ho can build the most powerful model? This is the key question of the AI race over the past three years. But it may not be the right one any longer.
On June 10, the Ministry of Industry and Information Technology released a roadmap for upgrading China’s telecommunications industry from 2016 to 2028. It signals a shift in China’s AI strategy from foundation models to integrated AI ecosystems.
The two have generally been treated as separate but complimentary domains, with communications providing the infrastructure on which AI applications run. The new policy points to a different vision: AI and telecommunications are developed as a single intelligent infrastructure, with networks, computing, data, security and applications evolving together. It is based on the recognition that as AI moves into real-world applications, the advantage will increasingly lie in integrating models with communications networks, distributed computing, data and industry-specific applications.
As future 6G and computing networks connect more devices and support increasingly complex services, traditional network management built around manual configuration and reactive maintenance is reaching its limits. Deploying AI across networks can automate tasks ranging from resource allocation and fault prediction to traffic optimization and cybersecurity, enabling networks to monitor, optimize and repair themselves with far less human intervention. The guideline calls it “high-level autonomous networks.”
The relationship also runs in the opposite direction. Just as AI can make communications networks smarter, advanced communications infrastructure will determine how far AI can be deployed across the real economy. Moving beyond chatbots and content generation into applications such as factory automation, autonomous driving and remote healthcare requires not only more capable AI models, but also ultra-fast, reliable networks and computing resources that are available where they are needed. That is why the plan sets a target of ensuring at least 75 percent of metropolitan areas have access to computing resources that can be reached with an end-to-end latency of one millisecond or less by 2028.
The policy shift also points to a new approach to computing infrastructure. Computing power will increasingly be distributed across the cloud, edge servers and end devices. Large data centers will continue to train AI models and handle the most demanding workloads, while distributed edge computing supports time-sensitive applications closer to users. Only by distributing computing resources throughout the network can AI deliver the low latency, reliability and responsiveness required for various applications such as autonomous driving, industrial automation and remote healthcare.
Generally speaking, rather than focusing primarily on breakthroughs in foundation models, the emphasis is moving toward integrating the infrastructure and real-world applications to translate those advances into economic value.
For the communications industry, the first phase of digitization was about connecting people and devices. In the AI era, its task will be about intelligent integration.
The implementation plan suggests that China increasingly sees the next stage of AI competition not as a race to build ever-larger models, but as a race to integrate AI into the infrastructure that will turn advances in AI into scalable, real-world applications and underpins the future economic growth engine.
Old Version
