At the Networking for AI Summit, Cisco executives outlined how the company is evolving its networking portfolio and AgenticOps to support AI workloads across the data center, campus, and branch. The conversations with Murali Gandluru, VP Product Management for Data Center Networking, Matt Landry, VP of Product Management for Cisco Wireless, and Grant Shirk, Senior Director of Product Marketing for Networking, underscored how enterprises must prepare for architectural shifts, operational transformation, and the new traffic patterns introduced by AI training, inference, and agentic workloads.
Read on for a summary of each session and the embedded video of each.
Data Center Networking for AI
Murali Gandluru emphasized that AI training, inference, and agentic workloads are fundamentally reshaping data center architectures, creating demands for massive east-west bandwidth, deterministic latency, and cross-site connectivity. Cisco sees a unified Ethernet fabric as the foundation for democratizing AI infrastructure, enhanced by technologies such as 800G+, RDMA/RoCEv2, DPUs, and ongoing SiliconOne innovations. To maintain predictability and accelerate troubleshooting across GPU- and CPU-based workloads, Cisco integrates telemetry, automation, and lifecycle management directly into the fabric. Security is also built into the architecture, with centralized policy design and distributed enforcement ensuring that performance and protection coexist. Finally, Cisco supports customers with validated designs and blueprints tailored by industry, including finance, healthcare, and autonomous vehicles, to accelerate deployment of scalable, secure AI infrastructure.
Bridging Data Center and Campus/Branch
In a summit panel discussion, Murali Gandluru and Matt Landry described how enterprises must unify data center, campus, and branch domains under a consistent operational and security framework. Within the data center, GPU-to-GPU traffic patterns are reshaping fabrics, requiring silicon-level differentiation, advanced observability, and real-time troubleshooting at scale. At the campus and branch, inference and machine-to-machine workloads are transforming traffic patterns, driving up wireless and edge demands, and underscoring the need for agentic operations to offset staffing and budget constraints. Customers continue to prioritize digital resilience, reliability, predictable latency, and security infused into the network, and Cisco innovations such as liquid-cooled switches and HyperShield were highlighted as enablers of scale and resilience. Real-world deployments in financial services, autonomous vehicle factories, and media production environments illustrate the breadth of vertical use cases AI networking must support.
Campus and Branch Networks in the AI Era
Grant Shirk extended the conversation by framing the campus as the frontline for AI adoption. With the proliferation of IoT devices, AI workloads, and machine-to-machine interactions, enterprises must rethink identity, security, and scalability in access networks. Shirk stressed that organizations need both networking for AI (high-capacity, low-latency infrastructure) and AI for networking (agentic AIOps via AI Canvas to manage complexity at machine speed). Building trust remains critical, with human-in-the-loop oversight ensuring context, transparency, and auditability. He shared examples of AI’s impact in action: a pharmaceutical company whose on-prem models overwhelmed a 400G campus core, and a quick-service restaurant using small language models at drive-throughs to improve service and address staffing shortages. Operationally, Shirk advised enterprises to adopt platform-based approaches for unified visibility, shift toward fabric-based designs, and prepare for sustained inference traffic, which will increasingly define AI-era campus workloads.
Our ANGLE
Cisco’s perspective on Networking for AI emphasizes a holistic, cross-portfolio approach spanning data centers, campuses, and branches. The strategy combines Ethernet innovation in the data center, agentic operations to manage complexity across domains, and fabric-based campus designs to support AI workloads at the edge. By embedding security, lifecycle management, and validated blueprints across its portfolio, Cisco is positioning itself to help enterprises deploy AI-ready networks that are scalable, resilient, and operationally trustworthy.
For more information on Cisco solutions, please visit their website.
