Trainium3 UltraServers now available: enabling customers to train and deploy AI models faster at lower cost

Powered by AWS’s first 3nm AI chip, Amazon EC2 Trn3 UltraServers help organizations of all sizes run their most ambitious AI training and inference workloads

Amazon Web Services, Inc. (AWS), an Amazon.com, Inc. (NASDAQ: AMZN) company, today announced the general availability of Trainium3 UltraServers, powered by the new Trainium3 chip, at AWS re:Invent.

Important points

With up to 4.4x more compute performance, 4x more energy efficiency, and nearly 4x more memory bandwidth than Trainium2 UltraServers, Trainium3 UltraServers deliver higher performance for AI workloads, enabling faster AI development at lower operating costs.
Trn3 UltraServers scale to 144 Trainium3 chips and deliver up to 362 FP8 PFLOPs with 4x lower latency, training larger models faster and delivering inference at scale.
Customers like Anthropic, Karakuri, Metagenomi, NetoAI, Ricoh, and Splash Music use Trainium to reduce training and inference costs by up to 50%. Meanwhile, Decart delivers 4x faster inference on real-time generated video at half the cost of GPUs, and Amazon Bedrock is already delivering production workloads on Trainium3.

For more information about AWS Trainium3 UltraServers, visit About Amazon.

About AWS

Amazon Web Services (AWS) is guided by a focus on our customers, a pace of innovation, a commitment to operational excellence, and a long-term mindset. Over nearly two decades, AWS has built one of the fastest-growing enterprise technology businesses in history by democratizing technology and making cloud computing and generative AI accessible to organizations of all sizes and industries. . Millions of customers trust AWS to accelerate innovation, transform their businesses, and shape their future. With the most comprehensive AI capabilities and global infrastructure footprint, AWS enables builders to turn big ideas into reality. For more information, visit aws.amazon.com and follow @AWSNewsroom.