Tech Desk 
Nov 12, 2024
Amazon is reportedly preparing its own artificial intelligence (AI) chips, as it pursues gains on its multibillion-dollar semiconductor investment and to reduce its reliance on Nvidia, according to the Financial Times.
Amazon has a previous record of developing in-house chipsets for data centre tasks, and it once again seems to be extending its 2025 investment in a chip design startup.
The progressive initiative is led by Annapurna Labs, an Austin-based chip start-up that Amazon purchased in the beginning of 2015 for $350mn.
Read more: Ola Electric announces plans to roll out first AI chip in India
Annapurna is likely to reveal more details regarding the custom AI processors and Trainium chip series ‘Trainium 2’ in December.
Interestingly, Anthropic AI, a rival to OpenAI is already utilising the Trainium chip lineup developed by Amazon’s Annapurna Labs.
The move is not only aimed at decreasing reliance on Nvidia but also at lowering costs. By acquiring Annapurna, Amazon has been able to develop its own chip lineup — such as the Graviton processors for standard data center tasks — thereby reducing its dependence on AMD and Intel chips. These chips are coupled by Trainium, which is specifically designed for large language models.
Furthermore, Amazon and Amazon web services (AWS) aim to outpace Nvidia, one of the world’s most valuable companies. However, other details regarding the latest chips are still under wraps, which are yet to be revealed by Amazon ahead of its launch.
