Cisco has announced the Silicon One G300, a 102.4 Tbps switch silicon intended for large AI cluster buildouts, along with new Cisco N9000 and Cisco 8000 Ethernet systems powered by the G300, new high-density optics, and Nexus One management upgrades. Cisco positions the release around scaling AI training, inference, and real-time agentic workloads, with stated goals of improving energy efficiency, lowering operating costs, and simplifying operations as AI networking extends beyond hyperscalers to other operators and enterprises.
The company says the G300 targets “massive, distributed AI clusters” and introduces Intelligent Collective Networking, which it describes as combining a “fully shared packet buffer,” “path-based load balancing,” and “proactive network telemetry.” Cisco says these features are designed to absorb bursty AI traffic, respond faster to link failures, and prevent packet drops that can stall jobs, including over long distances. Cisco reports that, with Intelligent Collective Networking, it can deliver “33% increased network utilization” and “a 28% reduction in job completion time versus simulated non-optimized path selection,” and notes a “28% improvement in job completion time” in the news summary. Cisco also says the G300 is “highly programmable,” enabling post-deployment upgrades for new network functionality, and that security is “fused into the hardware” for “at-speed security.”
On platforms, Cisco is introducing the next generation of Cisco N9000 and Cisco 8000 fixed and modular systems with 102.4 Tbps switching and both liquid-cooled and air-cooled designs. Cisco says the “100% liquid cooled systems” enable higher bandwidth density and “a nearly 70% energy efficiency improvement,” and that customers can get “the same bandwidth in a single system that would previously have required 6 prior generation systems.” For optics, Cisco is adding 1.6T OSFP (Octal Small Form-factor Pluggable) modules targeting “AI scale out solutions” for “1.6T switch to NIC links” and “1.6T, 800G, 400G, or 200G switch to server links.” Cisco also announced 800G Linear Pluggable Optics (LPO), stating that LPO “reduces optical module power consumption by 50% compared to retimed optical modules,” and that with LPO support on new N9000 and 8000 systems, customers can “reduce overall switch power by 30%.”
Cisco also expanded its Silicon One P200-powered systems portfolio. Building on previously introduced 51.2T systems for hyperscale deployments, Cisco says new P200-powered N9000 systems plus expanded operating system support on 8223 systems deliver “scale across, data center interconnect, universal spine, and core and peer routing capabilities” to “neoclouds, enterprises, and service providers.” Cisco also introduced new 28.8T modular line cards and tied the P200 expansion to “Cisco 800G ZR/ZR+ coherent pluggable optics” to support “a common architecture across multiple roles in their network.”
For operations, Cisco says Nexus One now provides a “unified management plane” spanning “silicon, systems, optics, software, and programmable intelligence.” It also introduced “AgenticOps for data center networking through AI Canvas,” described as guided, “human-in-the-loop” troubleshooting conversations. Nexus One updates include “Unified Fabric” for deploying and adapting networks “even across multiple sites,” with Cisco N9000 systems as “the common hardware” across fabrics including Nexus Hyperfabric, plus built-in API-driven automation. Cisco also announced “AI Job Observability and Native Splunk Platform Integration,” stating that the integration is “coming in March” and is intended to let customers analyze telemetry “directly where data resides,” which Cisco calls out as relevant for “sovereign cloud deployments and compliance-sensitive environments where data locality is paramount.” Cisco states that the Silicon One G300, G300-powered systems, and optics “will ship this year.”
Source: Cisco
