A few years back I worked shifts at Coupang's Bucheon 2 Fresh Center, and what people called "automation" back then was mostly conveyor belts and barcode scanners, not robots walking around picking up boxes. So when headlines now talk up humanoid robots on factory floors, I stay skeptical of the hardware. The money moving behind the scenes tells a different story, though.
Field Note: at a Seoul trade conference last month, a Samsung supply-chain manager mentioned that a chunk of their current robotics-related revenue has nothing to do with robots being sold — it's chips being sold for robots that barely exist commercially yet.
The compute stack comes before the robots do
A Barclays analysis covering 100-plus companies across the robotics value chain found that humanoid robots can drive real demand for data centers, AI accelerators, memory, and edge processors well before large-scale commercial rollout begins. Investors have focused mostly on the visible parts — motors, actuators, sensors. But the near-term bottleneck is intelligence: a robot has to perceive its surroundings and improvise, which needs three computing layers — data centers running simulation and generating synthetic training data, foundation models turning perception into action, and onboard processors executing decisions locally under strict power and latency limits.
Simulation matters because there's no internet-scale dataset for robot movement the way there is for language. Digital environments teach walking and grasping before real-world deployment, though physical testing still fills gaps simulation can't capture. Practically, computing demand can scale ahead of actual robot production — exactly the opening for Nvidia, AMD, Qualcomm, Micron, SK Hynix, Samsung, and TSMC.
Korea isn't waiting on the sidelines
Seoul just designated physical AI a national strategic industry inside a roughly $1 trillion package covering chip fabs, AI data centers, and robotics. The target: commercialize humanoids across 10 industries by 2028, growing Korea's global humanoid market share from about 1% today to 20%. Samsung is putting 60 trillion won into the Yeongnam region alone, and separately developing next-gen robot chips combining HBM, image sensors, foundry, and advanced packaging.
Hyundai is furthest along on actual hardware. Since acquiring Boston Dynamics in 2021, it's scaled Atlas production through Korea's supply chain, targeting 30,000 units a year by 2028, backed by a $5.8 billion buildout in Saemangeum. The government is also training 10,000 robotics specialists over five years and building a safety certification center in Daegu-Gyeongbuk, a lesson from the early 2010s collaborative-robot era, when weak certification standards slowed adoption for years.
Insider's Insight: what stands out compared to the US robotics scene is how tightly Korea plans the chip side and the robot side together from day one, rather than building robots and hoping the supply chain catches up later.
The market is small now — and that's the point
Current estimates put the humanoid robot market at roughly $2–3 billion, projected to reach $10–25 billion by 2030 and, optimistically, around $200 billion by 2035. Safety and autonomy problems remain unsolved, so full-scale deployment may land closer to 2035. Actuators account for 30–50% of a humanoid's component cost versus just 10–15% for onboard computing — a ratio that shifts as volume scales, which is why chipmakers are positioning years ahead of the robots themselves.
Behind the Scenes: SK Hynix already controls over half the global HBM market by revenue, sold out well into next year, mostly to Nvidia. Robot-specific compute demand on top of that means Korean memory makers aren't just riding the AI data center wave — they're positioning for a second one.
FAQ
Is Korea ahead in humanoid robots, or just in chips? Mainly chips. Korea's global humanoid market share is around 1%, but its HBM and memory position gives it leverage over whoever builds the robots.
Why does simulation matter for training robots? There's no internet-scale movement dataset like there is for text, so digital environments teach walking and grasping before real deployment.
When will humanoid robots show up in Korean factories? The government targets 2028 across 10 industries, with Hyundai aiming for 30,000 Atlas units a year by then.
Which Korean companies benefit most? Samsung and SK Hynix for memory and HBM, plus foundry and packaging suppliers tied to Samsung's robot chip roadmap.
Why is compute a bigger near-term opportunity than actuators? Compute infrastructure has to build out before robot production scales, so demand shows up earlier in the supply chain than actuator manufacturing does.
Explore More:
- Made in Korea: Tech Dominance in Semiconductors, EVs, and Defense
- How Samsung Became a Global Tech Empire
- Hyundai-Kia: How Korea Conquered the Auto Industry
- Why Naver and Kakao Dominate Korea's IT Ecosystem
#KTech #KoreanSemiconductors #Samsung #SKHynix #HumanoidRobots #PhysicalAI #HyundaiRobotics #KoreaTech2026