6 days ago
AWS and Nvidia Plan Two Million More AI GPUs
AWS and Nvidia want to add 2 million more AI computer chips to AWS data centres.
The chips are called GPUs and help computers perform difficult AI tasks.
They plan to add them around the world during 2027 and 2028.
AWS says businesses, researchers and governments need more computing power than expected.
The new systems could help with AI agents, scientific work, business automation and robots.
AWS will also offer Nvidia Vera CPUs, which can work with GPUs or operate by themselves.
The companies are developing ways for Nvidia chips and AWS Trainium chips to work together.
They also plan secure AI data centres for the US government.
These centres could contain up to 100,000 Nvidia GPUs for highly protected workloads.
AWS and Nvidia plan to deploy 2 million additional Nvidia GPUs across AWS data centres worldwide in 2027 and 2028.
The expansion will support agentic AI, scientific research, enterprise automation and physical AI workloads.
AWS previously planned to add more than 1 million Nvidia GPUs by 2026, but said demand has exceeded expectations.
AWS is preparing to offer Nvidia Vera CPU-based infrastructure and expand NVLink Fusion work with Nvidia.
The companies will build secure US government AI data centres with up to 100,000 Nvidia GPUs for IL6-or-higher workloads.
- Who
- Amazon Web Services and Nvidia.
- What
- The companies are expanding their partnership to deploy 2 million additional Nvidia GPUs and develop broader AI infrastructure.
- Where
- Across AWS data centres worldwide, including secure AWS infrastructure for the US government.
- When
- The additional GPUs are planned for deployment in 2027 and 2028; AWS previously targeted more than 1 million GPUs by 2026.
- Why
- AWS said demand for AI computing capacity has been stronger than initially expected.
Key facts
- Additional GPUs
- 2 million Nvidia GPUs planned for AWS data centres in 2027 and 2028.
- Earlier AWS target
- More than 1 million Nvidia GPUs planned by 2026.
- Government infrastructure
- Up to 100,000 Nvidia GPUs for secure US government AI data centres.
- Security level
- The government systems are intended to support workloads requiring Impact Level 6 security or higher.
- New CPU infrastructure
- AWS is working to bring Nvidia Vera CPU-based infrastructure to its cloud platform.
- Partnership scope
- The expanded collaboration includes GPUs, CPUs, networking, AI factories, open models and robotics.
- Data-processing claim
- AWS said Nvidia cuDF can provide up to 3.7-times faster Apache Spark processing than a CPU-based configuration and up to 30% improved price-performance.
Quotes
Matt Garman
AWS chief executive officer
“Customers want the freedom to choose the best tools for their AI workloads.”
livemint.com










