China has set a 2030 target of 9,800 EFLOPS of intelligent-computing capacity and 3.8 trillion yuan in cumulative information-infrastructure investment, giving its AI buildout a more explicit policy horizon. The figures appear in the Ministry of Industry and Information Technology’s 15th Five-Year Plan for the information and communication sector, released September 7 and reported by China Daily Asia. The document is a target-setting blueprint, not evidence that the capacity or investment has already been achieved.
The scale of the targets shows why artificial intelligence has become inseparable from telecom networks, data centers, broadband, and electric power. Models require chips and servers, but they also need networks to connect users and applications, storage to handle data, and enough infrastructure to support deployment outside a small group of labs. China’s plan places AI inside a wider communications strategy rather than treating computing capacity as a separate industrial niche.
The 9,800-EFLOPS goal is particularly notable because China Daily Asia compares it with an existing reported level of 2,185 EFLOPS. EFLOPS measure exa floating-point operations per second, a way of expressing computing capacity. The comparison does not mean every unit of capacity will be used for AI training or that all systems are directly comparable. It does indicate the magnitude of growth the government wants by the end of the decade.
A 3.8 Trillion Yuan Foundation for Intelligent Computing
The plan calls for 3.8 trillion yuan in cumulative information-infrastructure investment by 2030. It also sets a target for the information and communication industry to generate 4.1 trillion yuan in revenue by that year. These targets make clear that policymakers view infrastructure as both a public-service foundation and an industrial growth engine.
The plan sorts 13 indicators into five lenses: industry development, innovation, green progress, infrastructure, and adoption. China Daily Asia reports six priority tasks, including building information infrastructure with moderate overcapacity, expanding the information and communication industrial system, strengthening data and network security, widening applications, and improving global development capacity.
The phrase “moderate overcapacity” is significant. It suggests that policymakers do not want infrastructure built only for currently measured demand. AI services can need capacity before their usage is fully visible, and a network built only after demand arrives may become a bottleneck. At the same time, the plan does not define how much excess capacity is appropriate in each location or sector. Implementation will determine whether investment is coordinated with actual workloads.
China’s computing ambitions have been visible in prior infrastructure projects. EastFrontier reported that China’s AI computing capacity reached 962 EFLOPS in July, using a different official data point from earlier in the year. The new plan’s 9,800-EFLOPS target should not be treated as a simple tenfold forecast from that figure, because reporting methodologies and dates can differ. It does, however, show the policy direction toward much larger capacity.
Networks, Broadband, and Data Security Support AI Deployment
The plan is not only about AI accelerators. By 2030, China aims for 50 5G or 5G-A base stations per 10,000 people, 95% 5G or 5G-A user penetration, and 320 million gigabit-broadband subscribers. Those telecom measures can support AI applications that require low latency, large data transfers, or reliable connections across factories, vehicles, offices, and homes.
A large language model running in a distant data center still depends on the network path between a user and the server. The same is true for industrial inspection, remote-device management, and agent systems that connect to enterprise software. The plan therefore treats network coverage and computing capacity as related components of the same digital foundation.
Data and network security are also listed among the priorities. That matters for AI because more connected systems create more possible points of failure. A capacity target alone does not determine whether users trust an AI service with sensitive data or whether an enterprise can deploy it across a regulated operation. The plan’s security emphasis acknowledges that infrastructure growth creates governance demands as well as commercial opportunity.
EastFrontier’s coverage of China’s national supercomputing network adding 1,700 AI models showed an earlier effort to connect computing resources with model access. The new information-and-communication plan operates at a broader level. It provides the network, capacity, and governance framing within which such platforms could expand.
Targets Create a Scorecard for China’s AI Buildout
The most useful way to read the 2030 plan is as a scorecard. It supplies specific figures against which future progress can be measured: investment, computing capacity, revenue, base-station density, user penetration, and broadband subscriptions. Those numbers make it easier to identify where the buildout is proceeding and where delivery is slower than policy ambition.
The targets are not guarantees. Investment can be delayed, equipment supply can tighten, and demand can shift. Intelligent-computing capacity also does not automatically translate into domestically developed chips, profitable AI applications, or globally competitive models. The plan gives a destination, but companies and local governments will still need to decide what systems to build and how to use them.
For AI developers, the infrastructure agenda could lower some barriers over time by expanding access to networks and computing capacity. For telecom operators and equipment suppliers, it creates a policy framework for continued capital expenditure. For investors, it offers a way to separate short-term enthusiasm from long-term execution: the relevant question is whether reported capacity, revenue, and connectivity figures begin to move toward the plan’s benchmarks.
China’s policy message is clear. AI is no longer being planned only through model programs and pilot applications. It is being embedded in a 2030 vision for communications networks, data systems, broadband coverage, and intelligent computing. The hard part will be translating targets into capacity that is affordable, secure, geographically useful, and connected to applications that people and businesses actually use.