EP 146: The State of AI Networking with Austin Lyons

In this episode of The Circuit, Ben Bajarin, Jay Goldberg, and Austin Lyons (Chip Stratt) delve into the evolving landscape of networking, particularly in the context of AI and GPU-intensive workloads. They discuss the shift from traditional networking topologies to new paradigms like "scale up," "scale out," and "scale across," defining these terms with analogies and technical explanations. The conversation covers the roles of copper and fiber optic cables in data centers, the challenges and trade-offs associated with each, and the increasing importance of SerDes (Serializer/Deserializer) technology for efficient data transmission. The hosts also touch on the economic and practical considerations of building and networking AI-centric data centers, highlighting the balance between cost, performance, and technological limitations.