AWS announced 2 million additional NVIDIA GPUs (Blackwell Ultra, Rubin, Rubin Ultra) for 2027-2028, with 100,000 reserved for U.S. government use.
Following·since 27 Aug 2026·Day 7·2 sources
AWS Adds 2M More NVIDIA GPUs to Roadmap for 2027-2028
The gist
The expansion deepens the AWS-NVIDIA relationship across compute, networking, software, and silicon layers simultaneously. The Trainium-NVLink Fusion integration means AWS is weaving NVIDIA components into its own custom accelerator stack, not treating them as separate tracks.
The full picture
AWS expanded its NVIDIA partnership by committing to 2 million additional GPUs, covering Blackwell Ultra, Rubin, and Rubin Ultra generations, for deployment across AWS global data centers in 2027-2028. This comes on top of more than one million GPUs AWS had already committed to earlier in 2026. Of the newly announced GPUs, 100,000 are designated exclusively for U.S. government workloads. NVIDIA's role in AWS infrastructure spans Vera CPUs, Spectrum networking, Nemotron models, and CUDA-X libraries, as well as Amazon Robotics' physical AI platform. Future Amazon Trainium chips are set to integrate with NVIDIA NVLink Fusion and NVHBM memory technology, connecting Amazon's custom silicon to NVIDIA interconnect and memory infrastructure.
How it developed
Sources
Rohan Paul Twitter
Related
- Grew out ofAI capex overtaking oil and gas
The daily email
Want this in your inbox?
I send a short email each morning with the stories that moved. If you would rather just read here, that works too.
Subscribe free