From IoT to Physical AI: How 6G Transforms Connected Things into Intelligent Environments

Written by Dr Mallik Tatipamula, FRS FREng CTO, Ericsson Silicon Valley

This article explores Physical AI as the next step beyond IoT, and why 6G and Integrated Sensing and Communications (ISAC) are critical to enabling safe, intelligent, and resilient physical systems.

In recent years, leaders in the AI and robotics community have increasingly framed the next frontier of artificial intelligence as Physical AI or Physical Intelligence: systems that do not merely process information, but perceive, reason, and act in the real world. As Nvidia CEO, Jensen Huang and others have emphasized, the transformative impact of AI will come not from disembodied models alone, but from intelligence embedded in physical environments. In parallel, the Internet of Things (IoT) emerged as a defining concept of the 5G era, connecting the physical world to the digital one at unprecedented scale by instrumenting cities, factories, infrastructure, and environments with sensors and actuators. Through IoT, the physical world became observable and, to a degree, controllable. Yet these systems remained fundamentally passive: devices sensed, reported, and occasionally acted, while intelligence largely resided elsewhere, typically in centralized cloud or edge platforms.

Physical AI represents the next and more consequential step in this evolution. Rather than merely connecting things, Physical AI embeds intelligence directly into physical systems. Robots, autonomous vehicles, drones, industrial machines, and intelligent infrastructure no longer behave as simple endpoints. They perceive their surroundings, reason continuously, learn from experience, and act autonomously in the physical world. In doing so, they introduce agency, something classical IoT systems were never designed to support.

From this perspective, Physical AI is not a reinvention of IoT, but its natural successor. IoT made the physical world visible; Physical AI makes it intelligent. This distinction matters because it fundamentally changes system and network requirements. IoT could tolerate delay, intermittent connectivity, and partial failure. Physical AI often cannot. When intelligence is embedded in moving, acting systems, latency variability can destabilize control loops, inconsistent information can lead to unsafe behavior, and failures are no longer benign, they are safety-critical.

This is where 6G becomes essential. Physical AI systems do not operate in isolation; they coordinate continuously with other agents, digital services, and humans while acting in the physical world. As a result, networks are no longer passive transport fabrics. They sit inside perception–decision–action loops, shaping what agents observe, how quickly they react, and how they coordinate. In this context, mobility management and roaming are no longer about preserving network sessions as devices move, but about maintaining cognitive continuity, ensuring that intent, learning state, trust, and safety constraints remain consistent as autonomous agents cross physical locations and administrative boundaries. An autonomous vehicle, for example, must preserve safe and policy-compliant behavior as it moves across cities or national borders operated by different infrastructure providers. Likewise, a mobile robot transitioning from a factory floor to a logistics facility must retain task intent and safety guarantees while operating across different private networks. Resilience therefore shifts from keeping connections alive to preserving stable, aligned behavior under movement and change. This shift suggests that future telecom policy and regulation must increasingly treat networks as safety-critical infrastructure for autonomous systems, not merely as connectivity utilities.

Integrated Sensing and Communications (ISAC), a key pillar of 6G, is central to this transformation. In IoT systems, sensing and communication were largely decoupled. In Physical AI systems, perception becomes part of cognition. ISAC enables networks to simultaneously sense, localize, and communicate, closing control loops at machine timescales and supporting safe, real-time interaction with dynamic physical environments.

Sub-terahertz communications further reinforce this vision by enabling extreme spatial resolution, precise localization, and ultra-low latency over short ranges. In dense Physical AI environments—such as smart factories, laboratories, and robotic spaces; THz is not just about peak throughput and more about precision, synchronization, and the tight coupling between sensing and communication that autonomous systems require.

Advances in materials and device technologies amplify this transformation. Metamaterials, battery-less sensors, and energy-harvesting devices extend Physical AI beyond conventional endpoints, embedding perception and actuation directly into surfaces, structures, and environments. These technologies enable persistent, low-maintenance sensing at massive scale, transforming physical spaces into intelligent substrates rather than collections of discrete devices.

Seen together, Physical AI extends IoT along an architectural continuum: from connected devices, to intelligent agents, to an Internet of AI Agents coordinated at scale. The role of 6G is to make this transition safe, resilient, and sustainable: embedding determinism, semantic coherence, and governance into the network itself. In this sense, Physical AI does to IoT what the mobile Internet once did to fixed broadband: not replacing it, but transforming it into something fundamentally new. This evolution also redefines resilience, not merely as the ability to recover connectivity, but as the ability to preserve correct, aligned behavior at scale, setting the stage for the next chapter in AI-native network design.