AI models are growing at an unprecedented pace, driving exponential increases in compute demands. As a single GPU can no longer handle these workloads, the challenge has become one of distributed computing—requiring highly scalable, high-performance network infrastructure.
To address this, both scale-up and scale-out architectures are being adopted, with Ethernet emerging as the preferred solution thanks to its scalability, resilience, and open ecosystem. While Ethernet is already the standard for scale-out networks, it is now proving equally powerful for scale-up environments.
At the OCP EMEA Summit, Broadcom introduced its Scale-Up Ethernet Framework, a significant contribution to the open hardware community. This presentation will outline the key components of this framework, showcase innovative products built on this architecture, and highlight how the community is collaborating to advance the future of AI infrastructure.
