As physical artificial intelligence (AI), including humanoid robots, rapidly commercializes in manufacturing, experts are calling for the establishment of a domestic robot platform. There are concerns that if South Korea provides its manufacturing infrastructure to global companies, it may become a testing ground for foreign firms.
The Korea Institute for Industrial Economics and Trade released a report on September 16 outlining strategies for developing a physical AI ecosystem for manufacturing. The report predicts that physical AI, including humanoid robots, is more likely to be commercialized in manufacturing settings than in service sectors.
The market is growing rapidly. The global market for physical AI in industrial automation is projected to increase from $95 million in 2025 to $1.8887 billion by 2032, reflecting an average annual growth rate of about 53%. This is due to the relative ease of learning and commercialization in manufacturing environments.
Currently, the transition to manufacturing AI (AX) involves integrating AI into operational technology (OT) solutions, such as production management systems, or utilizing digital twins and predictive maintenance within fixed factory equipment. However, this approach still requires human workers for complex, unstructured tasks. Additionally, applying AI to older equipment commonly used by small and medium-sized manufacturers poses significant challenges.
In response, the institute proposed 'physical AI-based manufacturing AX' as an alternative. This approach involves AI robots, including humanoids, mimicking and learning human actions to perform tasks without significantly altering existing work environments, allowing robots to directly operate older equipment. This could help overcome the limitations of fixed-equipment-centered manufacturing AX.
South Korea has a competitive advantage in terms of behavioral data. The institute identified six key elements for a physical AI ecosystem: form factor development and design, manufacturing-specific AI behavioral learning, actuators, robotic hands, AI foundation models, and on-device AI semiconductors.
However, a challenge remains: the behavioral data that AI robots learn is dependent on the specific form factor of the robots. If domestic manufacturing sites are provided to foreign companies without securing local robots, the behavioral data generated could enhance the competitiveness of global firms' robots. Lee Jun-young, a senior researcher at the institute, expressed concern, stating, "If we only offer our manufacturing sites without our robots, our excellent manufacturing infrastructure could become a testing ground for global physical AI companies."
To address this, the institute recommends that South Korea secure its own physical AI robots and expand the domestic ecosystem around them. In the short term, it suggests collaborating with global companies to establish a market entry foundation focused on manufacturing-specific behavioral learning, while in the long term, it emphasizes the need to cultivate areas where domestic form factors and key components can be internalized.
The government’s ongoing 'AI Robot Global Power Leap' policy should also be supplemented with a focus on developing domestic robot form factors. The institute proposed a roadmap for expanding the introduction of robots in manufacturing settings based on technological maturity, starting with lightweight, general-purpose, and high-intensity robots. It also suggested creating AI robot learning infrastructure at the industrial complex level to support small and medium-sized manufacturers that lack the capacity for self-introduction of AX.
Lee emphasized, "Unlike the United States, which has a strong private-sector physical AI ecosystem, China is rapidly catching up with a state-led strategy. South Korea must also take a proactive role in creating a virtuous cycle of training our robots using our excellent manufacturing infrastructure."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.