en.Wedoany.com Reported - On June 5, Wang Yue, Chief Expert of China Telecom, shared his views on Physical AI and the evolution of communication networks. He believes that the breakthrough of Physical AI lies in enabling AI to move from "virtual thinking" to "real action," and that future communication networks need to evolve from a infrastructure for connecting data into an intelligent capability foundation that supports machine perception, judgment, coordination, and execution.

Physical AI targets real physical systems such as robots, intelligent vehicles, low-altitude equipment, industrial machinery, and embodied intelligent terminals. This type of AI application differs significantly from traditional mobile internet services: it requires continuous environmental perception, real-time upload of images and sensor data, inference and decision-making across cloud, edge, and terminal, and feedback of decision results into physical actions. For communication networks, this means business pressure no longer comes solely from downloading videos, browsing web pages, and accessing mobile applications, but from the comprehensive demands of massive terminals for low latency, high reliability, strong determinism, large uplink bandwidth, and edge computing collaboration. If networks remain stuck in the logic of "selling traffic," they will struggle to capture the new value brought by robots, smart factories, unmanned systems, and AI terminals.

In previous discussions on 6G, Wang Yue proposed that future access networks, in addition to ensuring basic capabilities such as coverage, speed, energy efficiency, and service quality, need to possess two core capabilities: first, achieving integrated and unified scheduling of network and computing resources to break down resource barriers between cloud and network; second, building a "cloud-edge-terminal" collaborative system by strengthening edge computing deployment to enhance real-time inference capabilities, meeting the low-latency and high-reliability requirements of AI services. This assessment aligns closely with the development direction of Physical AI. When robots, vehicles, and industrial equipment operate in real environments, they cannot rely on remote cloud for all inference over the long term. Edge nodes must take on more real-time computation, state synchronization, and local decision-making tasks, and communication networks must transform from transmission channels into system platforms that coordinate computing power, models, data, and connectivity.

The uplink will become a critical capability entry point in the Physical AI era. In the past, mobile networks were primarily optimized for downlink experiences, with user downloads of videos, images, and applications dominating traffic. In the era of AI terminals and embodied intelligence, devices need to continuously upload camera feeds, sensor data, environmental states, task logs, and model interaction information. Unmanned vehicles, inspection robots, AR glasses, industrial robotic arms, and low-altitude drones may all generate large amounts of uplink data simultaneously. If network capabilities cannot adapt to this shift in traffic patterns, real-time collaboration and cloud-edge synergy for AI terminals will be constrained. The future value of operators will extend beyond simply providing connectivity to include uplink assurance, edge inference, computing power scheduling, AI capability openness, and industry-specific services.

This imposes new business model requirements on operators. Physical AI will not automatically retain value on the network side. If operators only provide underlying channels, more revenue may flow to cloud vendors, chip manufacturers, model providers, and terminal platforms. China Telecom's emphasis on directions such as AI-RAN, communication-intelligence integration, and computing-network collaboration essentially seeks a deeper position for communication networks in the AI value chain. Future networks need to understand the operational characteristics of AI services, be capable of sensing terminal tasks, matching edge computing resources, scheduling wireless resources, ensuring key service experiences, and providing network capabilities to industry customers such as robotics, industrial, transportation, low-altitude economy, and smart cities through open interfaces.

Physical AI will also drive changes in 6G network architecture in reverse. The pace of AI development is faster than traditional communication generational cycles, with model capabilities, terminal forms, and application modes rapidly iterating. If network architecture continues to evolve slowly on a fixed cycle, it may miss the window of opportunity brought by the explosion of AI applications. For scenarios such as embodied intelligence, robots, and intelligent vehicles gradually emerging after 2027, communication networks need to form AI-native design concepts earlier, integrating perception, communication, computing, data, and intelligent scheduling into a unified framework. Wang Yue's assessment of Physical AI points to a larger industry shift: communication operators need to upgrade from network builders to participants in the intelligent infrastructure and industry capability platforms of the AI era.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com