Amazon is working on its own artificial intelligence chips to lessen its reliance on NVIDIA, according to a report from the Financial Times. The tech giant has a history of designing in-house processors for data center tasks, and its latest move is an extension of its 2015 investment in a chip design startup.
Amazon is expected to unveil more details about its custom AI processors next month, alongside updates about its Trainium chip series. These chips, created by Amazon’s Annapurna Labs, are already being used by Anthropic, a rival to Microsoft-backed OpenAI. Anthropic, Amazon’s primary AI partner, provides the e-commerce and cloud computing giant with access to its Claude foundational AI model.
According to the Financial Times, this move from Amazon is designed not only to reduce reliance on Nvidia, but to save costs as well. The tech giant has experience in designing custom chips. Its purchase of chip design startup Annapurna has allowed Amazon to create a series of processors to lower costs associated with using AMD and Intel chips for standard data center tasks.
These chips, known as Graviton processors, are paired with Amazon’s custom AI chips called Trainium. Trainium is designed to work with large language models. Amazon unveiled Trainium2 in November 2023.
The tech giant’s Annapurna Labs is also leading the development of these custom chips that will help Amazon reduce reliance on Nvidia.
While specific details about these new chips are limited, Amazon is expected to share more information at its event next month, where it will discuss its Trainium2 chips.