Amazon announced new AI servers with custom chips designed for major technology partners including Apple

In a significant challenge to Nvidia's dominance in the AI chip market, Amazon Web Services unveiled new data center servers on Tuesday equipped with custom-designed AI chips developed by Amazon itself. The servers, which AWS calls Trn2 UltraServers, are based on 64 of Amazon Web Services' Trainium2 chips and will be strung together in a massive supercomputer with hundreds of thousands of chips. Apple has emerged as an early adopter of this technology, becoming the first major customer to utilize these new AI chips, with Apple executive Benoit Dupin confirming that Apple is using Trainium2 chips.

The new servers are described as "purpose-built for the demanding workloads of cutting edge generative AI training and inference," according to AWS Chief Executive Matt Garman, who made the announcement at the company's event in Las Vegas on Tuesday. AI startup Anthropic will be the first to use the massive supercomputer built from these servers, marking a significant partnership in the development and deployment of Amazon's custom chip technology.

The new AWS offerings will directly compete against Nvidia's flagship servers, which pack 72 of its latest "Blackwell" chips. While both companies offer proprietary technology for connecting their chips, Gadi Hutt, who leads business development for the AI chips at AWS, said that AWS will be able to connect a greater number of chips together than Nvidia. Additionally, AWS Chief Executive Matt Garman announced that Trainium3, the company's next generation of AI chip, will debut next year.