Cisco warns agentic AI could triple enterprise network traffic by 2029

At Cisco Live in Las Vegas, Cisco Chair and CEO Chuck Robbins delivered a stark warning: the next major bottleneck for enterprise AI adoption won't be GPUs or data centers—it will be the network. As companies move from simple chatbots to agentic AI systems that autonomously act, communicate, and make decisions, network traffic could triple within three years. Robbins emphasized that every innovation from the past 40 years—including the current AI boom—depends entirely on the network's ability to carry the load. He cited Starbucks as an example, where AI-driven real-time inventory and supply chain management requires reliable, high-bandwidth networking across thousands of stores.

Jeetu Patel, Cisco's chief product officer, elaborated on the shift from spiky chatbot traffic to the sustained high demand of agentic AI. He noted that agents—trillions of them working together—generate 450% more traffic than a human user. This creates a 'networking supercycle' that demands massive upgrades in data centers, workplaces, and service provider infrastructure. Patel warned that agents operate at machine speed, swarming to handle tasks rather than clicking slowly like humans. Robbins also highlighted a parallel cybersecurity supercycle, noting that AI speeds up both defense and attacks—Cisco scanned 1.8 billion lines of code across 25 programming models in days to illustrate AI's power. For IT leaders, the message is clear: prepare the network for an AI-driven future or risk becoming a bottleneck.