Amazon and NVIDIA have expanded their partnership to supply 2 million additional NVIDIA GPU chips to Amazon data centers. This deal ensures Amazon Web Services can meet the surging demand from startups, enterprises, AI research labs, and governments. Buyers and developers relying on AWS for AI workloads will see sustained access to high-performance computing resources.
Deal secures Blackwell and Rubin chips for AWS AI infrastructure
The new chips include NVIDIA Blackwell Ultra, Rubin, and Rubin Ultra GPUs, scheduled for supply in 2027 and 2028. This order follows a previous agreement to deploy over 1 million NVIDIA GPUs on AWS infrastructure starting this year. The partnership also extends to robotics, with Amazon adopting NVIDIA's physical AI stack (Omniverse, Cosmos, Isaac, Jetson) for warehouse robots.
AWS will integrate NVIDIA's Nemotron open model family into Amazon Bedrock and SageMaker services. This integration allows developers to access open-source AI models directly through AWS platforms. The move strengthens the software ecosystem supporting the hardware infrastructure.
NVIDIA reported Q2 revenue of $96.2 billion, with data center revenue up 117% year-over-year to $89 billion. Amazon is simultaneously developing its own AI chips, including the Trainium accelerator and Arm-based Graviton CPUs, to reduce reliance on NVIDIA. This dual strategy balances immediate high-performance needs with long-term hardware independence.
Discussion
0 comments
Log in to join the thread with a thoughtful take, question, or correction.