Amazon Web Services (AWS) announced on Tuesday that it will provide $110 million in free computing credits to researchers using its custom AI chips, Trainium, aiming to boost competition with Nvidia in the AI chip market.
This program grants access to 40,000 first-generation Trainium chips, designed to support AI model development and compete with chips from Nvidia, AMD, and Alphabet’s cloud services.
Researchers from institutions like Carnegie Mellon University and the University of California, Berkeley, are already participating.
AWS, the largest cloud provider by sales, is adopting a distinct strategy from Nvidia to attract AI developers. While most AI projects on Nvidia’s chips use the company’s proprietary Cuda software, AWS plans to release the core instructions of Trainium’s architecture, allowing customers to directly program the chips.
Gadi Hutt, head of AWS AI chip development, says this approach could appeal to large-scale clients, enabling them to make custom adjustments to enhance performance across thousands of chips.
AWS faces intensifying competition from Microsoft in the cloud AI space and hopes Trainium’s customizability will attract clients needing scalable, efficient computing power for extensive AI projects.