion agreement to use AWS and its Amazon EC2 UltraServers immediately. Under the deal, OpenAI will access hundreds of thousands of NVIDIA GPUs and likely CPUs for AI agents.

For AWS, the OpenAI deal is another big win. AWS reported 20% revenue growth in the third quarter and launched its Project Rainer cluster for Anthropic. AWS will announce Trainium3 at re:Invent 2025.

The big takeaways from this OpenAI-AWS deal may be the following:

  • OpenAI and Anthropic are diversifying compute with the latter more likely to leverage custom silicon from hyperscalers. In the AWS case, Anthropic is using Trainium2. For Google Cloud, the Anthropic deal is about the TPUs. OpenAI is diversified across Microsoft Azure, Google Cloud, Oracle Cloud Infrastructure and now AWS. See: OpenAI, Anthropic increasingly diverge as strategies evolve
  • OpenAI is more of a Nvidia shop, but it'll be interesting to see whether it winds up diversifying into AWS custom silicon--especially for CPUs.
  • The AWS-OpenAI deal is another data point in the AI industry's circular economy. OpenAI has more than $1 trillion in future commitments. However, OpenAI appears to be paying AWS immediately.

In the statement, AWS and OpenAI said:

"OpenAI will immediately start utilizing AWS compute as part of this partnership, with all capacity targeted to be deployed before the end of 2026, and the ability to expand further into 2027 and beyond."

Other key points:

  • The deal is focused on Nvidia GPUs but can expand to "tens of millions of CPUs" for agentic workloads. Inference will be a huge market as AI agents scale.
  • The OpenAI infrastructure on AWS will include clusters of Nvidia's GB200s and GB300s via Amazon EC2 UltraServers.
  • AWS said the OpenAI deal includes inference for ChatGPT as well as training new models.
  • OpenAI and AWS started to work together when the AI company put its open weight foundation models on Amazon Bedrock.

More:

 

Business Research Themes

Data to Decisions, Tech Optimization

Roles

Chief Information Officer

Tags

OpenAI, Artificial Intelligence (AI), Large Language Models (LLMs), Amazon