The South Korean government announced a large industrial investment program in June centered on semiconductors, physical AI and AI data centers. The plan is intended to expand domestic production capacity and connect the country’s semiconductor industry with growing demand for computing infrastructure and industrial automation.
Samsung Electronics and SK Hynix are expected to invest a combined $541.03 billion USD (KRW 800 trillion) in new semiconductor capacity in southwestern South Korea. Samsung has selected Gwangju as a candidate location for a new production base, while SK Hynix is also considering an additional site in the region. A separate advanced semiconductor packaging cluster, involving approximately $54.78 billion USD (KRW 81 trillion) in investment, is planned for the Chungcheong region.
South Korean companies, including SK Group, GS Group and Naver, are also planning major investments in AI data centers. Initial projects are expected to total around $371.96 billion USD (KRW 550 trillion) and provide 8.4 GW of capacity. Construction is scheduled to begin by the first half of 2028, and cumulative investment could exceed $676.28 billion USD (KRW 1,000 trillion) by around 2035.
The physical AI program aims to make South Korea a leading producer of industrial and humanoid robots by 2030. The government plans to support the development of robots for ten major industries, with initial commercial applications targeted for 2028. The program also includes industrial data collection facilities, a domestic physical AI foundation model and training programs for engineers and researchers.
A robotics and component manufacturing cluster is planned for the Saemangeum area on the west coast. The government intends to combine robot production, component supply, AI development and industrial testing within the same regional base.
Source: Reuters
PSR Analysis: The plan is notable because South Korea is not treating AI as a software or semiconductor issue alone. It combines chip production, data centers, power and cooling infrastructure, robotics, component manufacturing and workforce development. The objective is to turn rising demand for AI computing into a broader cycle of domestic manufacturing investment.
The scale of the program will create substantial demand for industrial equipment before physical AI has a major direct effect on machinery markets. Semiconductor plants and data centers require construction equipment, standby and distributed power systems, cooling equipment, material-handling systems, uninterruptible power supplies and electrical control equipment. These investments are likely to generate measurable equipment demand earlier than the planned deployment of humanoid robots.
The decision to place new facilities outside the main semiconductor centers also reflects capacity constraints in existing industrial areas. Large semiconductor and data center projects require reliable electricity, water, logistics infrastructure and skilled labor, as well as access to equipment and material suppliers. Building an equivalent industrial base in new regions will take time, even with strong government support.
The physical AI component is more uncertain. Developing robot platforms and AI models is only one part of commercialization. Industrial users will also require reliable hardware, application-specific engineering, safety systems, maintenance networks and evidence that the equipment can reduce operating costs. Early adoption is therefore more likely in structured manufacturing and logistics environments than in applications requiring highly flexible human-like operation.
The main near-term impact of the strategy will come from the construction of semiconductors and data center capacity. Physical AI may become more important over the longer term, but its development will depend on whether South Korea can connect its strength in semiconductors and electronics with competitive robot hardware, components and practical industrial applications. PSR
Akihiro Komuro is Research Analyst, Far East and Southeast Asia, for Power Systems Research
