Martin Lund on Why AI Can’t Exist Without Networking

Why It Matters

The insight underscores that AI adoption hinges on networking capabilities, making Cisco's silicon and unified architecture critical for enterprises seeking scalable, secure AI deployments.

Key Takeaways

• AI depends on robust, secure networking infrastructure across edge devices
• Cisco's P200 silicon delivers up to 15 exabytes bandwidth across data centers
• Future AI workloads require unified IP/Ethernet architecture for scalability
• Power efficiency, performance, and security are core design priorities
• Edge AI and robotics will demand new networking and security models

Summary

In a recent interview, Cisco’s chief architect Martin Lund argued that artificial intelligence cannot function without a robust networking foundation, emphasizing that AI now permeates everything from Wi‑Fi access points to edge cameras and data‑center GPUs.

He highlighted three pillars—reliability, performance, and security—as non‑negotiable for the next wave of agentic AI workloads, noting that Cisco’s own silicon strategy, launched in 2019, culminated in the P200 chip capable of moving up to 15 exabytes of data across training clusters and between data‑center racks.

Lund quoted, “Without networking, there is really no AI,” and added that a single, unified IP/Ethernet architecture underpins both routing and switching, allowing Cisco to control innovation and optimize power usage, performance, and delivery timelines.

The message signals to enterprises and service providers that investing in high‑capacity, secure networking hardware is essential to unlock AI at scale, positioning Cisco’s end‑to‑end silicon and architecture as a strategic advantage in the rapidly expanding AI ecosystem.

Original Description

AI may dominate the conversation - but it runs on something far less visible. In this conversation, Steve Saunders speaks with Martin Lund, Executive Vice President of Cisco’s Common Hardware Group, about the infrastructure demands sitting beneath the AI boom.

From data centres to the edge, AI is fundamentally reshaping network requirements. Traffic patterns are changing, bandwidth demands are accelerating, and the number of connected endpoints is set to grow exponentially - driven by IoT, robotics, and what Martin describes as the next wave of “physical AI.”

At the same time, the scale of connectivity required between systems is reaching unprecedented levels. Cisco’s latest silicon is designed to handle massive data flows across racks, data centres, and geographies - all within a unified architecture that avoids the fragmentation of specialised networks.

But performance isn’t the only challenge. Power consumption, security, and manageability are becoming critical constraints as networks push against the limits of physics. The result is a new era of infrastructure design - where networking is no longer a supporting layer, but the foundation that makes AI possible.