Amazon Web Services (AWS), the cloud computing division of Amazon.com, has introduced new servers for data centers powered by the company's own AI chips, Trainium2, challenging NVIDIA's dominance in this area, Reuters reports. Apple has already confirmed that it uses Trainium2 chips in its work.
The newly announced servers, known as Trn2 UltraServers, are built on 64 Trainium2 chips and are designed for high-performance generative AI tasks, including machine learning. These servers will be deployed in a supercomputer with hundreds of thousands of chips, and the first user will be Anthropic, an AI startup.
NVIDIA currently controls more than 70% of the AI chip market, but AWS's Trainium2 servers are positioned as a cost-effective alternative. According to Gadi Hutt, Head of Business Development at AWS, servers with Trainium2 chips allow customers to train AI models at 40% lower cost than NVIDIA's offerings.
"We think with Trainium2, (customers) get more compute than what's available from Nvidia today, and they will be able to save cost," Hutt told Reuters in an interview.
AWS also announced plans to debut its next generation of Trainium3 chips in 2025. These chips are expected to form the basis of the next evolution of AI infrastructure, continuing AWS's drive to capture a larger share of the AI chip market.
AWS's new Trn2 ultra-servers will also compete with NVIDIA's flagship servers equipped with 72 Blackwell chips. Both companies have developed their own chip interconnect technologies, but AWS claims that its servers can interconnect more chips, offering more processing power for AI workloads.
Both AWS and NVIDIA rely on Taiwan Semiconductor Manufacturing Company (TSMC) to produce their chips. While NVIDIA has faced supply chain constraints, Hutt said AWS has largely avoided similar issues. "From a supply standpoint, we are in a pretty good shape across all of the supply chain," he said.
AWS management has announced that the new servers and supercomputer will be available in 2025, although the exact date was not specified. The race to bring advanced AI infrastructure to the market has intensified amid growing demand for artificial intelligence technologies.