AWS and NVIDIA to Add 2 Million More AI Chips by 2028

SEATTLE —

Amazon Web Services and NVIDIA said Tuesday they will deploy 2 million additional NVIDIA graphics processing units across AWS's global data centers in 2027 and 2028, a major expansion of a partnership the two companies have built over 16 years, according to a joint announcement from NVIDIA.

The companies said the buildout will include the newer Blackwell Ultra, Rubin and Rubin Ultra chips from NVIDIA, and will support what NVIDIA called "agentic AI," a term for AI systems that can act on tasks with less human oversight, along with scientific research, warehouse robotics and other automation work. NVIDIA said the expansion builds on a plan announced at its GTC conference in 2026 to add more than 1 million GPUs to AWS starting that year, and that customer demand has since exceeded those earlier projections, according to the company.

A portion of the new capacity, 100,000 GPUs, will run on what the companies described as secure AWS infrastructure built specifically for federal and national security workloads, according to the announcement. NVIDIA and AWS did not say which government agencies would use that capacity or when it would come online.

Chips, Memory and Government Work

Beyond raw GPU numbers, the companies said they are working on deeper technical integration. That includes extending NVIDIA's NVLink Fusion interconnect technology to work with new custom high bandwidth memory chips, a project NVIDIA said involves its own engineers working alongside Amazon's chip design unit, Annapurna Labs. AWS also plans to bring NVIDIA's Vera CPU technology into its infrastructure as an option for customers running AI workloads that need heavy computing power alongside graphics chips, according to the announcement.

The companies said they are also integrating NVIDIA's platform with AWS's Nitro security system and its Elastic Fabric Adapter networking technology, and expanding support for NVIDIA's open Nemotron AI models on Amazon's Bedrock and SageMaker cloud services. Amazon Robotics, the company's warehouse automation unit, will adopt NVIDIA's robotics platform, according to the announcement.

Executives Tout the Partnership

"Customers want the freedom to choose the best tools for their AI workloads, and they want confidence that everything works seamlessly together," said Matt Garman, chief executive of AWS, in the announcement. He said AWS has "invested deeply with NVIDIA to make AWS the best place to run NVIDIA AI technologies."

Jensen Huang, founder and chief executive of NVIDIA, said in the announcement that "demand is running ahead of every forecast" and described the expanded partnership as covering "the full stack," including chips, networking and software. Neither company disclosed a total dollar value for the expanded deal, and neither detailed how the added computing capacity would be priced for customers.