Distributed Computing @ Scale for AI Training & Inference – Main Stage Keynote
As AI models continue to scale, both training and inference are growing rapidly in operational importance. Training pushes the limits of compute density and interconnect scale, while inference now dominates production workloads. Together, these forces are reshaping AI system architectures.
Meeting these demands requires a next-generation networking fabric that can:
- Scale up within and across a small number of racks to tightly couple XPUs for high-throughput training and low-latency inference
- Scale out across entire data centers using flat, high-performance topologies that support large-scale training and high-fanout inference workloads
- Scale across geographically distributed data centers, enabling unified AI fabrics that support million-plus-XPU training and inference environments.
We will present the latest advancements in industry initiatives—including Ethernet Scale-Up Networking (ESUN), Scale-Up Ethernet Transport (SUE-T), and Open Cluster Design for AI—and show how Ethernet is democratizing large-scale AI deployments through insights from G42 and other AI operators.
Hasan Siraj is Head of Software Products and Ecosystem at Broadcom, where he oversees software product management across the company’s networking portfolio—including Tomahawk, Trident, and Jericho—and leads product strategy for Broadcom’s AI NIC solutions. He drives Broadcom’s open networking and open-source strategy, working closely with hyperscalers, enterprises, service providers, OEMs, and ODMs to build a thriving software ecosystem around Broadcom silicon. Hasan is also facilitating the build-out of some of the world’s largest AI infrastructure deployments and expanding Broadcom’s merchant silicon into adjacent markets through strategic partnerships and new go-to-market models. Prior to Broadcom, he led Cisco’s Enterprise Routing and Switching teams. He holds a Master’s in Electrical Engineering from Cornell University and an MBA from The Wharton School.
