According to Odaily, Amazon Web Services (AWS) has announced the "Ultracluster" initiative, a large-scale AI supercomputer powered by hundreds of thousands of internal Trainium chips, set to become operational in 2025. This announcement was made at the annual AWS re:Invent conference, where AWS also introduced Project Rainier, touted as the world's largest AI computing cluster. This project will be utilized by Anthropic, an AI startup in which Amazon recently invested $4 billion.
The Trainium chips, integral to this initiative, are designed by Annapurna Labs, an Amazon subsidiary based in Texas. Acquired in 2015, Annapurna Labs has been pivotal in advancing Amazon's AI chip development, aiming to reduce reliance on Nvidia. During the event, AWS also launched a new Ultraserver featuring 64 interconnected Trainium chips. Additionally, AWS revealed that Apple has become one of its latest chip clients. Benoit Dupin, Apple's Senior Director of Machine Learning and Artificial Intelligence, noted that they are testing the Trainium2 chips, which are expected to offer approximately 50% cost savings.