China’s 15th Five-Year Plan places artificial intelligence across industrial policy, public services, consumer technology and national infrastructure through 2030.
The plan groups AI with quantum computing, biotechnology and energy as strategic science priorities. It calls for domestic progress in high-performance chips, supporting software, model architectures and core algorithms, while linking deployment to communications, computing hubs and data organisation.
The scope shows that China is treating AI less as one technology sector than as a layer across the economy.
Compute, models and data form the foundation
The digital infrastructure strategy rests on three components: computing power, AI models and the movement and organisation of data.
National “intelligent computing clusters” would expand capacity, while leasing mechanisms are intended to give smaller companies and a wider population access to advanced resources. New government procurement methods could also provide demand for computing services.
Communications investments in satellite systems, 5G Advanced and 6G are expected to support data-intensive workloads. The plan also calls for continued research and manufacturing around training and inference, including multimodal, agent-based and embodied AI.
Access is an important part of the design. If compute remains concentrated among the largest companies, smaller firms cannot experiment or build competitive products. Shared clusters can lower that barrier, although allocation and pricing will determine who benefits in practice.
Deployment targets the physical and service economies
Manufacturing, energy, agriculture and services are all named. Industrial uses include design, production and operations. Energy systems could apply AI to management, while agriculture gains tools for production.
Finance, logistics and software are highlighted in services. For consumers, the plan promotes more AI-enabled phones, computers and robots, alongside applications in education, healthcare, elder care and social provision.
Adaptive learning, diagnostic assistance and welfare administration illustrate the intended public impact. Government bodies are also expected to expand digital services through integrated data systems and use models in administration and public-safety risk assessment.
These ambitions connect frontier research with ordinary devices and institutions. Success will depend on whether local implementations deliver reliable services rather than simply meeting deployment counts.
Regulation grows alongside adoption
The plan devotes meaningful attention to governance. It calls for legal and regulatory frameworks covering algorithm registration, security and transparency, while naming data misuse and deepfakes as risks.
International cooperation appears more cautiously, including possible participation in standards for data flows and infrastructure. The balance suggests a desire for domestic control combined with influence over the rules through which systems interact globally.
The source notes that China’s recent approach has emphasised smaller, efficient and openly available models more than the large proprietary systems associated with a handful of Western companies. The Five-Year Plan leaves room to see whether that path continues as deployment broadens.
Many implementation details remain unspecified. A national plan establishes priorities, but budgets, provincial programmes, procurement and measurable targets will reveal how those priorities become operational.
The direction is nevertheless clear. China wants AI capability in chips, networks, compute, models and data, then wants that stack embedded across production and public life.
Observers will watch whether shared infrastructure genuinely expands access, whether efficient models remain central and how regulation balances rapid adoption with control.
By 2030, the plan’s success will not be judged only by model benchmarks. It will be visible in factories, farms, hospitals, classrooms, government services and the devices used by hundreds of millions of people.
That breadth creates significant coordination risk as well as opportunity. Compute access, data standards, local implementation and regulatory enforcement must develop together. A strong chip programme cannot compensate for inaccessible data, while widespread deployment without transparent rules could amplify the misuse and deepfake concerns the plan itself identifies.
