Cisco Systems (CSCO) has entered the competitive AI infrastructure market with its new Silicon One G300 networking chip, directly challenging Broadcom (AVGO) and Nvidia (NVDA) for a share of the $600 billion AI spending boom. The company expects the new technology to help some AI computing jobs complete 28% faster by automatically rerouting data around network problems within microseconds.

Manufactured using Taiwan Semiconductor Manufacturing Co's (TSM) advanced 3-nanometer process, the chip features innovative "shock absorber" capabilities designed to prevent network congestion during data traffic spikes. Cisco plans to begin sales in the second half of 2026, targeting massive data centers that require efficient information transfer between AI training and delivery systems.

The launch intensifies competition in networking technology, a critical component of AI systems. When Nvidia unveiled its latest systems last month, networking chips represented one of six key components in its architecture. Martin Lund, Cisco's executive vice president, emphasized the company's focus on end-to-end network efficiency for networks with hundreds of thousands of connections.