Amazon Launches New AI Servers, Apple Joins as a Customer
Amazon (NASDAQ:AMZN) Web Services (AWS) has announced the introduction of new data center servers featuring its custom-built artificial intelligence (AI) chips, challenging Nvidia's dominance in the industry. It has been confirmed that Apple (NASDAQ:AAPL) Inc. plans to use these new Trainium2 chips. AWS's cloud unit stated that these servers will be part of a massive supercomputer containing hundreds of thousands of chips. This announcement was made on Tuesday.
The supercomputer powered by AWS Trainium2 chips will initially be utilized by the AI initiative Anthropic. Known for creating reliable and interpretable AI systems, Anthropic will leverage this computational power to enhance the capabilities of its AI models.
Benoit Dupin, an executive from Apple, confirmed the technology giant's use of Trainium2 chips, indicating significant adoption of AWS's new product.
AWS Chief Executive Matt Garman also revealed that the company is currently working on Trainium3, which is the next evolution of AI chips planned for release next year.
The new Amazon Elastic Compute Cloud (Amazon EC2) instances powered by AWS Trainium2 have now been made publicly available, introducing Trn2 UltraServers. These UltraServers are designed to provide exceptional performance and cost efficiency for training and deploying contemporary AI models, including large language models (LLM) and foundational models (FM).
Trn2 instances promise 30-40% better price performance compared to existing GPU-based EC2 instances, featuring 16 Trainium2 chips that deliver 20.8 peak petaflops of computing power. This makes them ideal for processing AI workloads with billions of parameters.
For even more demanding AI tasks, Trn2 UltraServers offer a new EC2 service with 64 interconnected Trainium2 chips providing up to 83.2 peak petaflops of computing power. This setup quadruples a single instance's computing, memory, and networking capabilities, allowing for the training and deployment of the world's largest AI models.
The collaboration project between AWS and Anthropic, known as Project Rainier, aims to build an EC2 UltraCluster composed of Trn2 UltraServers, which will become the world's largest AI computing cluster upon completion.
AWS also highlighted an upcoming Trainium3 chip, which is expected to be produced using a 3-nanometer process node and promises to quadruple the performance of the existing Trn2 UltraServers.
The AWS Neuron software development kit (SDK) facilitates the optimization of AI models to run on Trainium chips, supports popular frameworks like JAX and PyTorch, and integrates with the Hugging Face model hub, which hosts over 100,000 models.
Trn2 instances are currently available in the US East (Ohio) AWS Region, with plans to roll out in additional regions soon. Meanwhile, Trn2 UltraServers are available in a preview stage.