Physical AI

Sector Overview: Where AI Meets the Physical World

Physical AI, the application of artificial intelligence to machines that perceive, reason about, and manipulate the physical world, represents arguably K-Moonshot's most commercially visible and publicly captivating sector. While other K-Moonshot domains operate in laboratories or data centres, physical AI manifests in the tangible form of robots walking through factories, handling objects in warehouses, and eventually operating in households and public spaces. Mission 6 (Humanoid Robots) is the sector's flagship initiative, supported by Mission 7 (General-Purpose Physical AI Models and Computing Platforms) which provides the software intelligence that animates physical systems.

South Korea brings formidable industrial assets to the physical AI competition. Hyundai Motor Group owns an 80 percent stake in Boston Dynamics, the world's most recognizable robotics company. Samsung Electronics has acquired an approximately 15 percent stake in Rainbow Robotics, becoming the company's largest shareholder and embedding itself in Korea's indigenous humanoid development programme. Doosan Robotics, following a successful 2023 IPO that raised approximately 500 billion KRW, is expanding from collaborative industrial robots into AI-powered autonomous systems. LG Electronics has developed the KAPEX platform for AI-driven robotic intelligence. Together, these companies create a Korean robotics cluster that is, in aggregate, the most diversified national robotics ecosystem outside the United States and China.

The Humanoid Strategy Council: National Coordination

On February 27, 2026, the Korean government launched the Humanoid Strategy Council, a national coordination body designed to align government policy, corporate investment, and research priorities in humanoid robotics development. The council, chaired by the Minister of Science and ICT and including representatives from Hyundai, Samsung, LG, Doosan, and leading research institutions, provides the institutional framework through which K-Moonshot's physical AI ambitions are translated into coordinated action.

The council's formation reflects a recognition that humanoid robotics development requires integration across multiple domains: mechanical engineering, AI software, sensor technology, power systems, materials science, and manufacturing. No single Korean company, however large, possesses the full capability stack. The council's mandate includes coordinating research investments to avoid duplication, establishing common standards for humanoid robot components and interfaces, creating shared testing and certification infrastructure, and developing the regulatory framework for humanoid robot deployment in workplace and public environments.

Hyundai Motor Group and Boston Dynamics

Hyundai Motor Group's acquisition of Boston Dynamics in 2021 for approximately USD 1.1 billion represented a strategic bet that has only grown in significance. Boston Dynamics, founded at MIT in 1992, has produced some of the most technically advanced robots in history: Atlas (a bipedal humanoid capable of dynamic acrobatics), Spot (a quadruped robot deployed in industrial inspection, construction, and public safety), and Stretch (a mobile robot designed for warehouse box-moving operations).

Under Hyundai ownership, Boston Dynamics has transitioned from a research-focused organization into a commercial entity with growing revenue from Spot robot sales and Stretch warehouse deployments. The integration of Boston Dynamics' advanced locomotion and manipulation capabilities with Hyundai's manufacturing scale, automotive engineering expertise, and global supply chain creates a combination that few competitors can replicate.

Hyundai's planned 9 trillion KRW innovation hub in Gunsan, South Jeolla Province, includes a dedicated robotics manufacturing cluster targeting production of 30,000 humanoid units per year. This production target, if realized, would establish one of the world's first mass-manufacturing facilities for humanoid robots, a step that transforms humanoids from laboratory demonstrations into commercially deployable systems. The Gunsan hub also integrates with Hyundai's broader manufacturing operations, including automotive production lines where humanoid robots are expected to serve as flexible factory workers capable of performing tasks currently requiring human labour.

The strategic logic extends beyond manufacturing. Hyundai's Supernal urban air mobility division, its construction equipment division, and its logistics operations all represent deployment contexts for physical AI systems. The company's vision encompasses a mobility ecosystem where autonomous vehicles, delivery robots, aerial platforms, and humanoid workers operate within a unified AI coordination framework, a concept that aligns directly with Mission 7's computing platform objectives.

Samsung and Rainbow Robotics

Samsung Electronics' investment in Rainbow Robotics represents the company's primary vector into humanoid robot development. Rainbow Robotics, founded in 2011 by a team of KAIST robotics researchers, developed the HUBO humanoid robot that won the 2015 DARPA Robotics Challenge. The company has since commercialized collaborative robot (cobot) arms while continuing to advance its humanoid platform.

Samsung's approximately 15 percent stake, acquired through a series of investments beginning in 2023, makes it Rainbow Robotics' largest shareholder. The partnership combines Samsung's AI semiconductor capabilities (mobile processors, NPUs, and image sensors), manufacturing scale, and global distribution with Rainbow Robotics' mechanical engineering expertise and humanoid locomotion algorithms. Samsung's AI Research Centre in Seoul has established dedicated research programmes in robot perception, manipulation planning, and human-robot interaction that feed directly into the Rainbow Robotics platform.

The Samsung-Rainbow partnership positions Korea to develop a humanoid robot that leverages Samsung's semiconductor and display technology for sensing and computation, creating a vertically integrated system where the robot's brain (AI processor), eyes (image sensors), and skin (flexible sensors) are all Korean-developed components. This vertical integration mirrors the approach that has historically driven Korean competitiveness in smartphones, displays, and memory semiconductors.

Doosan Robotics: From Cobots to Autonomous Systems

Doosan Robotics, a subsidiary of Doosan Group, has established itself as Korea's leading collaborative robot manufacturer. The company's cobot product line, ranging from 1 kilogram to 25 kilogram payload capacity, serves applications in electronics assembly, food and beverage handling, and laboratory automation. Doosan Robotics' 2023 IPO on the Korea Exchange raised approximately 500 billion KRW, providing capital for expansion into AI-powered autonomous systems.

Doosan's physical AI strategy focuses on mobile manipulation: combining cobot arms with autonomous mobile platforms to create systems capable of navigating unstructured environments while performing manipulation tasks. This approach, sometimes described as the mobile cobot or AMR-cobot integration, addresses a market need distinct from full humanoid robots. Factories, warehouses, and service environments often need flexible automation solutions that can move between workstations, adapt to changing layouts, and collaborate safely with human workers, requirements that mobile manipulation platforms can satisfy without the full complexity of bipedal humanoid locomotion.

Doosan Robotics' AI capabilities include computer vision-based object recognition, force-torque sensing for delicate manipulation, and cloud-connected fleet management for multi-robot coordination. The company is investing in AI model development for robotic task planning and execution, drawing on the broader AI Science ecosystem's foundation model capabilities.

LG Electronics: The KAPEX Platform

LG Electronics has approached physical AI through a platform strategy centered on its KAPEX (Korean AI Platform for EXcellence) system. KAPEX provides a unified AI software framework for robotic perception, decision-making, and action across LG's diverse robotic product portfolio, which includes commercial cleaning robots (CLOi), hotel and restaurant service robots, and industrial guide robots deployed in Korean airports and shopping centres.

LG's physical AI strategy is differentiated from the humanoid-focused approaches of Hyundai and Samsung. Rather than pursuing bipedal humanoid form factors, LG has focused on application-specific robot platforms optimized for particular commercial environments. The company's deployment base, with thousands of service robots operating in Korean commercial facilities, provides real-world operational data for training and refining AI models, a data advantage that is difficult to replicate without comparable deployment scale.

LG AI Research, the company's dedicated AI research laboratory, has developed multimodal AI models that integrate visual, auditory, and spatial perception for robotic applications. The laboratory's work on EXAONE, LG's foundation AI model, includes physical AI extensions that enable robots to understand natural language instructions, interpret spatial environments, and plan multi-step actions. This integration of foundation models with robotic execution connects LG's physical AI programme directly to Mission 7's general-purpose physical AI model objectives.

The Global Humanoid Competition

Korea's physical AI ambitions unfold within an intensely competitive global landscape where multiple well-funded contestants are racing toward humanoid robot commercialization.

Tesla Optimus (United States)

Tesla's Optimus humanoid robot programme leverages the company's AI capabilities developed for autonomous driving (particularly computer vision and neural network training at scale), its battery technology expertise, and its manufacturing cost-reduction culture. Tesla's target of deploying thousands of Optimus units in its own factories by the mid-2020s, followed by external sales, represents the most aggressive commercialization timeline among major humanoid programmes. Elon Musk has publicly stated ambitions for Optimus to become Tesla's most valuable product line, with eventual production volumes in the millions. However, the gap between demonstration videos and reliable autonomous operation in unstructured environments remains substantial.

Figure AI (United States)

Figure AI, founded in 2022, has raised over USD 1.5 billion in funding from investors including Microsoft, NVIDIA, OpenAI, Amazon, and Jeff Bezos. The company's Figure 02 humanoid has demonstrated impressive manipulation capabilities in controlled environments and has established deployment partnerships with BMW for automotive manufacturing applications. Figure's approach integrates large language models with robotic control, enabling humanoids to respond to natural language instructions and reason about physical tasks.

1X Technologies (Norway)

1X Technologies, backed by OpenAI and Tiger Global, is developing the NEO humanoid for household applications. The company's approach prioritizes safe human-robot coexistence and cost-effective design over maximum performance, targeting a price point and safety profile suitable for consumer markets. 1X's partnership with OpenAI provides access to advanced AI models for robot reasoning and instruction following.

Chinese Competitors

China's humanoid robotics ecosystem has expanded rapidly, with companies including Unitree Robotics, Fourier Intelligence, Agility Robotics (US-China operations), Ubtech, Agibot, and Galbot all developing humanoid platforms. The Chinese government has designated humanoid robotics as a strategic industry, with multiple provincial and municipal governments offering subsidies, land, and regulatory support for humanoid robot development and manufacturing. The scale of Chinese investment, combined with the country's manufacturing cost advantages and enormous domestic market, presents a competitive challenge that Korean companies cannot afford to underestimate.

Korea's Competitive Positioning

In this crowded field, Korea's physical AI sector holds several distinctive advantages and faces specific challenges.

Advantages: Hyundai's ownership of Boston Dynamics provides access to the world's most advanced locomotion and dynamic balance technology. Samsung's semiconductor and sensor capabilities enable Korean-designed robots to use domestically produced AI processors, image sensors, and displays. Korea's manufacturing culture, honed through decades of automotive, electronics, and shipbuilding production, is directly transferable to robot manufacturing. The national coordination through the Humanoid Strategy Council ensures aligned investment and avoids duplication. And Korea's aging demographic, with one of the world's lowest birth rates (0.72 in 2024), creates urgent domestic demand for robotic labour substitution across manufacturing, services, and eldercare.

Challenges: Korea's AI software ecosystem, while growing, lags the US in foundation model capabilities for robotics. Boston Dynamics has historically prioritized hardware engineering over AI-driven autonomy, though this balance is shifting under Hyundai's ownership. Korean humanoid programmes have not yet demonstrated the production manufacturing readiness that Tesla claims for Optimus or that Chinese companies are targeting. And the talent pipeline for robotics AI, combining mechanical engineering with machine learning, is constrained by the same overall AI talent shortage that affects all K-Moonshot sectors.

The Industrial Robot Foundation

Korea's physical AI ambitions build upon an established industrial robotics base. Korea has the world's highest robot density in manufacturing, with approximately 1,000 robots per 10,000 manufacturing employees, the highest ratio globally according to the International Federation of Robotics. This installed base, concentrated in automotive production, electronics assembly, and semiconductor fabrication, provides both the operational context for physical AI deployment and the manufacturing expertise for producing robotic systems at scale.

Hyundai Robotics (the industrial robot division, distinct from the humanoid programme through Boston Dynamics) is Korea's largest industrial robot manufacturer, competing with Japanese firms Fanuc, Yaskawa, and Kawasaki, and European leaders ABB and KUKA. Korean manufacturers' deep understanding of factory automation workflows informs the design requirements for humanoid robots intended to operate in manufacturing environments, from the force limits needed for safe human-robot collaboration to the precision requirements for assembly tasks.

Software and AI Infrastructure

The software layer is where Korea's physical AI sector connects most directly to the broader AI Science and Semiconductor sectors. Humanoid robots require multiple AI capabilities operating simultaneously: computer vision for environment perception, language models for instruction understanding, motion planning algorithms for locomotion and manipulation, and reinforcement learning systems trained in simulation for adaptive behaviour.

Mission 7 (General-Purpose Physical AI Models and Computing Platforms) addresses this software requirement. The mission targets the development of foundation models specifically designed for physical world interaction, trained on robotics simulation data, human demonstration recordings, and real-world deployment telemetry. These models must operate on edge computing hardware embedded in the robot rather than relying entirely on cloud connectivity, creating demand for the domestic AI accelerator chips being developed under Mission 11.

Naver and Kakao, while primarily known as internet platforms, both operate robotics research programmes. Naver Labs has developed autonomous robots for building management and delivery, while Naver's ARC (AI Research Center) research in 3D spatial understanding and embodied AI directly supports physical AI model development. These capabilities complement the hardware-focused programmes at Hyundai, Samsung, and Doosan.

Manufacturing and Supply Chain

Humanoid robot manufacturing requires integration of actuators (motors and gears for joint movement), sensors (cameras, LiDAR, force-torque, inertial measurement units), computing hardware (AI processors, edge servers), power systems (batteries, power management), and structural components (lightweight alloys, carbon fibre composites, flexible coverings). Korea's industrial base provides domestic supply for many of these components: Samsung and SK produce AI processors and sensors, LG and Samsung SDI manufacture batteries, and Korea's steel and materials industry produces lightweight structural alloys.

The actuator supply chain is one area where Korea faces a dependency. High-performance robot actuators, particularly the harmonic drives and cycloidal gearboxes used in precision joint mechanisms, have historically been dominated by Japanese manufacturer Harmonic Drive Systems. Korean companies including SBB Tech and several startups are developing domestic alternatives, but achieving the precision, reliability, and cost-competitiveness of established Japanese products remains an ongoing challenge.

Hyundai's Gunsan hub production target of 30,000 humanoid units per year would require a fully industrialized supply chain for every component category. Achieving this scale will test Korea's ability to translate its manufacturing prowess in automotive and electronics into the distinct manufacturing requirements of humanoid robots, where smaller batch sizes, more complex assembly processes, and stringent safety testing create different production economics.

Demographic Driver: Korea's Aging Population

Korea's demographic trajectory provides perhaps the most compelling domestic demand signal for physical AI. The country's total fertility rate of 0.72 in 2024, the lowest among all OECD nations, implies a shrinking working-age population that will create labour shortages across manufacturing, services, healthcare, and agriculture within the next decade. The Bank of Korea has estimated that without productivity interventions, labour force contraction could reduce potential GDP growth by 0.5-1.0 percentage points annually through the 2030s.

Humanoid and service robots represent a direct response to this demographic challenge. Factory humanoids can supplement a declining manufacturing workforce. Service robots can fill hospitality, retail, and logistics roles. Care robots can assist an expanding elderly population (Korea is projected to become a super-aged society by 2025, with over 20 percent of the population aged 65+). This alignment between demographic necessity and technological capability creates a domestic market pull that both justifies K-Moonshot's investment in physical AI and provides a deployment environment for iterating and improving robotic systems.

Risk Assessment

Korea's physical AI sector faces several material risks that warrant analytical attention.

Technology integration risk is the most immediate concern. Building a reliable humanoid robot requires simultaneous excellence in mechanical engineering, AI software, sensor systems, power management, and manufacturing. Integration of these diverse technologies into a commercially viable product has proven difficult globally; no company has yet achieved mass-market humanoid robot sales. Korea's distributed corporate structure, with capabilities spread across Hyundai, Samsung, Doosan, and LG, creates coordination challenges that the Humanoid Strategy Council must overcome.

AI software gap relative to US competitors represents a structural challenge. The most advanced AI models for robotic reasoning and physical interaction are currently developed by US companies (OpenAI, Google DeepMind, NVIDIA) that have deeper talent pools and larger training compute budgets. Korean robot hardware may initially depend on foreign AI models, creating a dependency that Mission 7 is designed to address but that may persist through the near term.

Cost competitiveness against Chinese manufacturers is a medium-term concern. Chinese humanoid robot companies benefit from lower labour costs, government subsidies, and proximity to component suppliers. If humanoid robots become commodity products competing primarily on price, Korean manufacturers may struggle against Chinese competition, much as they have in other hardware categories. Korean strategy must therefore emphasize performance differentiation, brand positioning, and software integration that command premium pricing.

Regulatory and liability frameworks for humanoid robots operating in workplaces and public spaces remain undeveloped. Questions of liability when a humanoid robot causes injury, data privacy when robots equipped with cameras and microphones operate in homes, and labour law implications of robot-for-human substitution all require policy frameworks that do not yet exist. The speed of regulatory development will affect deployment timelines.

Strategic Outlook

Korea's physical AI sector combines world-class hardware capabilities with a compelling demographic demand driver and strong government coordination through the Humanoid Strategy Council and K-Moonshot. The Hyundai-Boston Dynamics combination provides a technology anchor, Samsung-Rainbow Robotics delivers vertical integration potential, Doosan Robotics offers commercial cobot expertise, and LG's platform approach addresses service applications. The aggregate capability across these companies creates a national robotics ecosystem of genuine global significance.

The critical question is whether Korea can close the AI software gap quickly enough to deliver fully autonomous humanoid robots before US and Chinese competitors achieve mass-market production. The answer will depend on the success of Mission 7's physical AI model development, the talent pipeline expansion under Mission 10, and the semiconductor sector's ability to deliver competitive edge AI processors. Physical AI is where all of K-Moonshot's technology threads converge into a tangible, commercially relevant product. Its success or failure will be among the programme's most visible and consequential outcomes.