AWS Trainium3-Powered Amazon EC2 Trn3 UltraServers | Amazon Web Services

Опубликовано 2 декабря 2025, 17:02

AWS announces the general availability of Amazon Elastic Compute Cloud (Amazon EC2) Trn3 UltraServers, powered by Trainium3 — our fourth-generation, first 3nm AWS AI chip purpose-built to deliver the best token economics for next-generation agentic, reasoning, and video-generation applications.

Each AWS Trainium3 chip delivers 2.52 PFLOPs of FP8 compute, boosts memory capacity by 1.5×, and increases bandwidth by 1.7× over Trainium2 — now providing 144 GB HBM3e memory and 4.9 TB/s of memory bandwidth. Trainium3 is engineered for both dense and expert-parallel workloads, featuring advanced data types (MXFP8, MXFP4) and an improved compute-to-memory balance for real-time, multimodal, and reasoning tasks.

Trn3 UltraServers can scale up to 144 Trainium3 chips (a total of 362 FP8 PFLOPs) and are available in EC2 UltraClusters 3.0, enabling scaling to hundreds of thousands of chips. The next-generation Trn3 UltraServer introduces NeuronSwitch-v1, an all-to-all fabric that doubles interchip interconnect bandwidth compared to Trn2 UltraServers.

Learn more about AWS Trainium3: go.aws/trn3

Subscribe to AWS: go.aws/subscribe

Create a free AWS account: go.aws/signup
Try AWS for free: go.aws/free
Connect with an expert: go.aws/contact
Explore more: go.aws/more

Next steps:
Explore on AWS in Analyst Research: go.aws/reports
Discover, deploy, and manage software that runs on AWS: go.aws/marketplace
Join the AWS Partner Network: go.aws/partners
Learn more on how Amazon builds and operates software: go.aws/library

Do you have technical AWS questions?
Ask the community of experts on AWS re:Post: go.aws/3lPaoPb

Why AWS?
Amazon Web Services is the world’s most comprehensive and broadly adopted cloud, enabling customers to build anything they can imagine. We offer the greatest choice of innovative cloud capabilities and expertise, on the most extensive global infrastructure with industry-leading security, reliability, and performance.

#GenerativeAI #LLM #AWS #AmazonWebServices #CloudComputing