en.Wedoany.com Reported - On June 10, China's Ministry of Industry and Information Technology issued the "Implementation Opinions on the Innovative Development of 'AI + Information and Communication' (2026–2028)", which proposes enhancing network intelligent computing service capabilities, with a focus on improving AI dedicated line services and the network's ability to support AI businesses.
The "Implementation Opinions" places "how networks serve AI" at a more concrete construction level. For intelligent agents, embodied intelligence, distributed inference, and industry model applications, traditional network connections are no longer just about solving the "can it access the internet" problem, but must meet new demands such as high uplink bandwidth, low latency, rapid scheduling, and stable computing access. The document proposes enriching various dedicated line types, including Optical Transport Network (OTN) dedicated lines, Passive Optical Network (PON) dedicated lines, and IP dedicated lines, to achieve rapid service provisioning and flexible scheduling. It also aims to enhance uplink bandwidth configuration for fiber access networks, promote the deployment of 5G-A networks supporting high uplink capabilities, optimize network experience, and reduce end-to-end latency. This implies that AI businesses will drive operators' networks to upgrade from general connectivity services to dedicated capabilities oriented towards computing resource invocation, model inference, and intelligent terminal collaboration.
AI dedicated lines are a key element in this round of action. The document specifies that related services will optimize network resource scheduling capabilities, provide models such as on-demand flexible billing and time-sharing, and leverage technologies like Segment Routing over IPv6 (SRv6), Optical Service Unit (OSU), and fine-grain Optical Transport Network (fgOTN) to achieve rapid end-to-end dedicated line provisioning and bandwidth adjustment.
The adaptation of network-side capabilities for AI businesses is further quantified. The document proposes conducting distributed inference technology trials across wide-area IP networks, with network resource utilization not less than 90%. It aims to enhance uplink service access capabilities for 5G-A and PON, and optimize wireless LAN access latency for homes and enterprises to no more than 5ms. For intelligent agents and embodied intelligence applications, terminals need to frequently upload images, voice, sensor data, and task instructions, with results returned after inference by cloud or edge computing. Any link congestion or latency fluctuation can affect the interactive experience. The inclusion of network resource utilization, uplink bandwidth, and LAN access latency in the action content indicates that AI applications are driving changes in communication network evaluation metrics. Operators also need to reorganize network capabilities around inference tasks, edge nodes, and computing channels.
This policy will drive adjustments across multiple segments of the information and communication industry chain. OTN, PON, IP dedicated lines, SRv6, OSU, fgOTN, 5G-A, WLAN optimization, edge nodes, computing scheduling platforms, and network operation and maintenance systems may all generate new construction demands centered on AI business carrying. The document also proposes promoting model application platforms, establishing an integrated service ecosystem of "computing power + data + models + AI applications", and building a full-process development environment from model training and scenario validation to commercial deployment. For operators, cloud service providers, equipment manufacturers, and industry customers, AI businesses are no longer just about deploying models or purchasing computing power, but require coordinated delivery of communication networks, intelligent computing centers, data resources, and application platforms.
Subsequent implementation will focus on three types of scenarios: enterprises and parks rapidly accessing intelligent computing resources via dedicated lines; intelligent agents and embodied intelligence achieving more stable uplink capabilities through 5G-A and fiber networks; and industry customers relying on model application platforms for training, validation, and deployment. By 2028, if related targets are advanced as planned, China's information and communication networks will be more deeply involved in the construction of AI industry infrastructure. Network services will further shift from bandwidth sales to computing connectivity, inference assurance, and support for industry AI applications. For the information and communication industry, the signal from this document is clear: network competition in the AI era will no longer depend solely on coverage scale and download speeds, but also on the ability to efficiently connect users, terminals, data, and computing power.
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
