Korea ranks No. 3 in global physical AI patent race

South Korea ranked third behind the United States and China in multicountry physical AI patent filings, placing it among the global leaders in a field expected to benefit from the country’s broad manufacturing base.

The findings suggest Korea, which lagged in the software-centric generative AI race, may have an opportunity to catch up in the next phase of AI competition by leveraging its manufacturing infrastructure.

A joint analysis by Maeil Business and global patent analytics company LexisNexis, released Sunday, examined physical AI patents filed worldwide from 2016 through 2025.

Korea accounted for 22 percent of multicountry patent families, defined as inventions filed in at least two jurisdictions. That placed it behind the United States at 29 percent and China at 25 percent.

Korea’s share was more than double its 11 percent in generative AI. Physical AI patent filings increased at an average annual rate of 55.6 percent over the decade, reflecting a rapid shift in the global industrial landscape.

Experts attributed Korea’s performance to its full-stack value chain. Unlike China, where physical AI patents are concentrated among research institutions, or the United States, where they are weighted toward AI platforms and software, Korea has domestic capabilities spanning AI chips, control algorithms, finished devices and manufacturing operations.

Samsung Electronics, LG Electronics and Hyundai Motor all ranked among the top 20 organizations in the Patent Asset Index, or PAI.

The index measures the overall strength of a patent portfolio by aggregating the technological and economic value of its patents.

Physical AI enables AI systems to perceive and control machinery and other physical devices in industrial settings. As AI competition increasingly shifts toward real-world applications, leading countries are racing to establish an early advantage.

China accounted for about 67 percent of all physical AI patent filings, dominating in volume. Its filings, however, were heavily concentrated among universities and research institutions such as the Chinese Academy of Sciences, Harbin Institute of Technology and Tsinghua University.

Their patents showed significantly less commercial impact than those held by large corporations and were concentrated in areas such as humanoid body structures, according to the analysis. That could limit their immediate application in industry.

The United States maintained strong patent influence through technology companies including Alphabet and Nvidia. Patent cluster analysis, however, showed that U.S. big-tech portfolios were concentrated in AI platforms and computing infrastructure.

Few big-tech patents appeared in physical application clusters such as precision manufacturing control, where intellectual property was fragmented among traditional industrial companies.

Korea’s patent clusters were distributed more evenly across the value chain, from AI semiconductors and algorithms to physical-control platforms. The country was also found to have built integrated portfolios connecting AI intelligence with real-world control systems.

By comparison, Korea’s share of multicountry patents in generative AI stood at 11 percent, far behind the United States at 43 percent and China at 27 percent.

“Government support can inflate domestic filing figures in national strategic technologies such as AI, so patents filed in at least two countries should also be compared,” said Kim Dong-hyun, a senior manager at LexisNexis.

“Multicountry filings help reveal where meaningful innovation is actually taking place and indicate companies’ and countries’ intentions to enter global markets and commercialize their technologies.”

Korea’s AI deep-tech ecosystem is also expanding alongside the platforms being developed by large companies.

Ultralow-power edge AI chipmaker DEEPX has built a portfolio of more than 90 active patents, rapidly expanding its intellectual property in system operations and computing. Medical AI has strengthened its portfolio in medical diagnostics and health information with 131 active patents.

Rebellions, which develops AI accelerator infrastructure; Nota AI, which specializes in lightweight vision AI technology; and MakinaRocks, which is pursuing a Kosdaq listing this year, have also increased their PAI scores.

The companies are emerging as participants in a domestic full-stack value chain linking AI chips, software and manufacturing platforms.

“Korea may have shortcomings when hardware or software is considered separately, but it is the only country with balanced capabilities across all areas,” said Jang Young-jae, a professor in KAIST’s Department of Industrial and Systems Engineering.

“The expertise Korea has accumulated in building and operating actual factories will become a major competitive advantage in the physical AI era.”

The government is already developing physical AI systems for operating factories. Physical AI demonstration laboratories at KAIST and Jeonbuk National University are running real manufacturing processes.

“In the physical AI market, an ecosystem strategy that allows AI models, semiconductors and hardware to operate together compatibly is essential,” said Lee Jun-woo, physical AI project manager at the National IT Industry Promotion Agency (NIPA).

“Korea’s long-developed manufacturing competitiveness is a blessing even from the perspective of the AI market.”

Physical AI technology developed in Korea has already produced measurable results at industrial sites.

Tests led by Jeonbuk National University Professor Kim Soon-tae increased production by 7.4 percent at DH Autolead, 11.4 percent at Daeseung Precision and 5.1 percent at Donghae Metal.

“If we combine Korea’s manufacturing capabilities with the physical AI technologies now being developed, the country could enter a new phase in both manufacturing and AI,” Kim said.