Amazon Web Services plans to add 2 million Nvidia GPUs to its global infrastructure between 2027 and 2028, expanding capacity for artificial intelligence workloads.
The deployment follows an earlier commitment to add more than 1 million chips starting in 2026 after customer demand exceeded expectations.
AWS and Nvidia also plan to introduce new Vera-based CPUs in Amazon’s cloud to support more advanced AI applications. The companies said networking upgrades will include extended NVLink Fusion technology paired with custom high-bandwidth memory for large computing clusters.
Amazon said upgrades to its analytics service using Nvidia’s cuDF software are expected to raise processing speeds to nearly 3.7 times those of standard configurations. The same changes are projected to improve price performance by about 30% versus setups that rely only on standard chips.
Vector search on Amazon’s search service is also being upgraded, with index construction completing about nine times faster when run on GPUs.
Amazon’s robotics division is working with Nvidia on simulation tools intended to speed training for next-generation warehouse robots.
The plan also includes 100,000 chips reserved for sensitive government and defense computing needs nationwide.
“Customers want the freedom to choose the best tools for their AI workloads, and they want confidence that everything works seamlessly together,” AWS Chief Executive Matt Garman said.
Garman said AWS had invested heavily with Nvidia to optimize performance across its infrastructure, including networking, security and deployment, and that the expanded collaboration would give research labs, enterprises and governments more ways to build and deploy AI on AWS.
“Nvidia and AWS have built one of the great growth engines of the AI era, and demand is running ahead of every forecast,” Nvidia Chief Executive Jensen Huang said.
Huang said the companies were expanding the partnership across GPUs, CPUs, networking, open models and software to accelerate agentic and physical AI.
AWS said new processors delivered under the expanded plan are expected to provide faster inference and improved graphics performance. Recent hardware upgrades already deliver up to 4.6 times faster inference and 2.1 times stronger graphics output than the previous generation of systems.
Featured image credit
