Samsung's answer to the most persistent problem in humanoid robotics is a film thinner than a fingernail. At its C-Lab Startup Demo Day on November 10 at the Seoul R&D Campus in Seocho-gu, the company demonstrated a 0.5mm thermoplastic polyurethane electronic skin wrapped around a humanoid robot hand that can detect exactly how hard it's gripping an object in real time. The demonstration did not announce a finished robot. It announced that the grip problem, the one unsolved variable between a powerful robotic hand and a useful one, may now have a commercially viable answer.
How the 0.5mm Film Solves the Problem Every Robot Hand Still Has
A humanoid robot hand can generate enormous force. That is the easy part. The hard part is knowing when to stop. The result is what engineers call the "robot hand dilemma": powerful actuators that crush what they should handle. A single egg or a ripe tomato becomes a live test of whether the robot can sense its own grip force and adjust before it destroys the object. For years, the solutions have been either too expensive, too fragile, or too thick to fit in tight assembly spaces. Samsung's demonstration at the C-Lab event represents the first time the company has shown a commercially oriented path around all three constraints at once.
The film is made of thermoplastic polyurethane, a rubber-like polymer that stretches continuously as the hand moves without losing sensing accuracy across the deformed surface. Pressure sensors are distributed across the film, registering not just whether contact is occurring but where on the hand it is happening and how much force is being applied at each point. The attachment method adapts to the geometry. When heat is applied, the film conforms precisely to individual finger contours; it can also wrap over the fingers as a full glove layer or be secured to the palm surface alone. When the sensor data reaches the motor controller and the signal crosses a pre-set threshold, the motors stop tightening. That feedback loop is what converts a powerful robotic hand into one that can handle fragile objects at the speed industrial production requires.
The skin is supplied by Midas H&T, a Korean startup that Samsung has brought into its C-Lab Outside program, the company's external startup incubation channel. CEO Chang Se-yoon confirmed the testing at the demo day and said the company is "building the supply chain needed to manufacture electronic skin so that we can begin production next year," according to Seoul Economic Daily. At 0.5mm, the film is thin enough to allow precision assembly in the tight workspaces characteristic of electronics and semiconductor manufacturing. Chang said the attachment method varies by application, suggesting that the same film technology could be tuned for different force-sensing profiles across different industrial tasks.
The testing falls under Samsung's RX Business Development Office, a robotics unit created in July 2026 that reports directly to the CEO rather than any single product division. CTO Yoon Jang-hyun of the DX division was appointed in September to lead both the hardware development team and the AI software development team for the robotics effort. The organizational structure signals that Samsung's leadership treats humanoid robotics as a platform-level initiative rather than a product-line extension, one that sits above the TV or smartphone business in terms of strategic attention. And according to Newsworks, the company is targeting deployment at industrial sites as early as 2027, with a broader rollout by 2028 at the latest.
The Tendon-Inspired Hand the Skin Is Designed to Wrap
The electronic skin is not the only hand technology Samsung has been developing. In October 2026, the U.S. Patent and Trademark Office published Samsung patent US20260326284A1 (via Parola Analytics), filed in May, for a robotic hand that mimics human tendon mechanics to coordinate multiple finger joints without requiring a separate motor at every joint. The two technologies were clearly designed in parallel.
The patent design uses active pulleys driven by motors to curl and extend each finger through two cables: one on the palm side that bends the finger, one on the back that extends it. Two categories of passive pulleys then link the joints mechanically, reproducing the effect of biological ligaments. An Oblique Retinacular Ligament cable coordinates the distal and middle finger joints; a Lumbrical Muscle cable links the middle and knuckle joints. A torsion spring on the lumbrical cable returns the fingers to a neutral position after external force moves them. The design keeps the hand compact by moving the drive motors into the palm module rather than distributing them across the finger segments.
The practical result is a hand that can adapt its grip pattern to an object's shape through cable tension rather than a separate control signal at every joint. The Newsworks report notes that Samsung has combined this cable-driven mechanism with a camera system that changes finger arrangement based on what the hand recognizes. The camera sees the shape; the cable system adapts the grip; the electronic skin monitors the actual contact force. The three layers together create a sensory-motor loop that is closer to how a human hand handles an unknown object than anything a purely vision-based system can achieve on its own. Samsung registered the foundational U.S. patent for the hand mechanism in April 2026, and the hand uses Rainbow Robotics actuators for the arm interface.
Robotics Lab head Chris Hauser acknowledged at a recent briefing that Samsung's current capabilities are behind the leading competitors. The admission is unusual for Samsung, whose public posture is typically about capability benchmarks rather than gaps. The timing suggests that the company is managing expectations before a possible CES 2027 debut, while simultaneously making the case that it has a credible technology foundation to close that gap.
Samsung enters a market where the leading robots already have tactile sensing. Figure 03 ships with 3g tactile fingertips and embeds tactile sensors and cameras inside the Helix 02 palm. Tesla Optimus Gen 2 has tactile sensing across all fingers and 22 degrees of freedom per hand, with its Gen 3 design reportedly pushing to 50 total actuators. Boston Dynamics Atlas carries 360-degree cameras paired with force feedback. XELA Robotics sells commercial 3D tactile sensor modules that fit standard humanoid grippers and had its first commercial orders available in early 2026 for the Tesollo DG-5F hand. These are production systems, not demos.
The competitive comparison looks different once you include Samsung's customer base, however. Humanoids Daily reported that Oh Jun-ho, head of Samsung's Future Robot Promotion Team, described the company's strategy at the International Robot Symposium as being both "a technology provider and a customer." That phrasing names a structural advantage that is easy to miss when comparing sensor specs. Samsung has semiconductor fabs and electronics assembly lines where the precision-grip problem the electronic skin solves is a live production constraint today. Figure's paying external customer is BMW Spartanburg. Tesla Optimus, as of early 2026, was still described by its CEO as primarily a learning platform, not a productive one. Samsung would be both the robot supplier and the first deployment operator, a feedback loop that Figure and Tesla cannot replicate without entering a customer's factory on someone else's terms.
South Korea's national K-Humanoid Alliance gives Samsung additional leverage. The alliance coordinates robotics R&D and manufacturing across the country, and Samsung's scale in semiconductor components, displays, and battery cells makes it a natural anchor company for the supply chain the alliance is developing. IDC reported approximately 25,000 humanoid shipments globally in H1 2026, up 432.1% year over year. At that growth rate, the company that solves the grip precision problem at scale and can supply its own factories with the first production units will be positioned to capture industrial orders as humanoid robots transition from pilot programs to full factory integration. The question is not whether Samsung has the best humanoid today. It is whether its integrated position gives it a faster path to the first sustained commercial deployment.
The Bear Case Against Samsung's Humanoid Bet
The bear case, however, is straightforward: Samsung is at least two generations behind the market leaders in demonstrated full-system humanoid capability, and it is entering a category where the gap between component demo and commercial deployment has historically taken three to five years to close. Figure 03 began paid work at BMW in early 2026 after two years of parallel R&D and customer negotiation. Tesla's Optimus production line at Fremont is running but has not yet announced a single external customer contract with published production volumes. Samsung's target of industrial sites "as early as next year" follows a CES debut that hasn't been confirmed.
Critics argue that the C-Lab Demo Day announcement reflects readiness to show the skin technology rather than readiness to deploy the robot system. The 2027 production timeline stated by Midas H&T covers the film alone, not the integrated humanoid system that the film must work inside. Samsung's hand design remains a patent, not a production part. And while the "provider and customer" framing is structurally compelling, deploying experimental robots inside its own semiconductor fabs introduces production risk that is difficult to quantify. A single incident in a cleanroom environment can interrupt manufacturing processes worth far more per hour than the robot is saving.
LG Electronics, Doosan Robotics, and Rainbow Robotics are all building humanoid capabilities inside South Korea. Sammy Fans reported that Samsung's go/no-go decision for a CES 2027 unveiling was set to be finalized in November, and that the standard for going ahead was "minimum readiness" rather than a finished product. That is a low bar for a company whose brand expectations are set by its display and semiconductor divisions. If the robot that appears at CES can only demonstrate simple object lifting, the contrast with Atlas and Figure 03 will be immediate and public, and it will shape how Samsung's humanoid program is perceived for the following twelve months.
The November Decision and the Signals That Will Define 2027
Samsung's go/no-go decision on CES sets a compressed timeline. If the company proceeds with the January 2027 show, observers will be watching whether its bipedal humanoid can demonstrate the skin's grip control under realistic conditions, handling varied objects at speed rather than moving a single pre-positioned item in a controlled demo. The skin's commercial story depends entirely on the hand it wraps, and the hand's commercial story depends on the robot platform it attaches to.
eWeek documented that Chinese researchers at City University of Hong Kong have developed a neuromorphic electronic skin that converts touch inputs into nerve-like pulse trains, triggering motor reflexes when force exceeds a threshold without routing the signal through a central processor first. That approach is a generation beyond pressure detection and sampling, and Samsung will need to show a credible multi-year roadmap to match it, not just 0.5mm sensing but sub-millisecond neural-style response, if it intends to compete in the precision manufacturing segment against Chinese suppliers who are iterating at speed.
Three signals will tell the story by mid-2027. First, whether Midas H&T discloses a production volume target, not just a 2027 start date. The difference between a supplier that can ship 100 units and one that can ship 10,000 is the difference between a demo program and a commercial product. Second, whether Samsung places a named factory-line deployment at one of its own facilities, specifically in the DS Division's semiconductor fabs, where the precision requirements are most demanding and the documentation of success would be most credible. Third, whether the C-Lab relationship with Midas H&T deepens into an acquisition, the most aggressive signal that Samsung views electronic skin as a strategic asset and not a component it sources from outside.
Samsung is not entering the humanoid market with a robot. It is entering with a supply chain, a customer base, and a 0.5mm film that solves the problem every other competitor is still routing around software.
Key Takeaways
- 0.5mm TPU film from Midas H&T, demonstrated at Samsung C-Lab Demo Day on November 10, detects contact pressure and controls grip force on humanoid robot hands, with production targeted for 2027
- 432.1% growth in global humanoid shipments in H1 2026 (IDC, approximately 25,000 units) puts grip precision at the center of mass-deployment readiness for the next phase
- Dual provider-and-customer strategy means Samsung's own semiconductor fabs and assembly lines become the first deployment customers, a feedback loop Figure and Tesla cannot replicate from outside
- Patent US20260326284A1, filed May 2026 and published October 22, describes a tendon-inspired cable-driven hand coordinating three joints per finger with active and passive pulleys and a torsion spring return mechanism
- November go/no-go decision on a CES 2027 humanoid unveiling is imminent, with Samsung's stated threshold set at minimum readiness rather than a finished product
Questions Worth Asking
- If Samsung's first humanoid deployment is inside its own semiconductor cleanrooms, how does it quantify the production risk of a robot incident versus the upside of reduced labor costs at current robot maturity levels?
- Chinese neuromorphic e-skin already converts touch into nerve-like reflexes without central processing latency. At what point does Samsung's pressure-film approach become a generation behind before it reaches scale production?
- The K-Humanoid Alliance coordinates Korean robotics manufacturing at a national level. Does Samsung's participation accelerate its timeline or slow it by aligning with a government pace rather than a market pace?
