Amazon Web Services (AWS), Amazon's cloud computing division, announced on Tuesday plans to produce "Ultracluster," a massive AI supercomputer powered by its Trainium chips, at its re:Invent [sic] conference.
Set to launch in 2025, the Ultracluster will support AI startup Anthropic and offer five times the computing capacity of its current systems. The Ultraserver integrates Amazon's networking technology, offering up to 83.2 petaflops of computing power, representing four times Nvidia's typical chip servers. Also today, AWS revealed a new server called Ultraserver, which is made up of 64 of its interconnected chips.
"Today, there's really only one choice on the GPU side, and it's just Nvidia," said Matt Garman, chief executive of Amazon Web Services. "We think that customers would appreciate having multiple choices."