Cisco unveiled Silicon One G300 networking chips and systems engineered to power and scale AI datacenters.

Cisco announced new Silicon One G300 technology and supporting systems designed to power AI data centers for what the company calls the agentic era. The new G300-powered Cisco N9000 and 8000 systems, combined with advanced optics and management upgrades, aim to deliver hyperscale-level performance, reliability, and efficiency for AI network builders of all sizes. According to Martin Lund, Executive Vice President of Cisco's Common Hardware Group, "As AI training and inference continues to scale, data movement is the key to efficient AI compute; the network becomes part of the compute itself. It's not just about faster GPUs - the network must deliver scalable bandwidth and reliable, congestion-free data movement." Lund added that the new technology enables "every customer to fully utilize their compute and scale AI securely and reliably in production."

The Silicon One G300 is a 102.4 Tbps switching silicon designed to power massive, distributed AI clusters. The chip uniquely offers Intelligent Collective Networking, which combines an industry-leading fully shared packet buffer, path-based load balancing, and proactive network telemetry. This architecture efficiently absorbs bursty AI traffic, responds faster to link failures, and prevents packet drops that can stall jobs. According to Cisco, Intelligent Collective Networking delivers 33% increased network utilization and a 28% reduction in job completion time versus simulated non-optimized path selection, making AI data centers more profitable with more tokens generated per GPU-hour.

The G300 is highly programmable, enabling equipment to be upgraded for new network functionality after deployment, which protects long-term infrastructure investments and supports emerging use cases. Security is fused into the hardware, allowing customers to embrace holistic, at-speed security to keep clusters running. Silicon One, introduced in 2019, is described as the industry's most scalable and programmable unified networking architecture, supporting AI, hyperscaler, data center, enterprise, and service provider use cases.

To address diverse AI deployment requirements, Cisco is introducing the next generation of N9000 and 8000 fixed and modular Ethernet systems powered by Silicon One, along with innovative optics designed for the extreme power and thermal demands of AI workloads. The company is advancing Nexus One with a unified management plane bringing together silicon, systems, optics, software, and programmable intelligence as a single integrated solution. Cisco is also introducing AgenticOps for data center networking through AI Canvas, which makes it easier to troubleshoot through guided, human-in-the-loop conversations that turn complex issues into actionable resolutions.

Cisco is partnering with strategic technology partners including AMD, DDN, Intel, NetApp, NVIDIA, and VAST to combine cutting-edge networking, compute, and storage to deliver optimized infrastructure. The Silicon One G300, G300-powered systems, and optics will ship this year.