Cisco has unveiled the Cisco Silicon One G300, a 102.4 Tbps switching silicon built for massive AI cluster deployments. The new silicon will power Cisco N9000 and Cisco 8000 systems. These systems push the boundaries of AI networking in data centers.

The devices feature innovative liquid cooling and support high-density optics. This ensures customers achieve maximum efficiency from their GPU investments. In addition, Cisco enhanced Nexus One to simplify AI network management. Enterprises can now operate networks on-premises or in the cloud. This reduces complexity and helps scale AI data centers faster.

Jeetu Patel, President and Chief Product Officer at Cisco, said, “We are leading performance, manageability, and security in AI networking. We innovate across the full stack, from silicon to systems and software. This supports every customer type from hyperscalers to enterprises shifting to AI-powered workloads.”

The Silicon One G300 introduces Intelligent Collective Networking. It combines a shared packet buffer, path-based load balancing, and proactive network telemetry. This improves performance and reliability for large-scale AI clusters. The technology absorbs bursty AI traffic, reacts quickly to link failures, and prevents packet loss. Cisco reports a 33% increase in network utilization and a 28% reduction in job completion time.

The silicon is highly programmable. Devices can receive new network functionality even after deployment. Security is integrated at the hardware level, allowing at-speed protection for AI clusters.

Nexus One now offers a unified management plane. It integrates silicon, systems, optics, software, and programmable intelligence. This gives organizations flexibility in running AI workloads efficiently.

The Cisco Silicon One G300 and G300-powered systems will ship later this year. This silicon is designed to set new standards for AI networking in data centers.

Related post

View all