Amazon is set to launch its new AI chip, Trainium 2, in December, as part of a strategy to reduce its reliance on Nvidia. Developed by Annapurna Labs, acquired by Amazon in 2015, Trainium 2 is designed to improve the efficiency of AI training models and cut costs for Amazon Web Services (AWS) customers.
The chip is already being tested by companies like Anthropic and Databricks and offers up to 40% savings on AI model inference compared to Nvidia's offerings. This move is part of Amazon's broader effort to build custom chips for its cloud infrastructure, alongside existing chips like Graviton and Inferentia.
While Nvidia dominates the AI chip market, Amazon’s push for custom solutions signals a desire for more competition and choice in the industry. AWS aims to provide alternatives to Nvidia while maintaining compatibility with its chips, hoping to offer lower-cost, specialized AI solutions for customers.
Related News