China’s Ministry of Industry and Information Technology (MIIT) has unveiled an ambitious roadmap aimed at embedding artificial intelligence into the country's vast telecommunications and information sectors. The newly released 'Implementation Opinions on AI + Information and Communication (2026–2028)' seeks to foster a self-reinforcing ecosystem where AI optimizes networks and networks, in turn, provide the massive compute and data throughput required by modern generative models.
By 2028, the ministry expects China’s telecommunications infrastructure to reach an 'advanced international level' of intelligent operation. The plan outlines the creation of over 30 high-value industrial scenarios and a series of specialized AI agents. Crucially, the policy mandates that metro computing power latency circles—the time it takes for data to travel from user to server—must achieve a 1-millisecond response time for at least 75% of targeted areas, effectively bringing high-speed processing to the network's edge.
The strategic shift focuses on three pillars: applying AI to network maintenance, supporting the broader AI industry via massive computing clusters, and merging AI with physical connectivity to create new services. China has already deployed over 5 million 5G base stations and reached a smart computing capacity of 1882 EFLOPS. However, the MIIT notes that the next phase requires moving from mere 'connectivity' to 'intelligent connectivity,' where networks can self-diagnose and optimize without human intervention.
Industrial experts believe this policy will spark a significant cycle of hardware replacement. Pan Helin, a prominent economist and member of the MIIT’s expert committee, suggests that telecommunications carriers will see revenue growth through integrated 'Network-Model-Compute' services. Meanwhile, equipment manufacturers are poised to receive a wave of orders for 5G-A (5.5G) and 6G-ready hardware capable of handling edge-based inference for robotics and 'dark factories.'
Despite the optimistic goals, structural hurdles remain. Integrating heterogeneous computing power—bridging different types of chips and architectures—remains technically complex, and the tension between data privacy and energy efficiency is mounting. Furthermore, the plan addresses a critical geographical mismatch in China’s 'East-to-West Computing' project; by reducing network latency, AI developers in the dense eastern coastal cities can finally utilize the cheaper, renewable-energy-powered data centers located in the country’s western provinces.