Korea makes physical AI one of three national mega projects, aiming for global top three in AI robots by 2030
Seoul on June 29 named physical AI alongside semiconductors and AI data centres as a national mega project and set out a '3M' plan for AI robots. The plan targets ten industry-specific humanoids commercialised by 2028, 1,000 robot deployments a year, data factories for ten sectors and a robot foundry in Saemangeum tied to Hyundai Motor Group investment.
South Korea's government on June 29, 2026 put robots at the centre of its industrial policy, naming physical AI as one of three 'mega projects' at a national briefing chaired at the presidential office, according to a reference document released by the Ministry of Trade, Industry and Energy. The other two projects are semiconductors and AI data centres.
The robot section sets a goal for Korea to become one of the world's three leading AI robot nations, and the physical AI section aims for global leadership in the field by 2030, the document said. The government framed the plan as a shift from a country that uses robots well to one that builds them.
The numbers behind the plan
The ministry's own diagnosis is blunt. It cites International Federation of Robotics data showing Korea has the highest robot density in the world at 1,220 robots per 10,000 workers, ahead of Singapore at 770 and China at 470. Yet Korea accounts for only 1% of global robot production, and the document cites OMNIA figures putting Korea's share of 2025 humanoid output at 1%, against 86% for China and 4% for the United States. It also quotes a Bank of America forecast that cumulative global humanoid production will rise from 20,000 units in 2025 to 300 million by 2040.
- South Korea
- 1
- 1220
- China
- 86
- 470
- United States
- 4
- No data
- Singapore
- No data
- 770
Humanoid share from OMNIA and robot density from IFR, as cited in the ministry's June 29, 2026 briefing document. Figures disclosed, not estimated; null where the document gives no figure.
As of Oct 1, 2026
What the 3M strategy commits to
The plan is organised under three headings the ministry calls 3M. The first, M.AX acceleration, uses the existing manufacturing AI alliance as the forum for robot adoption across industries, develops 'specialist' AI robots for individual sectors and targets ten industry-specific humanoids for commercialisation in 2028. It also links robot rollout to the government's AI factory programme with a goal of 1,000 deployment sites a year, followed by export packaging.
The second heading, Master, covers capabilities. The government will set up data factories for ten industries to produce robot training data, develop a Korean foundation model tuned to ten core sectors so that Korean industry does not depend on platforms led by NVIDIA and Google, and create dedicated R&D for three components it judges weaker than rivals: actuators, robot hands and sensors. It also plans rapid robot-specific development in areas where Korea is strong, citing all-solid-state batteries for robots as an example, and a cross-ministry programme to train 10,000 robot specialists.
The third heading, Mass Production, is the most concrete on geography. Saemangeum on the west coast is to host a robot foundry that manufactures for startups without their own production capacity, plus parts-cluster infrastructure intended to draw follow-on investment from automotive suppliers after what the document calls Hyundai Motor's large-scale investment. The Daegu and Gyeongbuk region is to get a robot test field tied to nearby car, shipbuilding and electronics plants, along with support for car-parts makers converting to robot components. The plan adds a cross-ministry robot demand team, help with purchase subsidies and replacement of ageing robots, standardisation of humanoid parts such as batteries and hands, and financing for new capacity through the National Growth Fund.
The Saemangeum Development Agency said separately on June 24 that Hyundai Motor Group signed an investment agreement in February covering AI data centres, robots and green hydrogen in the reclaimed-land zone, without disclosing a figure in that release.
The physical AI layer
A second chapter treats physical AI as a national strategic industry. It calls for large synthetic-data infrastructure, a world model that generates physically consistent training data, digital twins of regional sites, a government-wide data library combining public and private datasets, and a world-class general-purpose model the document calls a 'K-model', from which specialised models for tasks such as welding and parcel sorting would be derived with small amounts of field data. It also lists purpose-built devices such as humanoids, NPUs for physical AI services and low-latency networks.
On the same day the ministry, together with the science and SME ministries, published a 'Manufacturing AI 2030' strategy that targets 20 trillion won of public and private investment by 2030 and more than 100 trillion won of added value, and links to the 150 trillion won National Growth Fund. That strategy lists commercialisation of manufacturing-specialised humanoids, through core-component development and pilots, among its advanced-stage tasks.
Context and rivals
The document explicitly positions the plan against China, which it describes as already in mass production, and it adopts a matching tactic: rather than perfecting robots before launch, Korea will push rapid mass production for cost reduction and fix problems afterwards. The reference to NVIDIA and Google as the leaders in robot foundation models signals that Seoul sees software dependence, not only hardware share, as the strategic risk.
Risks and open questions
The document contains no budget for the robot chapter as a whole, no allocation from the National Growth Fund to robotics and no figure for Hyundai's Saemangeum investment. The 2028 humanoid target depends on component programmes whose funding was still to be set in the 2027 budget cycle, and the foundation-model goal pits a state-led effort against private rivals that spend at a far larger scale.
ROBOTNESS analysis
Korea's 3M plan is credible on deployment and weak on the two things that decide humanoid economics, component cost and model ownership.
The evidence for the deployment side is strong. Korea already runs the densest robot installed base in the world, the plan ties rollout to existing AI factory and alliance channels, and a target of 1,000 sites a year is modest enough to be met. The Saemangeum foundry addresses a real bottleneck for startups that can design robots but cannot build them.
The weakness is that the document identifies actuators, hands and sensors as gaps without saying how much will be spent closing them, and a 1% share of humanoid output means Korean parts makers lack the volume that drives cost down in China.
The strongest counter-argument is that Korea does not need to win on volume. Its automotive, shipbuilding and electronics plants are captive demand that can absorb specialist humanoids at premium prices, as the industry-specific design of the ten target robots implies.
Bull case. If the 2027 budget funds the component R&D and Hyundai's Saemangeum investment brings its supplier base into robot parts, Korea could have a domestic humanoid supply chain by 2028 that serves its own factories first and exports second. Captive demand would give suppliers the learning curve they currently lack.
Bear case. Without disclosed money the plan risks becoming a list of intentions, while Chinese makers keep cutting prices and US model developers widen their lead. Korean factories would then keep buying foreign robots and foreign models, repeating the pattern the document itself describes.
Signals to watch.
- The 2027 government budget proposal and its humanoid component line, expected in the autumn of 2026.
- Designation of the robot mega special zone and the robot foundry operator in Saemangeum during the second half of 2026.
- First field results from the ten industry-specific humanoid projects ahead of the 2028 commercialisation target.
- 10,000 robot specialists
- June 29, 2026, national briefing on three mega projects
- Global top three by 2030
- Ministry of Trade, Industry and Energy
- Saemangeum robot foundry; Daegu and Gyeongbuk test field
- Actuators, robot hands, sensors
- 1,000 sites a year
- 10 types, commercialisation target 2028
- 20 trillion won investment, over 100 trillion won value
- 1% (China 86%, US 4%)
Why it matters
This is the first time Seoul has placed robots in the top tier of its industrial agenda alongside memory chips, which historically received the deepest state support. Inclusion as a mega project gives the robot file presidential attention, cross-ministry coordination and a claim on the National Growth Fund, which matters more than any single programme budget.
The framing also changes. Earlier Korean robot plans focused on adoption and dissemination subsidies. The 3M plan targets production share, component sovereignty and model ownership, which puts Korean suppliers in direct competition with Chinese humanoid makers and US model developers rather than positioning them as buyers.
Rival analysis
China is the explicit benchmark. The ministry's own figures give China 86% of 2025 humanoid output, and the document's tactic of mass production first and refinement later is borrowed from Chinese practice. Korea's answer is specialisation: ten industry-specific humanoids rather than a general-purpose platform.
On software, the document names NVIDIA and Google as leaders of robot foundation models. Korea's proposed K-model would compete with privately funded US models, and with Japan's industrial alliances that are building on NVIDIA technology. The realistic niche is fine-tuned sector models trained on Korean factory data, which no foreign developer holds.
Valuation context
The plan sets no robot-specific budget, so its market impact runs through expectations. Listed Korean robot companies such as Doosan Robotics, Rainbow Robotics and Robotis are the obvious beneficiaries of demand-side measures like public procurement and purchase subsidies.
The National Growth Fund is described at 150 trillion won in the companion Manufacturing AI 2030 strategy, but no robot allocation is stated in the primary documents. Investors should treat any reported robot share of the fund as unconfirmed until the fund's managers publish it.
Supply-chain implications
The component list is the most important part for suppliers. Actuators, robot hands and sensors are named as weak points, while batteries and semiconductors are named as strengths to adapt. That points to Korean motor, reducer and sensor makers, and to battery makers developing robot-specific cells.
Saemangeum and Daegu-Gyeongbuk split the supply chain by role. Saemangeum gets contract manufacturing and parts clusters for car suppliers following Hyundai, while Daegu-Gyeongbuk gets testing and the conversion of auto-parts makers. Standardising humanoid batteries and hands is intended to let suppliers serve several robot makers at once.
Signals to watch
First, the 2027 budget: the humanoid component line will show whether the Master pillar is funded. Second, operator selection for the Saemangeum robot foundry, which will reveal whether Hyundai or a third party runs it.
Third, the data factories: the first sector chosen and its data-sharing rules will indicate whether the K-model can be trained on real factory data at scale. Fourth, any disclosed robot allocation from the National Growth Fund.
Analyst view
Thesis: the plan will deliver deployment and demand faster than it delivers component or model independence. Confidence: medium. Reasons: the deployment levers already exist and are cheap, while the component and model goals lack disclosed funding and face better-financed foreign rivals.
The main upside risk to this view is Hyundai Motor Group. If its Saemangeum investment includes in-house actuator and humanoid production, the supply-chain goals could move faster than state funding alone would allow.
Questions you should be asking
How much of the National Growth Fund will be reserved for robot companies, and on what terms? Who will own and operate the Saemangeum robot foundry?
Which ten industries will receive data factories and specialist humanoids, and will data generated in them be available to all alliance members or only to project participants?