Amazon Web Services Introduces New AI-Powered Servers

Amazon (NASDAQ:AMZN) Web Services (AWS) has introduced new data center servers featuring its proprietary artificial intelligence (AI) chips, challenging Nvidia (NASDAQ:NVDA)'s market dominance. Apple Inc (NASDAQ:AAPL) has confirmed it will utilize these new Trainium2 chips.

These servers are part of a massive supercomputer incorporating hundreds of thousands of chips, with the announcement made on Tuesday.

The supercomputer, powered by AWS's Trainium2 chips, will be utilized by AI startup Anthropic as the first company to leverage this technology. Anthropic specializes in creating reliable and interpretable AI systems, aiming to enhance its AI models' capabilities.

Apple executive Benoit Dupin confirmed significant adoption of Trainium2 chips by Apple.

AWS Chief Executive Matt Garman indicated that development is already underway for Trainium3, the next iteration of their AI chip, expected to launch next year.

The new Amazon Elastic (NYSE:ESTC) Compute Cloud (Amazon EC2) instances, powered by AWS Trainium2, have also been launched, marking the introduction of Trn2 UltraServers. These UltraServers deliver exceptional performance and cost efficiency for training and deploying contemporary AI models, including large language models (LLM) and foundation models (FM).

The Trn2 instances enhance price performance by 30-40% compared to current GPU-based EC2 instances, featuring 16 Trainium2 chips that provide 20.8 peak petaflops of compute—ideal for handling AI workloads with billions of parameters.

For more demanding AI tasks, the Trn2 UltraServers offer a new EC2 service with 64 interconnected Trainium2 chips, delivering up to 83.2 peak petaflops of compute, effectively quadrupling compute, memory, and networking capabilities for training the largest AI models.

AWS and Anthropic's collaborative effort, Project Rainier, aims to develop an EC2 UltraCluster of Trn2 UltraServers, destined to become the world's largest AI compute cluster upon completion.

Upcoming Trainium3 chips will be produced using a 3-nanometer process node, promising to quadruple the Trn2 UltraServers' performance.

The AWS Neuron software development kit (SDK) will optimize AI models for Trainium chips, supporting popular frameworks like JAX and PyTorch, and is integrated with the Hugging Face model hub, hosting over 100,000 models.

Trn2 instances are currently available in the US East (Ohio) AWS Region with plans for broader availability, while Trn2 UltraServers are in a preview phase.

This article was generated with the support of AI and reviewed by an editor. For more information see our T&C.