AWS Launches Amazon EC2 P6-B300 Instances for Large-Scale AI Model Training

P6-B300 EC2 instances with NVIDIA Blackwell Ultra GPUs offer enhanced memory and network throughput for trillion-parameter model workloads.

By Sofia Reed · Editorial pseudonym · 2026-05-06 · 1 min read

AI-ASSISTED REPORTING · RETROSPECTIVE ARCHIVE

Retrospective edition: 2026-05-06. Source published: 2026-05-06. Added to this archive: 2026-11-03. This briefing was prepared later from the linked announcement; it was not published on this website on the historical edition date.

This new instance type provides eight Blackwell Ultra GPUs, 2.1 TB of GPU memory bandwidth, and 6.4 Tbps Elastic Fabric Adapter networking, exceeding the P6-B200 generation by 2x network bandwidth and 1.5x GPU memory and TFLOPS. It supports 4 TB system memory to facilitate training and serving of very large foundation and language models with sophisticated optimization methods.

Available in key US regions including N. Virginia and GovCloud (US-East), P6-B300 instances target organizations deploying advanced AI models needing substantial compute and data throughput. The announcements are from AWS, with performance metrics based on internal specifications rather than third-party validation.

Source & further reading

Amazon EC2 P6-B300 instances are now available in the US East (N. Virginia) Region ↗

Source published 2026-05-06. This report was prepared with AI assistance from the linked primary-source material. Claims and limitations should be checked against the original announcement or research.