Amazon Web Services (AWS) has made compute instances with the Nvidia Blackwell Ultra GPUs generally available.

Dubbed the Amazon EC2 P6-B300 instances, AWS claims they deliver twice the networking bandwidth and 1.5 times more GPU memory compared to previous generation instances.

The P6-B300 instances offer 6.4Tbps Elastic Fabric Adapter networking bandwidth and 2.1TB of GPU memory. Each instance features eight Nvidia B300 GPUs and 192 VCPUs.

The instances are currently available in the US West (Oregon) AWS region.

AWS is not the first cloud provider to embrace Blackwell Ultra. In October 2025, Microsoft revealed that it had deployed a cluster of 4,600 Nvidia GB300 NVL72 systems. Each rack has 72 B300 GPUs.

Google said late last month that instances based on the GB300 NVL72 system were "shipping in production."

Lambda has also deployed at least one GB300 NVL72 system, while IREN said it had secured some 1,200 B300s and 1,200 GB300s in August 2025.