Amazon is launching its next-generation AI chip, Trainium 2, to reduce reliance on Nvidia, enhance efficiency for AWS customers and support AI research with $110 million in cloud computing credits. (Source: Image by RR)

Trainium 2 to Reduce Costs and Increase Efficiency for Amazon and AWS Customers

Amazon is set to launch its next-generation AI chip, Trainium 2, as part of its strategy to reduce reliance on Nvidia, which currently dominates 80% of the GPU market. Dave Brown, Vice President of Compute and Networking Services at AWS, stated that while Amazon continues to support Nvidia GPUs, it believes in offering alternatives to foster competition and lower costs. By designing its own chips, Amazon aims to enhance operational efficiency for both the company and its AWS customers, providing a cost-effective alternative for training large-scale AI models.

Trainium 2, developed through Amazon’s cloud computing division, AWS, is part of a significant investment in custom-designed chips to optimize its data center operations. The innovation is spearheaded by Annapurna Labs, an Israeli chip start-up acquired by Amazon in 2015 for $350 million. The chip, expected to debut next month, will be included in Amazon’s expanding lineup of AI chips, further solidifying its role as a major player in the AI hardware space. Trainium 2 is already being tested with key partners, including Anthropic, a prominent AI start-up backed by Amazon.

In addition to Trainium 2, Amazon’s existing “Inferentia” AI chips, designed for generating responses from AI models, have already proven to be 40% cheaper than comparable solutions. These cost-saving benefits highlight Amazon’s focus on providing its customers with efficient and scalable AI solutions. By developing its own chips, Amazon aims to create a competitive edge, reducing dependency on external providers while offering advanced technology to its clients.

AWS has also announced a new initiative to support researchers by providing free computing power to develop AI models using its custom-designed chips. With an estimated $110 million in credits available, researchers can access AWS cloud data centers to leverage Trainium and other specialized chips for training and deploying AI systems. This initiative not only promotes the use of Amazon’s hardware but also supports innovation in AI development, further positioning AWS as a leader in cloud-based AI solutions.

read more at entrepreneur.com