Barcelona — At a press roundtable during MWC 2026, Cisco President and Chief Product Officer Jeetu Patel outlined what he described as a pivotal transition for the telecom industry: the move from chatbot-driven AI to autonomous agents operating continuously across networks.
Patel was joined by Guru Shenoy (Senior Vice President, Provider Connectivity), Bill Gartner (Senior Vice President and General Manager, Optical Systems and Optics), and Kevin Wollenweber (Senior Vice President and General Manager, Data Center and Internet Infrastructure). But it was Patel who framed the core message: AI is entering a new phase that will fundamentally change infrastructure requirements — and create new opportunities for service providers.
1. From Chatbots to Autonomous Agents
Patel argued that the first wave of AI — conversational interfaces and coding assistants — introduced AI to the mainstream. The next phase is agentic: systems that execute tasks autonomously for extended periods, not just respond to prompts.
Unlike chatbots, which generate spiky traffic patterns, agents create sustained, continuous demand for compute and bandwidth. As these agents multiply across enterprises, traffic shifts from intermittent bursts to persistent machine-to-machine activity.
For networks, that means higher baseline utilization, more east-west data flows, and greater architectural complexity.
2. A New Infrastructure Model
- Scale-up: GPU-to-GPU connectivity inside racks
- Scale-out: clustering racks inside large data centers
- Scale-across: interconnecting data centers across metro and regional distances
Power constraints will prevent everything from being built in a single location. Instead, data centers will need to operate as distributed but tightly synchronized clusters — sometimes hundreds of kilometers apart.
In that context, Cisco is positioning itself as a full-stack infrastructure provider for the AI era: silicon, systems, optics, networking, security, and observability.
3. Telcos and the Monetization Question
A central question at MWC: will AI simply consume more bandwidth, or can operators capture new value?
Patel suggested that service providers can differentiate by offering:
- Assured performance and low latency
- Security and identity controls for AI agents
- Sovereignty guarantees for data and compute
- Distributed inference capabilities closer to users
As inference workloads grow — potentially outpacing training over time — telcos with fiber assets, metro presence, and edge infrastructure may find new roles in hosting and transporting AI workloads.
4. Trust and Sovereignty as Structural Themes
Patel repeatedly returned to trust. AI systems are probabilistic and autonomous; securing them requires new validation, runtime controls, and observability across the stack.
He also described sovereign AI as a sustained global trend. Countries and enterprises will increasingly want control over their own AI infrastructure and “token generation” capacity — not just rely on centralized hyperscale platforms.
For Cisco, this reinforces the need for infrastructure that enforces both security and policy constraints across data at rest and data in motion.
The Bigger Picture
Patel acknowledged short-term volatility in AI infrastructure investment cycles, but argued that the long-term demand for compute, networking, and secure connectivity is likely underestimated if autonomous agents become embedded across enterprises. Tthe shift to agentic AI is not incremental. It changes traffic patterns, architecture, and economics. And for telcos willing to modernize, it may offer a second chance to move beyond commoditized connectivity toward differentiated AI-era infrastructure.
