China Telecom's Li Xuelong: Industry Shifts from Bit Traffic to Token Value Operations

Among the "six networks," the next-generation communications network and the computing power network both fall within the information and communications industry's core responsibilities. The next-generation communications network is a further upgrade and evolution of the traditional communications network, while the computing power network mainly provides computing services for artificial intelligence. The two sit within the same system and have a foundation for integrated development and mutual support. The next-generation communications network can provide connectivity for computing access and inter-computing links, while the computing power network supplies computing power to the next-generation communications network for AI-based network management and AI-enabled intelligent networking. From an infrastructure construction perspective, the two networks can be planned as an integrated whole, achieving integrated construction and joint deployment of network management and platform-layer capabilities.

Tokens, crossing beyond bits, have become the smallest unit of information flow in the agent era, and operators are stepping up token operations. The industry is undergoing a transition from traditional bit traffic operations to token value operations. Operators no longer merely serve as pipeline movers of data bits; they are building a closed loop of token "production, scheduling, transmission, consumption and settlement" based on a transmission paradigm of generative intelligent transmission via AI Flow. On the application side, the intersection of the next-generation communications network and the computing power network thus takes on new meaning.

At present, token operations still face practical bottlenecks. In mobile and weak-network environments such as air-space-ground-sea domains, field operations and emergency response, tokens can be generated on cloud computing platforms but are difficult to transmit efficiently and stably to remote terminals. Massive multimodal sensing services create enormous transmission pressure, and bandwidth constraints directly affect the implementation of token services. Some government and enterprise AI scenarios face a situation where computing results can be produced but cannot be transmitted out, or cannot be transmitted well.

China Telecom's self-developed AI Flow builds a "device-edge-cloud" collaborative network architecture. Through technologies including generative intelligent transmission, family-based homologous models and distributed swarm intelligence, it fills the network transmission gap in delivering tokens from the cloud to terminals and returning on-site data tokens to the cloud. Previous token operations focused more on the computing side and platform side, addressing token generation, scheduling and billing, while the network transmission of tokens constitutes a necessary part of the full-chain token closed loop.

Under the traditional model for multimodal AI services, terminals need to send back complete original images, consuming enormous network bandwidth. Government and enterprise customers bear both computing token costs and dedicated line or satellite bandwidth costs, and this dual cost pressure constrains large-scale procurement of such services. Relying on AI Flow can substantially reduce the bandwidth cost of token services and improve the viability of business models. At the same time, token services can extend to narrowband satellite communications, emergency scenarios involving the loss of all three networks, marine operations and low-altitude inspections, fields that traditional networks struggle to support, forming integrated packaged products of "transmission plus computing power plus token services" and expanding new scenarios and new business growth for token operations.

AI Flow integrates artificial intelligence, communications and networking, enabling tokens to circulate not only within computing platforms but also in on-site operations across industries. For operators to conduct token operations, they need computing platforms as well as a next-generation network foundation suited to tokens. Together, these form the basis for the closed-loop operation of tokens across devices, edge and cloud.