AWS and OpenAI Ink $38B NVIDIA Deal

This is a quick one today. AWS and OpenAI inked a deal, but it was notable for its size. The deal was valued at $38B, spanning years. The other big impact is that this was not a deal for AWS’ AI accelerators nor their NICs, but instead was focused on NVIDIA GPUs.

AWS and OpenAI Ink $38B NVIDIA Deal

Notable in this announcement is that the $38B deal spans years, and is for NVIDIA GPUs. From the AWS release:

Clustering the NVIDIA GPUs—both GB200s and GB300s—via Amazon EC2 UltraServers on the same network enables low-latency performance across interconnected systems, allowing OpenAI to efficiently run workloads with optimal performance. (Source: AWS)

That means that OpenAI is running NVIDIA GB200 and GB300 in AWS. While that may not seem like a big deal at first, since OpenAI is out signing contracts with hyper-scalers and neo-clouds with power capacity, the notable part is that it is demanding NVIDIA. That NVIDIA is GB200 and GB300 targeted to be deployed by the end of 2026. As a result, AWS needs to deploy a lot of NVIDIA gear in its data centers. The end of 2026 also puts the GB200 and GB300 deployments in the same timeframe as the next generation of NVIDIA racks and AMD Helios racks start being deployed at scale in late 2026.

Final Words

For NVIDIA, this means a nice chunk of revenue selling its parts into AWS and OpenAI. For AWS, it gets some sizable revenue. At the same time, this is a leading AI company buying more NVIDIA. AWS would probably have preferred OpenAI to ask for its silicon stack. For OpenAI it is adding to commitments, but is also getting access to more power and data center capacity from AWS. Another notable benefit, is that by OpenAI investing in infrastructure at major hyper-scalers, it is clearly creating a path to bring its services closer to the applications that are running inside AWS.