Nvidia will supply Amazon Web Services with more than one million AI chips by the end of 2027, marking one of the largest known chip supply agreements between a semiconductor manufacturer and a cloud service provider. The companies revealed the plan during Nvidia’s GTC 2026 conference, highlighting a major expansion of their long-running partnership.
According to Ian Buck, Nvidia’s vice president of hyperscale and high-performance computing, shipments will begin this year and continue through 2027. The timeline aligns with CEO Jensen Huang’s earlier forecast of a trillion-dollar sales opportunity across Nvidia’s Blackwell and Rubin chip families. Financial terms of the agreement were not disclosed.
The arrangement extends beyond GPUs to include Nvidia’s Spectrum networking components, ConnectX systems, and inference-focused processors. AWS intends to deploy seven different types of Nvidia chips to manage a wide range of AI workloads. Buck noted that the complexity of inference operations requires an integrated approach across multiple chip architectures.
In a blog post, AWS said the deployment will integrate Nvidia’s Blackwell and Rubin architectures into its global data centers, underscoring more than 15 years of collaboration between the companies. The addition of Nvidia’s networking gear also marks a shift for AWS, which has typically relied on internally developed infrastructure.
Even as AWS deepens its reliance on Nvidia processors, it continues to develop its own Trainium chips. The companies plan to use Nvidia’s NVLink Fusion platform alongside Amazon’s in-house silicon to optimize efficiency and performance across cloud-based AI operations.
The agreement underscores the intensifying race among major cloud providers to expand AI computing capacity. It positions AWS as a key customer within Nvidia’s growing lineup of hyperscale partners amid record demand for advanced AI infrastructure.
Read Full Article: Jensen Huang Urges AI Leaders to Stop Scaring Public About Artificial Intelligence
