Cisco Systems on Tuesday launched the Silicon One G300 switch chip and a complementary router designed to speed data flow across large AI data centers. The products aim to capture part of the $600 billion AI infrastructure market, competing with offerings from Broadcom and Nvidia.
The G300 chip, expected to ship in the second half of 2026, is manufactured using Taiwan Semiconductor Manufacturing Co’s 3-nanometer technology. It incorporates “shock absorber” features that reroute data around network bottlenecks within microseconds, allowing AI workloads to run up to 28% faster, according to Martin Lund, Cisco’s executive vice president of the common hardware group.
Networking has emerged as a critical field for AI performance, as efficient data movement between AI chips can significantly impact training and inference speed. Nvidia’s newest systems include networking chips targeting the same market, while Broadcom continues to push its Tomahawk series. Cisco’s new offerings are designed to improve end-to-end network efficiency for data centers handling large-scale AI computation.
