Amazon Expands Nvidia Partnership With 2 Million More AI Chips for AWS

Amazon and Nvidia are expanding their artificial intelligence infrastructure partnership with a new agreement that will bring an additional 2 million Nvidia GPUs to Amazon Web Services data centres.

The companies said the new deployment will include Nvidia Blackwell Ultra, Rubin, and Rubin Ultra GPUs, with the chips expected to arrive at AWS data centres in 2027 and 2028. The announcement expands on a previous agreement announced five months earlier, when Amazon committed to deploying more than 1 million Nvidia GPUs across its cloud infrastructure.

Nvidia said demand for its AI computing technology has grown beyond earlier expectations since that initial agreement. Neither company disclosed financial terms, but the scale of the GPU expansion suggests a deal potentially worth tens of billions of dollars at current AI chip prices.

Amazon and Nvidia expand AI infrastructure partnership

The expanded relationship goes beyond Amazon purchasing more Nvidia GPUs. Nvidia’s networking technology, CPUs, data processing software, open models, and robotics platforms will also be integrated across AWS infrastructure.

The companies said demand from startups, enterprises, AI research organisations, and governments contributed to the decision to deepen their collaboration.

AWS will also provide Nvidia’s open AI models through Amazon Bedrock and SageMaker, allowing enterprise customers to access Nvidia’s Nemotron family of models through Amazon’s managed AI services.

AWS continues developing its own AI chips

The larger Nvidia order comes as Amazon continues investing in its own semiconductor efforts. The company has developed custom chips, including Trainium AI accelerators and Graviton processors, as it works to reduce its dependence on external chip suppliers.

Amazon has previously said its custom chip business is expanding, supported by commitments from AI companies using AWS infrastructure. The company has also explored offering Trainium chips to other organisations as an alternative for AI workloads.

Despite Amazon’s internal chip development, Nvidia remains the leading supplier of advanced AI GPUs used for training and running large AI models. Nvidia will also provide Vera CPUs as part of the expanded AWS deployment, with some versions integrated with Rubin systems and others available separately.

Nvidia extends AI technology into robotics and enterprise tools

The partnership will also include Nvidia’s physical AI technologies for Amazon’s warehouse robotics operations. The company’s robotics technology stack includes Omniverse for simulation, Cosmos for world models, Isaac for robotics development, and Jetson computing hardware for edge AI applications.

Nvidia recently introduced a new version of Jetson designed to make robotics computing more accessible for entry-level edge AI applications.

Nvidia’s expansion with Amazon comes as the chipmaker reports strong demand for AI infrastructure. The company reported $96.2 billion in quarterly sales, with data centre revenue accounting for the majority of its business during the period.

Nvidia CEO Jensen Huang has argued that additional computing capacity is becoming increasingly valuable as AI systems move into more productive applications. The company continues to invest heavily in supply and manufacturing capacity to meet expected demand for future data centre projects.

The Amazon partnership shows how cloud providers are expanding access to advanced AI hardware while also developing their own chip strategies. The next challenge will be determining how effectively this growing computing capacity translates into profitable AI services.