Cisco raises the bar with 51.2T AI routers built for inter-data center scale

Cisco has unveiled its 8223 routing system and new Silicon One P200 chip, setting a new benchmark in AI networking. The innovations promise unmatched scalability, power efficiency, and security, enabling AI workloads to seamlessly connect across multiple data centres.

The 8223 routing system marks a leap forward for distributed AI workloads. Launched on October 9, 2025, the system is powered by the company’s newest Silicon One P200 chip, combining 51.2 Tbps Ethernet capacity with deep-buffer routing to handle the heavy data exchange AI models demand between distant data centres.

Martin Lund, EVP of Cisco’s Common Hardware Group, summed up the challenge: “AI compute is outgrowing even the largest data centres, driving the need for reliable, secure connections between facilities hundreds of miles apart.”

The 8223 directly targets this need, promising scale-across performance, connecting data centres to operate as one unified network capable of moving massive volumes of AI traffic seamlessly and securely.

Built for the AI era: scalable, efficient, and secure

The 8223 system addresses three of the biggest challenges in modern data centres: bandwidth bottlenecks, power efficiency, and space constraints. Designed as a 3RU deep-buffer routing solution, it delivers switch-like power efficiency, helping data centres manage soaring energy costs while maintaining high performance. With its compact form factor, the 8223 also conserves physical space—an increasingly valuable commodity as AI clusters grow.

Scalability lies at the heart of the 8223. It is the only fixed routing system offering 64 ports of 800G and an interconnect bandwidth that scales up to 3 Exabits per second, providing the capacity required to handle even the most demanding AI workloads. Thanks to the P200’s deep buffering capability, the system can absorb sudden traffic surges during AI training or data-intensive operations, maintaining smooth performance without network slowdowns.

Intelligence and adaptability further strengthen the 8223’s credentials. Built around programmable silicon, the system can adjust to real-time network conditions and support emerging protocols without costly hardware upgrades. This flexibility allows networks to stay current as AI traffic patterns evolve.

Security has also been integrated into every layer of the system. With line-rate encryption, post-quantum resilient algorithms, and continuous monitoring through Cisco’s observability platforms, the 8223 ensures that data moving between data centres remains both secure and reliable.

Flexibility takes centre stage.

Cisco’s approach reflects the fast-changing demands of AI infrastructure. The 8223 will initially support open-source SONiC deployments, catering to operators who prefer customizable environments. The company also plans to expand support to its IOS XR platform in the near future.

Beyond the fixed 8223 system, the P200 silicon itself will be deployable in modular and disaggregated platforms, ensuring consistent architecture across various network scales. Cisco has also announced that future Nexus systems running NX-OS will incorporate the P200, providing seamless integration across its networking portfolio. This level of flexibility allows organisations to design infrastructures that match their unique operational requirements without compromising scalability or efficiency.

Industry backing and early adoption

Industry response to Cisco’s new launch has been overwhelmingly positive. Microsoft, an early adopter of Silicon One, emphasised that the P200’s architecture simplifies its expansion across data centre, wide-area, and AI/ML environments. Alibaba Cloud described the P200 as a pivotal step in advancing its eCore architecture, citing its high bandwidth, lower power consumption, and full P4 programmability as key enablers of next-generation datacenter interconnects. The company plans to use P200-powered devices to replace traditional chassis-based routers, improving reliability and simplicity while supporting its growing AI infrastructure.

Lumen, another major network provider, noted that Cisco’s 8000 Series and Silicon One technology align closely with its vision for scalable, efficient multi-cloud connectivity. Meanwhile, Patrick Moorhead, CEO and Chief Analyst at Moor Insights & Strategy, called the 8223 “a significant step forward,” pointing to its potential to redefine distributed AI networking as data centres move toward greater interconnectivity.

A foundation for AI-driven infrastructure

Since introducing Silicon One in 2019, Cisco has steadily expanded its footprint across hyperscalers and service providers. The new P200 chip extends that vision, combining programmability with ultra-high throughput to meet the growing demands of AI-driven workloads.

As data centres continue to spread geographically and operate as interconnected networks, Cisco’s 8223 and Silicon One P200 provide the essential backbone for this transition. The company’s innovations in routing performance, power efficiency, and security position it as a central player in the infrastructure evolution, powering the global AI economy.