As AI transforms industries, revolutionizes the global workforce, and drives new levels of digital innovation, the demands on network infrastructure have never been greater. Meeting the requirements of the AI era means that a new level of hyper-resilience is vital-delivering unprecedented bandwidth, energy efficiency, performance, and automation-across data center, core, and edge environments.
At Cisco, we're leading the way with a validated, adaptive, and unified architecture approach to AI networking, introducing breakthrough infrastructure and management solutions that enable providers to scale seamlessly, optimize performance, and unlock the full potential of next-generation AI workloads.
How do we meet the demands of the new era of AI networking?
To move towards hyper-resilience and meet the demands of AI, a few things are essential:
- High-bandwidth switches to handle the transition to 1.6-terabit speeds.
- Advanced thermal management, including liquid cooling, to handle unprecedented power density.
- Open networking architectures that support power-efficient optics.
- Deep-buffer routers that absorb bursts and prevent packet loss for connecting distributed data centers.
- Carrier-grade software that scales seamlessly across the core and the edge.
Today at Cisco Live in Amsterdam, Cisco is announcing several additions to our Agile Services Networking architecture to support AI connectivity requirements.
First, the 8132/8133 switches deliver 102.4T of capacity, powered by Silicon One G300, featuring 64 ports of 1.6T Ethernet in a single fixed system. To meet the extreme power demands of AI, these systems are available in both advanced liquid-cooled and air-cooled configurations, supporting Cisco OSFP 1.6T optics. This ensures maximum performance, power efficiency, and reliability, whether you are refreshing a traditional row or building a next-generation liquid-cooled AI pod.
Second, we're announcing the 8122X-64EF-O, a G200-based 64x800G switch. Running the SONiC OS, this system features support for Cisco 800G Linear Pluggable Optics (LPO). By utilizing LPO, customers can reduce power consumption by up to 30%, enabling greater port density per rack and providing a more efficient and sustainable path for 800G networking.
Third, we have new use cases for the 8223 fixed routers, which were previously announced in October (see prior announcement), and now feature expanded applicability across the network. Powered by a Cisco Silicon One P200, and running the industry-leading IOS XR software, these systems are built for massive scale. In addition to supporting scale-across networking, the 8223 now supports Core, Data Center Interconnect (DCI), and Peering use cases, bringing proven reliability and deep-buffer performance with IOS XR to the most-demanding network environments.
Finally, we are introducing the new P200-powered 88-LC2-36EF-M line card, which delivers a massive 28.8T of capacity. Available for both our 8-slot and 18-slot modular systems, this line card enables up to an unprecedented 518.4T of total system bandwidth, the highest in the industry. When paired with Cisco 800G ZR/ZR+ coherent pluggable optics, these systems can easily connect sites over 1,000 kilometers apart, providing the high-density performance needed for modern data center interconnects and core routing.
In a world where AI-driven innovation is reshaping every industry, the ability to build hyper-resilient, scalable, and efficient networks has never been more critical. With these advancements-spanning breakthrough switching and routing platforms, advanced thermal management, and open, power-efficient architectures-Cisco is not just responding to the demands of AI but setting new standards for what's possible with AI infrastructure. We are empowering organizations to unlock the full potential of AI-confidently and sustainably.
Scalable performance measurement for AI workloads now embedded in hardware fabric
The need for scale and per-path visibility becomes even more critical in emerging environments that require hyper-resilience, such as large-scale AI data centers. AI workloads are highly sensitive to latency variation and congestion and often rely on deterministic path selection across massive IP transport fabrics. In these environments, understanding performance per individual path-not just per aggregate-is essential.
Cisco, recognizing these evolving demands, has pioneered Integrated Performance Measurement (IPM). This innovative approach embeds performance measurement directly into the network hardware fabric, delivering a new era of scale, richness, and cost-efficiency in network performance monitoring.
IPM makes it possible to measure every path, one by one, in large, complex networks. Importantly, IPM can be deployed in existing networks to dramatically improve visibility compared to legacy probing approaches. With SRv6, IPM becomes even more powerful: SRv6 provides deterministic traffic steering, while IPM provides deterministic, per-path measurement aligned with business intent and performance targets.
Building upon advances in Silicon One hardware, observability, and Cisco IOS XR segment routing performance monitoring, IPM represents the next evolution in performance visibility, introducing one-way hardware-based generation and collection of up to 14 million probes per second across all Equal-Cost Multi-Path (ECMP) routes. With IPM, latency, loss, and liveness metrics are captured within a single probe packet.
The real value of IPM to network operations teams is that it is not a standalone hardware feature, but a foundational element within a powerful ecosystem designed for holistic network assurance, embedded SLA monitoring, and automation. When combined with assurance and routing analytics (BGP-LS path information), IPM provides extensive visibility into IP transport networks, enabling intelligent path optimization, robust SLA validation, and the real-time insights necessary for closed-loop automation and AIOps.
Cisco Crosswork Network Automation enhanced with agentic AI
Building on this robust foundation of real-time, per-path network intelligence, combined with trusted, cross-domain data to train models, we're extending intelligence and AI directly into the heart of network operations.
With the introduction of a multi-agentic AI framework, organizations can move beyond basic scripted automation towards human-assisted autonomy and on their way to full autonomy. This will enable a self-healing, self-optimizing infrastructure that doesn't just report issues, but understands and adapts to prevent or fix issues as they arise. Through multi-agentic AI frameworks, network teams can manage complex, multi-vendor networks with greater speed, accuracy, and confidence.
Today, we are introducing enhanced agentic AIOps capabilities that tackle the most persistent challenges in managing modern networks:
- Risk factor identification: In large-scale networks, multiple small, benign factors can combine to create a harmful situation that causes network issues like connection failures. These failures are hard to detect, especially in complex environments. AI analyzes network event data and operational factors to identify common factors during these network events. This reduces mean time to identify (MTTI) and helps prevent widespread downtime by catching issues early.
- Configuration drift detection: Small, untracked changes can silently undermine network stability by causing configurations to drift from their intended state. AI learns what normal looks like, so it can intuitively flag unexpected variances without needing a golden config baseline, highlighting what the correct configuration should be. This proactive approach reduces operational risks and downtime and is complementary to existing configuration management processes.
Our vision for the autonomous network is one that is open and extensible. With the release of a developer SDK (currently in beta), customers will be able to design their own capabilities, building unique operational requirements directly into the AI framework to meet innovation goals.
By aligning breakthrough hardware, integrated performance measurement, and AIOps capabilities, we're creating the blueprint to drive ease and efficiency across the entire network lifecycle. This isn't just about managing devices, but about creating a resilient, intelligent foundation that powers the AI era with zero friction.
See these innovations live in the World of Solutions this week
This week at Cisco Live in Amsterdam, these AI networking innovations will be on display in the Secure Global Connectivity zone of the World of Solutions. Stop by to see a wide range of demos, from SRv6 and non-terrestrial network architecture to the latest provider security solutions, live and in person.
Don't miss the Product Wall, where the newly announced platforms and latest Silicon One chipsets are on display.
Get AI-ready connectivity and learn more about our architecture blueprint
Additional resources:
- Cisco Announces New Silicon One G300
- Learn more about Cisco Optics
- Explore Integrated Performance Measurement (IPM) in the latest IOS XR data sheet and IPM technical documentation
- Cisco Crosswork Multi-Agentic AI Framework At-a-Glance
Related blogs:
- Networking for the Agentic Era: Cisco Unveils New Innovations in Scale and Simplicity