By Dr. Mallik Tatipamula, FRS
Over the past five decades, communication networks have undergone four architectural shifts, not simply improving speed or capacity, but transforming the core abstraction that defines their purpose. From circuits to packets to workloads to agents, each shift not only changed networks but expanded society’s capabilities.
In the 1980s, connectivity, compute, and control (AI) technologies existed in static, centralized silos: telecom switches carried voice, mainframes processed data, and rule-based symbolic AI reasoned in isolation. Over time, each evolved independently toward distributed, composable, and dynamic systems: from circuit-switched 2G to cloud-native 5G, from mainframes to distributed cloud, and from symbolic AI to deep learning and multi-agent intelligence. Today, they are no longer evolving alone, they are converging into a unified, intelligent, and autonomous fabric. Connectivity moves data, compute processes data, and AI understands and acts on data, all within one cohesive system.
Every era has been defined by its dominant abstraction. The IP-native era used packets to unify heterogeneous media, enabling universal connectivity: any device, anytime, anywhere. The cloud-native era introduced workloads, delivering elastic programmability: any service, anywhere. Now, the AI-native era introduces agents: intelligent entities that perceive, reason, collaborate, and autonomously pursue goals—enabling any intent, anywhere.
This evolution is additive, not replacement. AI-native networks will build on programmable, orchestrable foundations, embedding meaning, context, and autonomous decision-making into the network fabric. Tomorrow’s Internet will operate as a dual stack: the existing syntactic Internet of bits, packets, and workloads, and the emerging semantic Internet of meaning, intent, and agents. This shift demands new semantic protocols, trusted agent identity frameworks, and mechanisms for context-aware collaboration.
Engineering autonomy is not about adding AI to networks, it is about architecting networks for autonomy. This requires transdisciplinary education and research to develop the next-generation workforce, and ecosystem-level innovation enabled by deep partnerships between industry, academia, and government.
As agents become first-class citizens of the Internet, we enter the Internet of Agents, not just connecting systems, but connecting purpose. This will unlock intelligent, trusted, and autonomous operations in healthcare, mobility, manufacturing, education, and smart cities.
Circuits carried our voices, packets carried our information, and workloads carried our services. Now, agents will carry our intentions, turning connectivity into capability, and networks from transport systems into partners in purpose.
