Amazon Invests $110M in AI Research of Trainium Chips

Amazon is making a major investment in AI research as the company looks to reduce its reliance on Nvidia and develop its own in-house chips.

The AWS Trainium is a custom-built machine learning (ML) chip designed for deep learning training and inference tasks.

The investment would support generative AI research at universities using Trainium chips. The program, dubbed Build on Trainium, would provide researchers the ability to develop new AI architectures, machine learning libraries and performance enhancements for large-scale distributed AWS Trainium UltraClusters, groups of AI accelerators working in unison on complex computational tasks, Amazon said. 

“The program caters to a wide range of AI research, from algorithmic advancements to increase AI accelerator performance, all the way up to large distributed systems research,” Amazon said in a statement. “As part of Build on Trainium, AWS created a Trainium research UltraCluster with up to 40,000 Trainium chips, which are optimally designed for the unique workloads and computational structures of AI.”

Amazon said any AI advances created through the program would be open-sourced allowing researchers and developers to continue to advance their innovations. 

In August, Amazon announced Amazon’s $4 billion investment in Claude developer and OpenAI competitor Anthropic. 

Related:AWS Showcases Pathways to Market With Generative AI

The Build on Trainium program is also dedicating funding for new research and student education and Amazon plans to conduct multiple rounds of research awards, which would give selected proposals AWS Training credits, as well as access to the large Trainium UltraClusters for their research. 

Carnegie Mellon University’s Catalyst research group is participating in the program.

“AWS’s Build on Trainium initiative enables our faculty and students large-scale access to modern accelerators, like AWS Trainium, with an open programming model,” said Todd C. Mowry, a professor of computer science at Carnegie Mellon University. “It allows us to greatly expand our research on tensor program compilation, ML parallelization, and language model serving and tuning.”