Cisco Silicon One G300 unveiled: The new AI networking chip rivaling Nvidia and Broadcom
The Silicon One G300 is engineered for what Cisco describes as the "agentic era" of AI. It offers a staggering 102.4 terabits per second (Tbps) of bandwidth.
- Feb 10, 2026,
- Updated Feb 10, 2026 3:53 PM IST
Cisco Systems has launched a powerful new networking chip and router system. The announcement, made at the Cisco Live EMEA conference on 10 February, marks a direct challenge to industry heavyweights Broadcom and Nvidia.
The centrepiece of the launch is the Silicon One G300, a high-performance switching chip designed specifically to handle the massive data demands of modern AI workloads.
As tech giants and enterprises race to build increasingly complex AI models, the infrastructure required to connect thousands of processors has become a critical bottleneck. Cisco’s latest offering aims to eliminate these "traffic jams" within the data centre.
The Silicon One G300 is engineered for what Cisco describes as the "agentic era" of AI. It offers a staggering 102.4 terabits per second (Tbps) of bandwidth. According to the company, this allows operators to build massive AI clusters using significantly fewer switches.
As the environmental impact of AI comes under increasing scrutiny, Cisco has placed a heavy emphasis on sustainability. The new hardware, including the N9000 and 8000 series systems, features 100% liquid-cooled designs. When paired with Cisco’s advanced optics, these systems can improve energy efficiency by nearly 70%.
"We are still in the early innings of AI buildouts," said Nick Kucharewski, Senior Vice President at Cisco. While much of the initial spending has come from "hyperscalers" like Microsoft and Google, Cisco expects the next wave of investment to come from standard enterprises looking to integrate AI into their daily operations.
The release intensifies the rivalry between Cisco, Broadcom, and Nvidia. While Nvidia currently dominates the AI processor market and offers its own "InfiniBand" networking technology, Cisco is betting on the more open "Ethernet" standard, which is widely used in traditional networking.
Cisco claims its new implementation offers a 28% improvement in job completion times for AI training compared to rival offerings. To complement the hardware, the firm also unveiled AgenticOps, a suite of software tools designed to automate the management of these complex networks.
With the AI infrastructure market projected to reach $600 billion in the coming years, Cisco’s latest silicon represents a high-stakes bid to remain at the heart of the global internet, and the intelligence that now runs across it.
The Silicon One G300 is expected to be commercially available in the second half of 2026.
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine
Cisco Systems has launched a powerful new networking chip and router system. The announcement, made at the Cisco Live EMEA conference on 10 February, marks a direct challenge to industry heavyweights Broadcom and Nvidia.
The centrepiece of the launch is the Silicon One G300, a high-performance switching chip designed specifically to handle the massive data demands of modern AI workloads.
As tech giants and enterprises race to build increasingly complex AI models, the infrastructure required to connect thousands of processors has become a critical bottleneck. Cisco’s latest offering aims to eliminate these "traffic jams" within the data centre.
The Silicon One G300 is engineered for what Cisco describes as the "agentic era" of AI. It offers a staggering 102.4 terabits per second (Tbps) of bandwidth. According to the company, this allows operators to build massive AI clusters using significantly fewer switches.
As the environmental impact of AI comes under increasing scrutiny, Cisco has placed a heavy emphasis on sustainability. The new hardware, including the N9000 and 8000 series systems, features 100% liquid-cooled designs. When paired with Cisco’s advanced optics, these systems can improve energy efficiency by nearly 70%.
"We are still in the early innings of AI buildouts," said Nick Kucharewski, Senior Vice President at Cisco. While much of the initial spending has come from "hyperscalers" like Microsoft and Google, Cisco expects the next wave of investment to come from standard enterprises looking to integrate AI into their daily operations.
The release intensifies the rivalry between Cisco, Broadcom, and Nvidia. While Nvidia currently dominates the AI processor market and offers its own "InfiniBand" networking technology, Cisco is betting on the more open "Ethernet" standard, which is widely used in traditional networking.
Cisco claims its new implementation offers a 28% improvement in job completion times for AI training compared to rival offerings. To complement the hardware, the firm also unveiled AgenticOps, a suite of software tools designed to automate the management of these complex networks.
With the AI infrastructure market projected to reach $600 billion in the coming years, Cisco’s latest silicon represents a high-stakes bid to remain at the heart of the global internet, and the intelligence that now runs across it.
The Silicon One G300 is expected to be commercially available in the second half of 2026.
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine