Amazon P6e-GB200 UltraServers now available for the highest GPU performance in EC2

Today, Amazon announces the general availability of Amazon Elastic Compute Cloud (Amazon EC2) P6e-GB200 UltraServers, accelerated by NVIDIA GB200 NVL72. Amazon EC2 P6e-GB200 UltraServers offer the highest GPU-based AI training and inference performance in EC2. They are designed to accelerate training and inference of foundation models (FMs) including next-generation capabilities like reasoning models and agentic AI at the trillion-parameter scale.

With P6e-GB200 UltraServers, customers can access up to 72 Blackwell GPUs within one NVLink domain to leverage 360 petaflops of FP8 compute (without sparsity), 13.4 TB of total high bandwidth memory (HBM3e), and up to 28.8 Tbps of Elastic Fabric Adapter (EFAv4) networking. P6e-GB200 UltraServers are powered by the AWS Nitro System, allowing customers to reliably and securely scale AI workloads within EC2 UltraClusters to tens of thousands of GPUs.

P6e-GB200 UltraServers are now available through Amazon EC2 Capacity Blocks for ML in the Dallas Local Zone ("us-east-1-dfw-2a"), an extension of the US East (N. Virginia) region. P6e-GB200 UltraServers are available in two sizes: u-p6e-gb200x72 (72 GPUs within NVLink) and u-p6e-gb200x36 (36 GPUs within NVLink).

To learn more, see Amazon EC2 P6e-GB200 UltraServers and P6-B200 instances.