Cisco Expands AI Infrastructure Push with New High-Performance Networking Platform
The new silicon will power upcoming Cisco Nexus 9000 (N9000) and Cisco 8000 systems, purpose-built for the extreme performance, density, and efficiency demands of AI training and inference workloads.
Cisco has announced a major step in its strategy to turn the network into an AI innovation platform with the launch of the Silicon One G300, a 102.4 terabits-per-second (Tbps) switching chip designed for massive AI cluster deployments. The new silicon will power upcoming Cisco Nexus 9000 (N9000) and Cisco 8000 systems, purpose-built for the extreme performance, density, and efficiency demands of AI training and inference workloads.
The G300 introduces Intelligent Collective Networking, combining a fully shared packet buffer, path-based load balancing, and proactive telemetry to improve performance across distributed AI clusters. Cisco says this enables 33% higher network utilization and 28% faster job completion times compared to non-optimized networks, helping organizations maximize GPU output and improve AI data center economics. Security is embedded directly into the hardware, and the chip is highly programmable, allowing future upgrades without replacing infrastructure.
Cisco also introduced liquid-cooled and air-cooled 102.4T systems, delivering nearly 70% energy efficiency improvement compared to prior generations. New 1.6T optics and 800G Linear Pluggable Optics (LPO) further boost bandwidth while reducing optical power consumption by up to 50%, lowering overall switch power use by 30%. These advances target hyperscalers, enterprises, neoclouds, and service providers building AI-ready infrastructure.
On the software side, Cisco enhanced Nexus One, a unified AI networking platform that simplifies operations across on-premises and cloud environments. New features include AI job observability, Splunk integration, API-driven automation, and AgenticOps capabilities that use guided AI-driven troubleshooting to streamline network management. Cisco says the approach reduces operational complexity and supports sovereign and compliance-sensitive deployments.
The Silicon One G300, associated systems, and optics are scheduled to ship this year. Cisco is working with partners including NVIDIA, AMD, Intel, NetApp, DDN, and VAST to deliver integrated AI infrastructure stacks.
