China’s Ministry of Industry and Information Technology (MIIT) has released a comprehensive three-year policy directive to deeply integrate artificial intelligence into the nation’s telecommunications infrastructure. The “Implementation Opinions on Innovative Development of ‘AI+ICT’ (2026-2028),” reported by Futunn, outlines aggressive targets for network latency, bandwidth, and deployment of advanced optical technologies, signaling a strategic push to build the foundational infrastructure for the next generation of AI applications.
The directive underscores Beijing’s recognition that the widespread adoption of embodied intelligence, autonomous systems, and complex AI agents will require telecommunications networks capable of delivering unprecedented speed and reliability.
Aggressive Latency and Coverage Targets
At the core of the MIIT’s three-year roadmap are stringent performance metrics designed to eliminate network bottlenecks that currently hinder real-time AI processing. The most prominent target mandates that by 2028, at least 75 percent of metropolitan-area computing power must be accessible within a 1-millisecond latency circle.
This ultra-low latency requirement is critical for applications that demand instantaneous feedback, such as autonomous driving, remote robotic surgery, and synchronized industrial automation. By ensuring that computing resources are physically and logically closer to the end-user, the MIIT aims to facilitate the seamless operation of these latency-sensitive technologies.
Furthermore, the directive explicitly requires that Wireless Local Area Network (WLAN) access latency for both residential and enterprise environments be reduced to 5 milliseconds or less. This target is specifically tailored to support the proliferation of AI agents and smart home devices that require constant, high-bandwidth, and low-latency connections to cloud-based or edge-based AI models.
The roadmap also sets a high bar for network efficiency, mandating a network resource utilization rate of at least 90 percent for distributed inference tasks. This efficiency target is essential for maximizing return on investment in massive computing clusters and for ensuring that AI workloads can be dynamically routed across the network without significant performance degradation.
Elevating Optical Technologies to “Strategic Foundation”
To achieve these ambitious performance metrics, the MIIT directive elevates several advanced hardware technologies to the status of “strategic foundation.” Specifically, the roadmap prioritizes the development and deployment of photonic chips, Optical Circuit Switches (OCS), and Co-packaged Optics (CPO).
The emphasis on optical technologies reflects the physical limitations of traditional electronic switching and transmission. As data volumes explode, optical solutions offer significantly higher bandwidth density and lower power consumption.
•Optical Circuit Switches (OCS): These devices establish direct optical paths between input and output ports, eliminating the need for optical-to-electrical-to-optical conversion. This direct routing drastically reduces both latency and power consumption within data centers and across core networks.
•Co-packaged Optics (CPO): This technology involves packaging the switch Application-Specific Integrated Circuit (ASIC) and the optical modules on the same substrate. By minimizing the physical distance between the electronic and optical components, CPO significantly increases bandwidth density while reducing the energy required to drive the signals.
The strategic focus on these technologies aligns with broader national efforts to secure domestic supply chains for critical infrastructure components. The MIIT’s push for optical innovation provides a strong policy tailwind for domestic companies working in this space, such as Huawei’s recent investments in domestic optical chip autonomy.
Upgrading the Backbone and Metro Networks
The implementation opinions also detail specific upgrades required across different tiers of the national telecommunications network. For the backbone network, the MIIT mandates the accelerated deployment of 400Gbps, 800Gbps, and higher-capacity transmission systems. This massive increase in core bandwidth is necessary to handle the aggregated data flows generated by millions of AI endpoints.
At the metropolitan level, the directive promotes the adoption of 400Gbps+ optical transmission and all-optical cross-connect technologies. This ensures that the high-capacity backbone connects seamlessly to local distribution networks without creating localized bottlenecks.
To manage these complex, high-speed networks efficiently, the MIIT is pushing for the adoption of advanced routing protocols, including Segment Routing over IPv6 (SRv6) and Optical Service Unit (OSU) technologies. These protocols enable end-to-end rapid provisioning and dynamic traffic engineering, allowing the network to automatically adjust to the fluctuating demands of distributed AI workloads.
The roadmap also explicitly includes provisions for 5G-Advanced (5G-A) and 6G convergence innovation, ensuring that China’s wireless infrastructure evolves in tandem with its optical networks to support the massive uplink bandwidth required by embodied intelligence and mobile AI agents.
By setting these aggressive targets and prioritizing specific technological pathways, the MIIT’s three-year roadmap aims to ensure that China’s telecommunications infrastructure remains a competitive advantage rather than a limiting factor in the global AI race. The successful execution of this plan will be critical to cultivating 30+ high-value AI application scenarios by 2028 and achieving significant breakthroughs in AI+ICT integration by 2030.