Cisco Takes on Nvidia with New 102.4 Tbps AI Networking Chip
Cisco Systems has launched its Silicon One G300 networking chip, a powerful 102.4 terabit-per-second switch designed to accelerate data movement in massive AI data centers. This move positions the networking giant to compete directly with Nvidia and Broadcom in the booming AI infrastructure market.
The G300 chip was announced at Cisco Live EMEA in Amsterdam on February 10. It will power new Cisco N9000 and Cisco 8000 systems, which are expected to hit the market in the second half of 2026. The chip is manufactured using Taiwan Semiconductor’s advanced 3-nanometer technology and includes innovative “shock absorber” features designed to prevent network bottlenecks when processing large spikes of data traffic.
Cisco claims the G300 can improve AI job completion times by 28% by automatically rerouting data around network problems within microseconds. The chip maximizes GPU utilization in gigawatt-scale AI clusters used for training, inference, and real-time agentic workloads.
The new systems will be available in both air-cooled and fully liquid-cooled designs. The liquid cooling option enables nearly a 70% improvement in energy efficiency. Martin Lund, Cisco’s Executive Vice President of Common Hardware, emphasized the company’s focus on total end-to-end network efficiency rather than just raw bandwidth.
The launch comes as networking has become a critical competitive battleground in AI infrastructure. The AI buildout market is estimated to reach $600 billion as enterprises, cloud providers, and sovereign operators invest heavily in AI capabilities.
Source: Cisco Newsroom