Cisco Introduces AI Networking Chip G300, Targets $600B Data Center Market with TSMC 3nm

Cisco (CSCO-US) announced on February 10, 2026, the Silicon One G300 AI networking chip and router aimed at high-speed data transmission in large data centers, competing directly with Broadcom (AVGO-US) and NVIDIA (NVDA-US) for $600 billion in AI infrastructure spending. The G300, manufactured using TSMC's (2330-TW/TSM-US) 3nm process, is designed to improve communication efficiency across hundreds of thousands of connections during AI training and deployment, reducing congestion and latency. According to Cisco, it can boost AI compute workloads by about 28% through microsecond-level dynamic re-routing around bottlenecks. The chip is targeted for release in the second half of 2026, emphasizing end-to-end network efficiency rather than individual component performance. The move comes as AI expands, intensifying competition in data center networking with NVIDIA's latest AI system already including a core networking chip and Broadcom's Tomahawk series.