As generative AI drives demand for training and inference workloads, the tech industry has turned to innovations in chip, board and server design to provide the infrastructure and compute power to meet these demands. Annapurna Labs, which Amazon acquired in 2015, plays a key role in AWS's ability to innovate in this area, having developed the Nitro, Graviton, Inferentia, and Trainium families of processors. As a follow-up to RedMonk's tour of Annapurna Labs in Austin, TX, Senior Analyst Kelly Fitzpatrick sits down with Chetan Kapoor (Director of Product Management for the Amazon EC2 Accelerated Computing Portfolio) to discuss how the custom silicon developed at Annapurna Labs is a differentiator for AWS, some of the differences between the infrastructure needs of training and inference workloads, and what AWS is doing to support developers building genAI applications.
This RedMonk Conversation was originally published in video form on February 20, 2024.