Driven by the "New Whole-of-Nation System" to cultivate "New Quality Productive Forces", Beijing positions AI as the core industrial engine. The 2025 official roster validates 151 typical cases and 40 core scenarios, yet these remain heavily concentrated within highly automated sectors like electronics, automotive, and steel. Current adoption is dominated by SOEs (e.g., Baosteel, State Grid) and industry champions (e.g., Huawei, CATL), favouring "single-point breakthroughs" like visual inspection over complex process integration. To scale this, the "AI + Manufacturing" initiative mandates the standardised embedding of AI models directly into core industrial software, transitioning AI from a peripheral tool to the central operating system of R&D and production. Simultaneously, the government is executing a "benchmark-and-replicate" strategy: validating these high-value templates to incentivise rapid, low-cost adoption across the broader supply chain.
MIIT and seven other agencies jointly issued "Implementation Opinions on 'Artificial Intelligence + Manufacturing' Special Action", specifying:
- by 2027, China’s AI core technologies will achieve safe and reliable supply
- industry scale and empowerment capabilities to rank among world’s top
- promote 3-5 general large models in deep manufacturing applications
- launch 1,000 high-level industrial intelligent agents
- create 100 high-quality industrial datasets and promote 500 typical application scenarios
- cultivate 2-3 globally influential ecosystem-leading enterprises and 1,000 benchmark enterprises
- strengthen AI computing power and model foundation
- support breakthroughs in high-end training chips, edge inference chips, and AI servers
- develop "cloud-edge-endpoint" model systems and high-performance algorithms adaptive to manufacturing reliability
- implement "model-data resonance" action: establish chief data officer systems and convert basic data into high-quality industry datasets
- expand high-value application scenarios across manufacturing value chain
- r&d design: promote intelligent auxiliary design, software code generation, and AI-driven drug development
- pilot verification: accelerate virtual simulation and multi-modal fusion to lower testing costs
- production: deepen AI in core process control and scheduling; promote machine vision and unmanned intelligent inspection
- marketing & service: deploy digital humans and personalised recommendation; improve predictive maintenance
- accelerate intelligent equipment and product iteration
- industrial equipment: develop new generation AI numerical control systems for machine tools and industrial robots
- smart terminals: cultivate AI mobile phones, PCs, AR/VR devices, and brain-computer interfaces
- humanoid robots: build pilot bases and training grounds; apply in typical manufacturing scenarios
- industrial agents: develop "agent store" and collaborative agent protocols; focus on task planning and group collaboration
- empower key industries (specific domains)
- raw materials:
- steel: develop visual/decision large models for real-time sensing and global scheduling optimisation
- petrochemical: build industry large models for safety warning and equipment predictive maintenance
- new materials: use AI for "component-structure-performance" inverse design
- equipment manufacturing:
- auto: generate styling/layout schemes automatically; optimise wind resistance via simulation
- aerospace: use AI for aerodynamic simulation and intelligent assembly of complex components
- ships: optimise navigation energy efficiency and "cutting-welding-coating" processes
- consumer goods:
- pharma: accelerate target identification and virtual screening to lower R&D costs
- textiles: implement virtual try-on systems and defect self-repair in production
- electronics:
- components: enable virtual simulation debugging and millisecond-level process tuning
- raw materials:
- optimise industrial ecosystem and security
- standards: formulate standards for safety, ethics, and "AI + manufacturing" applications
- open source: build high-level AI open source communities and promote adapted open source licences
- security: construct industrial safety large models; strengthen defence against deepfakes and adversarial attacks
- talent: cultivate composite talent understanding both AI and manufacturing via "school-enterprise" platforms