[DigitalToday reporter Chi-gyu Hwang (황치규)] As generative AI spreads and investment in AI infrastructure expands, South Korea's AI business is maintaining rapid growth, a study finds. It also finds that profitability and cash flow are increasingly concentrated in specific industries. Growth is relatively concentrated among semiconductor and materials, parts and equipment makers.

AI and digital market research firm Knowledge Research Group (KRG) analysed 80 AI-related companies listed on the KOSPI and KOSDAQ. Total revenue rose 55.8 percent over two years, to 382.8 trillion won in 2025 from 245.7 trillion won in 2023.

Net profit increased over the same period to 81.0 trillion won from 46.9 trillion won, posting a compound annual growth rate of 79.3 percent.

By sector, semiconductors and materials, parts and equipment showed overwhelmingly strong growth.

Combined revenue for the two sectors rose to 336.2 trillion won in 2025 from 207.8 trillion won in 2023, recording a compound annual growth rate of 27.2 percent over three years. Net profit rose over the same period to 76.1 trillion won from 21.9 trillion won, an increase of nearly 50 trillion won. Net profit margin also climbed sharply to 22.6 percent in 2025 from 10.5 percent in 2023.

KRG analysed that this shows the core value of the global AI industry is concentrated in the AI infrastructure supply chain, including GPUs, HBM, advanced packaging, inspection and testing, and semiconductor materials.

KRG said, "The area generating the most profit in the AI industry now is not general-purpose generative AI services but the AI semiconductor supply chain." It added, "This means the AI industry is evolving beyond a software-centred market into a computing infrastructure-centred industry."

The LLM and agent segment is also showing steady growth. Total revenue rose to 10.2 trillion won in 2025 from 8.5 trillion won in 2023. Net profit expanded to 2.1 trillion won from 1.6 trillion won. Profitability is also holding at around 20 percent.

South Korea's market, however, has yet to sufficiently form a recurring revenue structure based on large APIs like the United States. Many companies maintain business structures centred on public projects, build-to-order projects, customised enterprise AI and proof-of-concept work. It was analysed that they are securing profitability above a certain level as demand expands for enterprise-operational AI such as document AI, work automation, AI search, enterprise AI and AI agents.

The data centre and cloud sector plays a role as core foundational infrastructure for the AI industry, but faces relatively heavy profitability burdens, the study shows. Total revenue rose to 15.0 trillion won in 2025 from 13.4 trillion won in 2023, but net profit fell to 1.1 trillion won in 2025 from 1.7 trillion won in 2024. Net profit margin also slipped to 7.6 percent from 11.5 percent.

That is attributed to a full-scale race to build AI data centres, which is simultaneously expanding spending on securing GPUs, building data centres, power infrastructure, cooling facilities and network investment. KRG said, "The data centre industry is currently at a stage where pre-emptive facility investment comes before top-line growth." It added, "For domestic cloud operators, achieving economies of scale is emerging as an important task amid competition with global CSPs."

The physical AI and robot industry is still showing strong early-market characteristics, the study said. Revenue is rising to 4.5 trillion won in 2025 from 4.0 trillion won in 2023, but profitability is highly volatile. That is because most areas, including humanoid robots, industrial robots and autonomous manufacturing systems, remain in an investment stage aimed at expanding the market and securing technology. In the global market, securing technological competitiveness and pre-empting ecosystems are emerging as key tasks, and full-scale profit generation is likely to come after the next few years, KRG analysed.

The medical AI sector recorded a loss of about 62.6 billion won in 2025 despite revenue growth. That was analysed to be because high entry barriers, including FDA approval, securing clinical data, global certification and insurance reimbursement systems, mean commercialisation takes a long time. The autonomous driving and mobility sector is also seeing rising revenue, but with various tasks remaining, including technological stability, regulation and liability systems, it showed a return to losses in 2025.

A KRG official said, "The AI industry is now moving from a competition in technology to a competition in industrial operation, and from a competition in expectations to a competition in cash flow." The official added, "The winners in the AI market will be companies that can generate profit in actual industrial sites rather than the technology itself."