Korea plans to put 200,000 AI robots into service by 2030, starting with more than 1,700 government purchases in 2027
South Korea's National Advanced Strategic Industry Committee adopted a plan on September 30 to deploy 200,000 AI robots by 2030, about 15,000 through public procurement and 185,000 through private adoption. The state will buy more than 1,700 units in 2027 for public safety, labour-short sites and education, and will widen purchases to public enterprises and local governments from 2028.
South Korea will use government purchasing to build a home market for AI robots. At the ninth meeting of the National Advanced Strategic Industry Committee, chaired by Prime Minister Han Seong-sook at the Government Complex Seoul on September 30, ministers adopted a plan to put 200,000 AI robots into service by 2030 through public and private channels, according to the press release issued after the meeting by the Office for Government Policy Coordination together with the Ministry of Trade and Industry and the Ministry of Science and ICT.
The plan, presented as a programme to spread AI robots and physical AI in areas citizens can feel, splits the target in two. Public procurement is set at about 15,000 units from 2027 to 2030 and private adoption at about 185,000 units from 2026 to 2030, according to the agenda summary attached to the release. In 2027 the government itself will buy and distribute more than 1,700 robots. From 2028 public enterprises and local governments join as buyers, and the attachment says purchases will be expanded sharply in fields such as infrastructure management and defence.
- Government purchases, first year
- 2027
- 2027
- 1,700
- Public procurement incl. public enterprises and local governments
- 2027
- 2030
- 15,000
- Private-sector adoption
- 2026
- 2030
- 185,000
- Total target
- 2026
- 2030
- 200,000
Figures as disclosed in the attachment to the Sept 30, 2026 committee press release. The 2027 figure is stated as 1,700 or more. Public plus private equals total (15,000 + 185,000 = 200,000).
As of Oct 1, 2026
The first 1,700 units go to three fields, the attachment says. Public safety covers disaster response, policing and defence sites. Labour-short work covers agriculture and welfare, where repetitive and heavy tasks are hard to staff. Education and culture covers research humanoids for universities and patrol robots for historic sites. Only robots that have finished development and field testing will be bought first. To answer concerns that machines underperform or sit unused after delivery, the government said it will run an after-sales management system and feed field requirements back into R&D. Procurement rules will encourage domestic components, and outstanding products will receive certificates.
On the private side, the plan offers purchase subsidies, rental support and regulatory changes. The trade ministry and the SME ministry will support adoption of at least 1,000 robots a year by small and mid-sized companies with a 50% subsidy. A rental programme that lets companies use robots without upfront cost is to be designed by the first half of 2027. Safety guides for humanoids on industrial sites and for quadrupeds operating outdoors will be drawn up, and 30 billion won is earmarked in 2027 to develop and commercialise humanoids specialised for ten industries.
Manufacturing capacity is the third pillar. The government plans a large robot cluster in Saemangeum and North Jeolla built around a robot foundry that produces for startups without factories of their own, with stronger R&D incentives for Hyundai Motor suppliers investing there. The Daegu and Gyeongbuk region, designated a national advanced-industry zone for robotics at the same meeting, becomes the second production base. The attachment lists 16 lead companies for that zone, among them HD Hyundai Robotics, SL, Neuromeka, Jahwa Electronics, Samsung Electronics and LG Innotek, and total investment of 5.3839 trillion won through 2032, of which 3.7754 trillion won comes from the lead companies. The zone's stated goal is 20% of global humanoid output by 2030, against 1% today.
The fourth element is data. The government will pool field data from cross-ministry physical AI pilots into a shared library, build a training centre that produces and validates data for general-purpose scenarios, and train a sovereign physical AI foundation model on what the robots collect. The release describes a data flywheel in which state-bought robots improve from the work they do. Postal logistics is named as the first labour-short site for physical AI services, alongside regional manufacturing in North Jeolla and South Gyeongsang. Public safety examples include rescue drones on bridges, surface robots that collect marine litter and quadrupeds that spray water at fires.
The government's own diagnosis explains the urgency. The attachment puts China's share of the humanoid market at 87%, the United States at 3% and Korea at 1%. In quadrupeds China holds 51%, the US 3% and Korea 1% or less. Korean companies remain focused on R&D because demand is thin and technology lags, and mass production is still at the preparation stage, the release says, warning that the industry could lose its competitive base if the early market is not opened now.
- Humanoid
- 87
- 3
- 1
- Quadruped
- 51
- 3
- 1
Shares as stated in the committee attachment. Korea's quadruped share is given as 1% or less and shown here as 1. Measurement basis and period were not disclosed.
As of Oct 1, 2026
Market data show the size of the gap. IDC said on September 29 that China shipped more than 19,000 humanoids in the first half of 2026, about 77.9% of a global total of nearly 25,000. Korea's 1,700 public units in 2027 would be modest beside Chinese volumes, yet they would give Korean makers such as Rainbow Robotics and Doosan Robotics, and the conglomerate units named for the Daegu and Gyeongbuk zone, a guaranteed first buyer of a kind they have not had.
The release does not give the budget for the 2027 purchases or the unit prices the government expects to pay, and that money still has to pass through the annual budget process. The 20% output goal implies a twentyfold rise in Korea's share in about four years while Chinese makers keep adding capacity. A preference for domestic parts may protect suppliers but raise costs, and deploying robots at disaster and defence sites sets a reliability bar that early humanoids have yet to clear in public data.
The attachment sets out the next steps. Each new specialised zone will form a promotion team and publish a detailed development plan in the second half of 2026, and a cross-ministry support council will start in the same period. The rental programme design is due by the first half of 2027, and the first government purchases follow during 2027.
ROBOTNESS analysis
Korea is moving its robot policy from funding research to buying output, and the 15,000 public units matter less for their volume than for the field data and reference sites they create.
The plan ties every procurement channel to data. State-bought robots feed a shared library and a sovereign foundation model, and procurement favours domestic parts. The design echoes the path that helped Chinese makers scale, where IDC data show research, education, demonstration and government data-collection centres still took 69% of humanoid shipments in the first half of 2026.
The strongest counter-argument is that public buying cannot replace commercial demand. China's early volumes rested on a deep domestic supply chain, while the Korean attachment itself concedes that local makers lack both demand and technology. Buying immature robots at scale could lock in weak products.
In the bull case the 2027 tranche reaches users on time, after-sales rules keep robots working, and Korean makers turn public reference sites into private orders under the 50% SME subsidy. Lead companies in Daegu and Gyeongbuk convert auto-parts lines to robot components, and Korea becomes a credible non-Chinese source of humanoid parts.
In the bear case budget allocations fall short of the unit targets, orders concentrate on a few incumbents and robots become showcase equipment. Chinese price competition keeps Korean hardware uncompetitive outside the public channel, and the 20% output goal fades from view.
Signals to watch
- Second half of 2026: the detailed development plan for the Daegu and Gyeongbuk robot zone and the launch of the cross-ministry support council.
- December 2026: final 2027 budget lines for robot procurement and the 30 billion won industry humanoid programme.
- First half of 2027: the design of the robot rental programme and the first public tender documents.
- 50% subsidy for at least 1,000 robots a year
- 9th National Advanced Strategic Industry Committee, Sept 30, 2026
- about 185,000 units, 2026 to 2030
- 200,000 AI robots by 2030
- 10 types, 30 billion won in 2027
- about 15,000 units, 2027 to 2030
- 20% of global output by 2030, from 1% now
- more than 1,700 units
- 5.3839 trillion won through 2032 (16 lead companies)
Why it matters
Korea has funded robot R&D for two decades, but this is the first time the state has committed to a multi-year unit target for buying AI robots itself. The 15,000 public units over 2027 to 2030 create a predictable first customer for domestic makers that, by the government's own account, are stuck at the R&D stage because no one is placing volume orders.
The plan also binds procurement to data collection and a sovereign physical AI foundation model. That makes the programme an industrial data strategy as much as a subsidy, and it gives Seoul a lever over which robot software stacks get trained on Korean field data.
Rival analysis
China is the benchmark the plan is written against. The government puts China's humanoid share at 87% and Korea's at 1%, and IDC counted more than 19,000 humanoid shipments in China in the first half of 2026 alone. Chinese makers reached volume partly through research, education and government data-centre demand, a pattern Korea's public procurement now imitates.
The US route is different, built on private capital for companies such as Figure AI, Apptronik and Agility Robotics and on factory pilots with carmakers. Korea's model sits between the two, with state demand at the front and conglomerate investment in the Daegu and Gyeongbuk zone behind it.
Valuation context
No procurement budget or unit price is disclosed, so the revenue pool for Korean makers cannot be sized yet. The disclosed money is on the supply side, 5.3839 trillion won of investment in the Daegu and Gyeongbuk zone through 2032 and 30 billion won for ten industry humanoids in 2027.
For listed Korean robot makers, the plan lowers demand risk for 2027 to 2030 but adds execution risk, because orders will likely be split across many vendors and fields, and public buyers will demand after-sales service that eats into margins.
Supply-chain implications
Procurement rules will encourage domestic parts, and the Daegu and Gyeongbuk zone is meant to convert regional auto-parts suppliers into robot component makers. The Saemangeum robot foundry is designed to give startups contract manufacturing they cannot build themselves, with incentives for Hyundai Motor suppliers to invest.
The constraint is that actuators, reducers, sensors and compute for humanoids are still largely sourced from China, Japan and the US. A domestic-parts preference may raise unit costs in the first tranches and push vendors to localise reducers and motors faster than their volumes justify.
Signals to watch
The second half of 2026 should bring the detailed Daegu and Gyeongbuk zone plan, which will show how the 16 lead companies split their 3.7754 trillion won. The final 2027 budget in December will reveal whether the 1,700-unit tranche is fully funded.
By the first half of 2027 the rental programme design and first tender documents should name robot categories, technical requirements and any domestic-content thresholds, which will determine which makers can bid.
Analyst view
Thesis: the plan is a credible demand signal but its success depends on budget follow-through and after-sales discipline. Confidence is medium. The unit split and timeline are specific and tied to named ministries, which raises credibility, but the absence of a budget figure and unit prices leaves the largest uncertainty open.
We would raise confidence if the 2027 budget funds the full 1,700 units and if tender rules reward field performance rather than headline specifications. We would lower it if procurement is spread so thin that no vendor reaches meaningful production volume.
Questions you should be asking
What is the 2027 budget for the 1,700 units, and how many of them will be humanoids rather than quadrupeds, drones or service robots?
How will domestic-parts preferences be measured, and will they apply to compute and AI models as well as hardware? Who will own and access the data collected by state-bought robots, and will private makers be able to train on it?