Nvidia and AWS are expanding their partnership with plans to deploy 2 million additional Nvidia GPUs across AWS's global infrastructure as demand for AI compute continues to increase.

The GPUs, expected to be deployed in 2027 and 2028, will include Nvidia's Blackwell Ultra, Rubin and Rubin Ultra platforms. The added capacity will support workloads including agentic AI, scientific discovery and physical AI.

The companies said the expansion brings Nvidia's total GPU capacity across AWS's global infrastructure introduced this year to more than 3 million.

The partnership will also move beyond GPU capacity, with the companies working together across CPUs, networking, memory, open models, data processing and robotics.

The deal comes as AI workloads increasingly move from experimentation into production, with enterprises and governments deploying AI agents, automation and physical AI applications at larger scale. The companies said this is creating demand for infrastructure that can handle not only model training and inference, but also data processing, indexing and the workloads required to operate AI systems. The buildout is part of a wider capital spending surge across the cloud industry, with Amazon, Microsoft, Alphabet and Meta all sharply increasing investment in AI data centers.

"Demand is running ahead of every forecast," Nvidia CEO Jensen Huang said in a release. "We are expanding our partnership across the full stack to make agentic and physical AI real at an unprecedented pace."

Under the agreement, AWS will also integrate Nvidia's Vera CPU-based infrastructure into its cloud, giving customers another option for workloads that require CPU compute alongside accelerated infrastructure. Nvidia introduced Vera earlier this year as a CPU designed for AI agent workloads. AWS, the largest cloud provider, also develops its own AI chips, including its Trainium and Inferentia accelerators, giving customers a mix of Nvidia and in-house silicon.

The deal also includes a plan to build AI factories for the U.S. government, including deploying 100,000 Nvidia GPUs on AWS infrastructure for federal and national security workloads. Federal AI adoption and access to advanced computing have become growing policy priorities in Washington.

The expanded partnership places greater emphasis on robotics and physical AI, with Amazon Robotics set to collaborate with Nvidia to develop next-generation robots using Nvidia's Jetson platform, Omniverse libraries and Isaac robotics development platform. Robotics and physical AI have become an increasing focus for Nvidia beyond its data center chip business.

The latest developments come as Nvidia continues to report rapid growth in demand for AI infrastructure. The AI chipmaker reported $96.2 billion in quarterly revenue on August 26, with data center revenue reaching $89 billion, up 117% year over year. Cloud providers are among Nvidia's largest customers, and multi-year commitments of this size are watched as an indicator of expected AI infrastructure demand.