Amazon has significantly expanded its collaboration with Nvidia by tripling its original order of GPU chips to add an additional 2 million units to its AWS data centers over the next two years, set for deployment in 2027 and 2028. This comes shortly after Amazon’s initial deal to deploy over 1 million Nvidia GPUs, reflecting a stronger-than-expected surge in demand for AI computing power. The GPUs ordered include Nvidia’s latest Blackwell Ultra, Rubin, and Rubin Ultra models, which are designed to handle the intensive compute requirements of AI training and inference workloads.

Although Amazon is intensifying its partnership with Nvidia, it is simultaneously investing in developing its own AI chips, including the Trainium series designed as a competitive alternative to Nvidia's offerings. Amazon’s AI chief, Peter DeSantis, has hinted at plans to sell Trainium chips to other companies, signaling an ambition to lessen dependency on Nvidia. However, Nvidia continues to lead in AI hardware, with its upcoming Vera CPUs and networking technologies also slated for integration into AWS, alongside open models and platforms designed to power not only AI infrastructure but Amazon’s warehouse robotics.

This broadened partnership between Amazon and Nvidia goes beyond chip supply, incorporating Nvidia's complete AI stack aimed at enhancing Amazon’s robotics and enterprise services. Amazon will adopt Nvidia’s Omniverse for simulation, Cosmos for world modeling, Isaac for robotics development, and Jetson hardware for edge AI in robotics. Additionally, Nvidia’s Nemotron family of open AI models will be made available through Amazon’s Bedrock and SageMaker platforms, further deepening the integration of Nvidia technologies into Amazon’s service offerings.

The deal is valued in the tens of billions of dollars, though exact financial terms remain undisclosed. Nvidia reported Q2 revenues of $96.2 billion, with data center sales surging 117% year-on-year, driven largely by AI demand. The company plans to invest $279 billion to secure manufacturing capacity, ensuring supply to meet this growing demand. Nvidia CEO Jensen Huang emphasized the productive and profitable work AI is enabling, underpinning the strategic importance of this expanded collaboration with Amazon amid a booming AI infrastructure market.