Group leaders' strategic keywords
"A shift to a software-centric manufacturing paradigm is needed"
Warning lights are flashing for 'K-Manufacturing.' Alarms are sounding across major industrial complexes including shipbuilding, steel, and petrochemicals. The shutdowns and closures are due to outdated machinery and equipment, the aging of skilled workers, and the intense competition from China.
What else could drive Seosan Mayor Lee Wan-sub to write a handwritten plea to President Lee Jae-myung saying "Save us"? In July, he requested that Daesan Petrochemical Industrial Complex, one of Korea’s three major petrochemical hubs, be designated as an industrial crisis preemptive response region. Prime Minister Kim Min-seok assessed the situation as "a crisis greater than the IMF crisis."
Physical AI is emerging rapidly as almost the only solution to the complex crisis and structural decline facing K-Manufacturing. Physical AI refers to AI that integrates with robots or automated machines to perform tasks or control the real-world physical environment.
"Physical AI is the answer"
In January, Jensen Huang, CEO of Nvidia, a superstar in the AI industry, appeared at CES, the world's largest IT and electronics fair. Throughout his two-hour keynote speech, he emphasized that an era has arrived where AI interacts with the real world beyond the digital realm. It's "Physical AI follows generative AI."
Manufacturing accounts for 24–28% of Korea's GDP, around double the OECD average of 14%, and higher than major manufacturing powerhouses such as Germany and Japan. Thus, analyses suggesting that the nation's fate hinges on physical AI have followed.
SK Group Chairman Chey Tae-won consistently advocates a transformation of manufacturing through physical AI. He remarked at the Korea Chamber of Commerce and Industry's summer forum that "a significant portion of Korea's manufacturing sector will be phased out within a decade due to China’s rise. The hope lies in AI specialized for manufacturing."
Recently, GS Group Chairman Huh Tae-soo addressed about 150 executives from subsidiaries, stating, "Physical AI necessary for process optimization and robotic integration innovation will be a pivotal turning point." He also remarked, "Technology is central to strategy, and insensitivity to technological change disqualifies an executive."
National Assembly members Chung Dong-young and Choi Hyung-doo are recognized as 'Physical AI evangelists' at the National Assembly. On Jul. 18, they held the '2025 Physical AI Global Alliance Colloquium' at the National Assembly Members' Office Building, discussing strategies for Korea's physical AI. Over 420 manufacturing company and research institute officials attended. An industry-academia-research consortium is expected to launch in Gyeongnam and Jeonbuk regions as early as August.
The two lawmakers were instrumental in including the 'Physical AI Core Technology PoC (Proof of Concept) Budget (42.6 billion won)' in the 2025 supplementary budget plan. Rep. Chung said, "There are no leading nations in physical AI yet; manufacturing powerhouse Korea must take the lead." Rep. Choi added, "Gyeongnam is the mecca of K-Manufacturing. AI transformation (AX) in factories is vital not only for Gyeongnam but also for the future of Korea's young people."
K-Manufacturing revival: The three major strategies
① Dark factory
The strategies for physical AI corporations and policy to revive Korean manufacturing can be summarized into three major strategies. First is the 'dark factory' pioneering strategy. A dark factory refers to a futuristic plant where all production processes are autonomously conducted by robots and AI, eliminating the need for lighting.
An example of this is the solution by Daim Research, created by Professor Jang Young-jae's team from KAIST's Department of Industrial and Systems Engineering. The company develops smart operation solutions controlling hundreds or thousands of logistics robots. Professor Jang stated that "physical AI embodies 'moving intelligence'" and emphasized the need for "a shift to software-centered manufacturing."
The explanation is that, unlike the current hardware-centric approach requiring replacement of entire products, future factories must evolve through virtual simulations and frequent updates.
② Materials, parts, and equipment
Next is the materials, parts, and equipment (소·부·장) innovation strategy. It involves harnessing physics-based AI models to develop not only new materials but also parts and equipment. This is akin to how Google's DeepMind developed the protein structure prediction AI 'AlphaFold,' significantly reducing drug development time.
Aiming to achieve this ambitious goal is the Korea University of Gyeongnam's Global Center for Co-Manufacturing AI. Professor Yoo Nam-hyun of Korea University of Gyeongnam (Deputy Director of the Global Co-Manufacturing AI Center) noted, "Korea University of Gyeongnam possesses the most extensive manufacturing equipment modeling data in the country," adding that "by collecting vast manufacturing data and refining physics-based models (PINNs), the development time for new equipment can be drastically reduced."
He further mentioned, "However, there is an urgent need to develop device architecture to extract physical data from factories and solutions for refining and synthesizing the data."
③ Humanoid
Lastly, there is the humanoid (human-like robot) development strategy tailored to the environments, tools, and workspaces familiar to people. This approach is akin to Tesla’s development of the humanoid 'Optimus' for revolutionizing manufacturing processes. Unlike conventional industrial robots that operate in specially designed spaces, humanoids can be deployed in nearly all human-used spaces without modification.
Recently, ROBOTIS, with LG Electronics as its second-largest shareholder, gained attention for supplying the humanoid 'AI Worker,' capable of tasks using both arms, to the U.S. OpenAI. ROBOTIS develops actuators, key components equivalent to a robot's joints.
ROBOTIS CEO Kim Byung-soo said, "With worldwide interest in physical AI, we've received numerous collaboration proposals domestically and internationally," explaining its renewed focus on humanoid development after a brief hiatus.
In April, the 'K-Humanoid Alliance' was launched under the initiative of the Ministry of Trade, Industry and Energy. The alliance includes robot manufacturing conglomerates, promising small and medium-sized enterprises related to components, software corporations, and universities. K-Humanoid Alliance Chairperson Jang Byoung-tak (Seoul National University professor) stated, "Initially, humanoids will be demonstrated based on specific industry demands, such as logistics and construction, before generalization."
Obstacles to overcome
To lead in physical AI, there are numerous challenges to address. Currently, the manufacturing industry lacks a standardized ontology that AI can comprehend. Standardizing the definition and interrelationship of manufacturing elements like processes, equipment, and materials is essential for AI to perform analysis and reasoning.
It is also crucial to expand the digital infrastructure enabling data collection and processing across manufacturing sites, including local small and medium-sized enterprises. According to the first smart manufacturing innovation status survey announced in April by the Ministry of SMEs and Startups and the Smart Manufacturing Innovation Promotion Group, only 0.1% of small and medium manufacturing corporations have adopted AI. Experts view physical AI as an issue of industrial ecosystem transformation, not solely a technical challenge.
Escaping fierce competition with China won't be easy. Last year, during its government work report at the National People's Congress, China announced the AI+ plan, aiming to integrate AI with industries like manufacturing, finance, healthcare, automobiles, and agriculture to fuel new economic growth. There are assessments that China has outpaced Korea in industrial policy terms as well.
KAIST President Lee Kwang-hyung stated, "If Korea succeeds in creating a physical AI example, a renaissance for K-Manufacturing will commence," emphasizing that "universities should cultivate physical AI experts to solve problems in manufacturing sites."