Amazon is developing new AI chips to improve its return on semiconductor investments and reduce its reliance on Nvidia.
Unlike Nvidia’s general-purpose tools, Amazon’s AI chip development is aimed at improving data center efficiency and giving customers customized options in the cloud AI market by optimizing chips for specific tasks.
The company plans to widely release its “Trainium 2” AI training chip next month, whose development was led by Annapurna Labs, which Amazon acquired in 2015 for $350 million. Trainium 2 is currently being tested by Anthropic. (FT)