Cisco unveils its Silicon One G300 switch chip, made using TSMC's 3nm tech, to take on Broadcom and Nvidia in AI infrastructure, set to go on sale in H2 2026
Cisco Systems (CSCO.O) on Tuesday launched a new chip and router designed to speed information through massive data centers that will compete …
Context & Ripple Effects
Cisco’s G300 extends a long-running effort to sell standalone networking silicon to major data-center customers, rather than confining the company to finished network equipment. Its recent P200 chip and routing-system launch showed the same push toward higher-capacity links across data centers.
The new chip also follows Cisco’s broader AI-networking buildout, including Nexus HyperFabric AI clusters developed with Nvidia. That makes the G300 notable as a bid to compete in a layer of AI infrastructure where Cisco also partners with a named rival.
First-order effects
- Cisco adds a 3nm Silicon One switch chip and accompanying router to its AI-data-center portfolio, with commercial availability planned for the second half of 2026.
- Broadcom and Nvidia gain another named competitor for the networking and infrastructure silicon used to move data inside large AI deployments.
Second-order effects
- Cisco can present customers with a broader in-house networking-silicon option alongside its systems, increasing pressure on incumbent suppliers to defend design wins through performance, integration, or pricing.
- The use of TSMC’s 3nm process ties Cisco’s networking roadmap more directly to leading-edge foundry capacity, extending AI-infrastructure demand beyond compute chips.
Third-order effects
- If Cisco converts its chip roadmap into deployments, AI data-center networking could become less concentrated around a small set of silicon vendors and more defined by competing integrated chip-and-system platforms.
- The move is another sign that AI infrastructure competition is spreading from accelerators to the network fabric that determines how effectively large systems operate.
The trend: AI infrastructure is broadening into a contest for high-performance networking silicon, as vendors seek to control more of the data-center stack around AI workloads.