Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Learn about next-generation networking solutions for ultra-large AI clusters in this keynote presentation by Ram Velaga, SVP & GM of Core Switching Group at Broadcom. Discover how AI model complexity and scale are driving the need to reimagine compute, memory, and I/O architectures that span from single racks to entire geographies, interconnecting hundreds of thousands to millions of XPUs. Explore the three critical scaling dimensions: scaling up within and across racks to connect thousands of XPUs, scaling out across data centers in flat, high-performance topologies supporting tens to hundreds of thousands of XPUs, and scaling across geographically distributed data centers to create globally unified AI fabrics of million-plus XPUs. Examine key innovations powered by open, standards-based Ethernet that unlock this unprecedented scale, and review the latest advancements in Scale-Up Ethernet (SUE) architecture, including detailed product roadmap highlights for in-rack, inter-rack, and inter-data center connectivity solutions.
