MIIT Drafts New-Generation Smart Manufacturing Policy, Will Steadily Carry Out R&D and Application Validation of Industrial Agents

On October 20, the State Council Information Office held a press conference. It was noted at the meeting that the CPC Central Committee and the State Council attach great importance to the development of next-generation intelligent manufacturing. At present, the Ministry of Industry and Information Technology (MIIT) is studying and drafting relevant policies and measures for next-generation intelligent manufacturing, to accelerate the deep integration of artificial intelligence and manufacturing.

The meeting emphasized that the key to developing next-generation intelligent manufacturing lies in promoting the deep integration of AI technology with the entire manufacturing chain. The MIIT will give full play to China's unique advantages of vibrant AI innovation and rich manufacturing scenarios, and focus on doing well in four areas: "tiered cultivation of smart factories," "innovative development of intelligent equipment," "accelerating breakthroughs in and the promotion and application of intelligent technology," and "deepening international cooperation in intelligent manufacturing."

Among these, in the field of intelligent technology breakthroughs, the MIIT will focus on scenarios such as intelligent production scheduling, parameter tuning, defect detection, and equipment operation and maintenance, implement the Model-Data Resonance initiative, and prudently carry out the R&D and application validation of industrial agents, promoting each one as it matures.

According to the definition of the China Academy of Industrial Internet, an industrial agent refers to a software entity or an integrated software-hardware intelligent system that, oriented toward clear industrial objectives and a specific industrial environment, integrates industrial data, knowledge, models, and mechanisms, possesses capabilities of environmental perception, task understanding, planning and decision-making, tool invocation, collaborative interaction, controlled execution, and feedback optimization, and can complete digital or physical industrial tasks within set permissions and security boundaries.

The institution believes that developing industrial agents is an inevitable choice for conforming to the evolutionary trend of next-generation AI technology and promoting the deep integration of AI and manufacturing.

On the policy front, the State Council's Guiding Opinions on Deeply Implementing the "AI+" Initiative lists "promoting the intelligent development of all factors of industry" as a key task, and proposes accelerating the implementation and application of AI across all links of design, pilot testing, production, service, and operations, with the application penetration rate of next-generation intelligent terminals, agents, and other applications exceeding 70% by 2027. The Implementation Opinions on the "AI+Manufacturing" Special Initiative, issued by the MIIT and eight other departments, proposes launching 1,000 high-level industrial agents by 2027.

Haitong International Securities summarized that since 2025, "AI+Manufacturing" policies have been implemented intensively, and from the State Council's overall outline to the "Model-Data Resonance" initiative and the tiered cultivation of smart factories, the policy focus has shifted from digital infrastructure building to reshaping production processes with large models and agents.

However, the institution stated that the penetration of industrial AI is currently uneven—the adoption rate among large enterprises is 15.5%, while that among small enterprises is only 5.4%; the equipment manufacturing industry leads with 13.5% of AI investment, while mining and chemicals remain below 7%. Scale thresholds and industry foundations have affected the pace of implementation, and growth space in the next 3–5 years will mainly come from industrial software, intelligent manufacturing solutions, robots, and AI infrastructure.

It emphasized that general large models provide "intelligence," industrial software provides "organization," and agents can only be implemented and executed when mounted in existing workflow containers. Although the market once worried that general large models would "bypass" industrial software and directly serve customers, in fact, process parameters, equipment operation data, and knowledge accumulated from historical projects reside in software systems and are the fuel for training vertical models and agents—these data are not on the internet, and model vendors cannot obtain them on their own.

Guohai Securities believes that, as policies continue to advance, leading domestic enterprises will accelerate their layout of industrial large models/industrial agents, which is expected to rapidly promote AI-enabled innovation and upgrading of industrial software and advance the development of new industrialization. Driven by both domestic substitution and intelligentization, coupled with continuous increases in domestic policy support, China's industrial software industry is expected to usher in a period of rapid development.

Soochow Securities judges that the future industrial world model will not develop in a way where "pure neural networks replace traditional simulation," but will form an integrated system in which "mechanism models provide boundaries, neural networks learn residuals, real-time data corrects states, and agents execute closed-loop control." It is expected that the first round of commercialization of physical AI will occur along three main lines:

· Autonomous operation of process industries: recommend focusing on Supcon Technology;

· Physical simulation and engineering world models: recommend focusing on Pinming Technology, Orbbec, MEMSensing, and Suochen Technology;

· Industrial agents and digital threads: recommend focusing on Nengke Technology.