Amazon Web Services (AWS) and NVIDIA have expanded their strategic partnership, aiming to meet the growing demand for artificial intelligence systems. Under the partnership announced by the two companies, 2 million additional NVIDIA graphics processing units (GPU) will be integrated into AWS’s global cloud infrastructure between 2027 and 2028. This step will continue AWS’s previously announced commitment to add more than 1 million chips to its data centers starting in 2026.
A New Phase in Hardware Architecture
In addition to GPU supply, the agreement also includes full-stack technology integration. Under the targets outlined in our report titled Amazon and NVIDIA Expand Partnership: 3 Million GPU to Be Deployed Across AWS Infrastructure, the systems will feature NVIDIA’s Blackwell Ultra, Rubin and Rubin Ultra architectures. The companies also plan to bring NVIDIA’s new Vera-based processors (CPU) to the AWS cloud to support advanced artificial intelligence applications.
To boost performance in large-scale computing clusters, the infrastructure will also receive expanded NVLink Fusion network updates that work alongside custom high-bandwidth memory technology.
Increase in Data Processing Performance
The innovations in the infrastructure are also reflected in software and data analytics services. By using NVIDIA’s cuDF software in Amazon’s analytics services, processing speeds are expected to increase by up to 3.7 times compared with standard configurations. This increase is expected to improve the price/performance ratio by approximately 30% compared with systems based solely on standard processors. Within the vector search capabilities of search services, index creation processes performed on GPU are completed approximately 9 times faster.
Cooperation in Government and Robotics
The project also includes a dedicated capacity reserved for the sensitive computing needs of US government agencies and the defense sector. AI facilities with a capacity of 100 thousand GPU will be built on secure AWS infrastructure to support critical national security workloads across the country.
Amazon’s robotics unit is also working with NVIDIA on simulation tools to accelerate the training of next-generation warehouse robots.
The latest hardware updates are said to deliver up to 4.6 times faster inference and 2.1 times more powerful graphics output than previous-generation systems. It is emphasized that the massive investments indicate that the momentum in AI spending across the industry will continue.