AI’s infrastructure problem keeps moving.
First, companies needed faster chips. Then they needed more GPUs, bigger data centers, and enough electricity to keep those AI factories running. At Cisco Live in Las Vegas, Cisco Chair and CEO Chuck Robbins argued that the next major constraint is the network itself.
Robbins said AI traffic could triple over the next three years as enterprises move from chatbots to agentic systems that act, communicate, and make decisions at machine speed. For IT leaders, the message was clear: AI ambitions may soon depend less on access to models and more on whether the network can carry the load.
“Every technical innovation over the past 40 years and all the current buzz about GPUs, inferencing, agentic AI, and the latest models — it is all dependent on the network,” Robbins explained. “We expect at least a tripling of networking traffic over the next three years due to the evolving capabilities and consumption demands of AI.”
Starbucks and networking
Robbins concluded his keynote by introducing Brian Niccol, chairman and CEO of Starbucks Coffee Company. They chatted on stage about how his organization uses AI and what it is doing to expand its networking capabilities.
“The original function of the network in Starbucks was to provide customers with a reliable Wi-Fi network in every store as a way to attract customers,” said Niccol. “Now we need the network to facilitate forecasting, scheduling, and supply chain management.”
He gave an example of how the company is transitioning from sending cases of coffee, syrup, and other supplies to stores. Instead, it is sending smaller amounts based on real-time and near-real-time knowledge of in-store inventory, traffic volumes, supply chain availability, and predicted consumption.
But this is only possible with AI and improved networking. More information means better prediction about what people will order, when, and what the inventory needs will be.
“If it’s Tuesday on the East Coast and 80°F, we have much better predictability on how many bottles of water a specific store will need,” Niccol said.
Agentic coworkers supercharge the network
Jeetu Patel, president and chief product officer at Cisco, continued Robbins’ theme by explaining the implications of the market pivoting from AI chatbots to agentic AI coworkers that can act autonomously. This is fundamentally shifting infrastructure requirements.
“Chatbots have a very spiky network demand pattern, whereas agentic AI and physical AI have a high, sustained demand signal,” he said. “With agents operating at machine speed, the volume of traffic will be meaningfully higher than before. Humans click, but agents swarm.”
Patel foresees a near future in which trillions of agents will work together to assist humans in decision-making and task execution. These agents are far more consumptive — generating 450% more traffic than a human, according to Patel.
“This should be looked upon as a networking supercycle that will require massive data center, workplace, and service provider upgrades to prepare the network and the underlying infrastructure for this upsurge in agents.”
More must-read AI coverage
- SS&C Intralinks DealCentre AI vs. Datasite: Which platform is built for the future of dealmaking?
- SS&C Intralinks FundCentre AI vs. Juniper Square: Which platform better supports modern private markets fund managers?
- Why Data, Not Models, Determines AI Success
- The Rise of the AI-Native Factory: How Physical AI Is Transforming Manufacturing
Security supercycle
In tandem, a cybersecurity supercycle is upon us, unveiled by Mythos in April of 2026.
“AI is changing the speed of cyber defense, empowering our adversaries at a pace we have never seen,” said Robbins.
He gave an example of AI’s power: Cisco scanned 1.8 billion lines of code across 25 programming models in a matter of days. This, he said, would have been eight years in the past. Now the bad guys have access to that kind of speed of action. Hence, many Cisco customers are clamoring for an air-gapped, AI-ready, secure, and resilient infrastructure.
Liz Centoni, executive VP of Customer Experience at Cisco, hammered this point home during the second day’s keynote.
“40% of exploits directly impact end-of-life devices, and AI now can map your network in minutes to find vulnerabilities no one has ever spotted before,” she said.
The solution? The show touted Cisco IQ, which is built into Cisco Cloud Control. It enables IT and CISOs to see their entire landscape with clarity, inventory every asset, know their current security status, and take automated remedial actions.
Cisco’s larger argument is that AI infrastructure is entering a new phase. The first wave was about compute. The next one may be about whether enterprises can connect, monitor, secure, and automate everything that compute makes possible.
For IT teams, that means AI readiness is no longer just a data center or cloud question. It is becoming a network architecture question.
Also read: Gartner SRM 2026 shows why cybersecurity leaders are shifting from pure prevention to resilience, identity, and AI agent governance.