South Korea invests $1 billion in "physical AI" to push AI into production control and create the "K Manufacturing Plant".

South Korea invests $1 billion in "physical AI" to push AI into production control and create the "K Manufacturing Plant".

The South Korean government has launched a large-scale physical AI manufacturing plan, investing 1.4131 trillion won (approximately US$1.01 billion) over the next five years to promote AI from factory defect detection to production control, equipment coordination, and factory operations. The plan also includes promoting mature technologies to manufacturing bases across the country to create "K-manufacturing factories" that can be exported overseas.

On September 16, according to the Seoul Economic Daily, South Korea's Ministry of Science and ICT held a briefing in Seoul to launch a physical AI research project for North Jeolla Province and South Gyeongsang Province, announcing a five-year research and development roadmap from 2026 to 2030. A key change in this project is the integration of manufacturing companies' long-accumulated process experience and "tacit knowledge" data, previously mainly held within factories, into the national R&D system. This aims to propel AI from simply identifying anomalies to understanding production processes and directly controlling equipment.

According to the plan, the projects in the two locations will first conduct technology verification, and the mature results will then be promoted to manufacturing bases nationwide. The relevant technologies involve multiple aspects such as neural processors, digital twins, robot collaboration, and industrial data infrastructure.

With an investment of 1.4 trillion won, the two regions are respectively tackling key challenges in precision manufacturing and factory collaboration.

The total R&D investment amounted to 1.4131 trillion won, of which 736.8 billion won was allocated to the North Jeolla Province project, focusing on the construction of factory platforms; and 676.3 billion won was allocated to the South Gyeongsang Province project, focusing on precision manufacturing. The two projects respectively addressed "how factories operate collaboratively" and "how AI understands and controls the production process."

The Gyeongsangnam-do project will focus on developing "large-scale motion models," which will aggregate motion data, physical laws, and spatial information from workers and robots, enabling AI to understand specific production processes and control actual equipment. Digital twins, data pipelines, and neural processor infrastructure will be used to connect data training, on-site applications, and results feedback.

The Jeollabuk-do project focuses on "factory orchestration," aiming to enable robots and equipment from different brands to operate collaboratively, allowing the entire factory to be scheduled as a system. AI will be responsible for factory layout and logistics robot path design, and will be verified in advance through 3D digital twins before the solution is deployed to actual equipment, ultimately moving towards a "lights-out factory" where AI participates in factory design, construction, and operation.

According to the project team, during the proof-of-concept exercise last year, AI completed the design of a robotic factory in about 3 hours, whereas in the past this process usually required 3 to 4 experts and nearly a month.

Thirteen companies have opened up their "tacit knowledge," enabling AI to move from detecting anomalies to controlling production.

Another key aspect of this project is that, for the first time, manufacturing companies have incorporated the "tacit knowledge" accumulated on the production floor over a long period into a national R&D project.

Cha Suk-won, a professor at Seoul National University who is in charge of the Gyeongsangnam-do project, said that 13 companies have already agreed to share tacit knowledge data used in their own factories. The project will collect and process this data, build physical models for different process stages, and then apply the models to other manufacturing companies with similar production processes.

Previously, AI in manufacturing was mostly used for tasks such as defect detection and anomaly identification. This project, however, attempts to further utilize the data and experience accumulated by the factory over a long period of time to enable AI to understand the specific operations and technological rules in the production process, and participate in equipment control accordingly.

The Jeollabuk-do project is led by Professor Chang Young-jae of the Korea Advanced Institute of Science and Technology (KAIST). He proposed viewing the factory as "a giant robot composed of various types of robots," achieving overall factory collaboration through unified scheduling of different equipment. He also revealed that Shinsung E&G, a supplier of semiconductor and display materials, components, and equipment, has decided to invest 20 billion won to participate in the development of the related physical AI industry ecosystem.

South Korea is betting on a holistic "manufacturing + AI" solution, expanding from demonstration factories to overseas markets.

The South Korean government's ultimate goal is not limited to demonstration projects in two regions, but to gradually promote proven technologies to manufacturing bases across the country and further develop "K-manufacturing plants" that can be exported.

According to the roadmap, from 2026 to 2030, technological research and development and verification will be carried out primarily in North Jeolla Province and South Gyeongsang Province, followed by the implementation of related AI capabilities in more manufacturing bases. The government hopes to combine the process data and production experience accumulated by South Korean manufacturing with technologies such as AI, robotics, and digital twins to form a comprehensive solution covering factory design, equipment collaboration, and production operations.

This means that South Korea's AI deployment in manufacturing is shifting from single-point applications to factory-level system transformation: AI is no longer just identifying anomalies on the production line, but is gradually participating in factory design, equipment control, and production scheduling, and is exploring the export of complete "AI factory" solutions to overseas markets.

Risk warning and disclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.