Samsung RX tests Midas H&T electronic skin on humanoid robot hands for grip force control

Humanoid robots still stumble on the same hand problem biology solved in childhood: grip hard enough to move the load, soft enough not to crush it. On November 10, Korean startup Midas H&T told reporters at Samsung Electronics' 2026 C-Lab Startup Demo Day in Seoul that Samsung is testing the company's electronic skin on the in-house humanoid hand RX is developing. DigiTimes followed on November 12. Seoul Economic Daily published English detail on November 11. For buyers, sellers, and operators on Buy My Robots, the story is tactile infrastructure: whether the hand on a listing can prove closed-loop force control, and who owns spare skin films after resale.

What Samsung and Midas H&T said at C-Lab Demo Day

The demo day ran at Samsung's Umyeon R&D campus in Seocho, the same Seoul complex where RX has staged prior humanoid work. Midas H&T participates through C-Lab Outside, Samsung's channel for external startups. Verification is happening with Samsung's RX Business Development Office, the robotics unit that reports into the CEO tier.

Midas H&T's film is about half a millimeter thick, built on stretchable thermoplastic polyurethane with distributed pressure sensing. Seoul Economic Daily describes heat application so the layer conforms to finger geometry, similar to applying protective film on a curved surface. CEO Chang Se-yoon said configurations vary: wrap along fingers, add palm coverage, or layer over existing camera and depth sensors so vision fixes the target while skin reports contact pressure.

The tomato example both Korean and English coverage repeat is deliberately mundane. Strong industrial grippers can lift housings, but they often lack the gradation needed for produce, fragile connectors, or uneven cartons. Midas H&T argues electronic skin closes the loop: once finger pressure crosses a threshold, motor commands stop tightening. Chang also said the thin profile targets tight assembly cells where bulkier tactile stacks do not fit.

Commercial timing remains forward looking. Chang told outlets that Midas H&T is building a supply chain to begin production in 2027. Aju Press on November 10 quotes CEO Steve Chang calling the work proof of concept while Samsung's Future Robotics Office tests the sensors on robotic hands. ZDNet Korea adds that Midas H&T positions the silver-and-rubber sensor as far thinner and cheaper than many conventional tactile modules, with agricultural and fishery handling among the named use cases. None of the releases publish Samsung serial counts, task success rates, or pricing.

Why hands matter more than another walking reel

RX's broader plan still points toward affiliate factory work first and public product talk closer to CES 2027. Hands are where that plan meets physics. Boston Dynamics' October four-finger Atlas hand story in our archive emphasized dense fingertip and palm sensing before drill triggers and in-hand rotation. Figure's Figure 03 materials cite palm cameras and tactile sensing. Tesla Optimus messaging still leans on hand yield and endurance as production bottlenecks.

Samsung's November disclosure does not ship a glove you can bolt onto a used Unitree or Digit. It does signal that Samsung's captive humanoid path treats distributed touch as a gating item, not an optional demo upgrade. Midas H&T also said it is working with the Korea Institute of Machinery and Materials on full-body electronic skin, aligned with talk of a standardized Korean humanoid platform. Marketplace users should read that as ecosystem building around RX bodies, not as an open accessory catalog yet.

Secondary effects show up fast on listings. End effectors already dominate repair bills on humanoid fleets. A conformal skin film adds calibration drift, cut damage, and vendor-specific replacement rolls. If Samsung humanoids eventually list alongside Rainbow Robotics platforms or pilot surplus from affiliate plants, buyers need to know whether tactile layers are field replaceable or factory recalibration only.

How this differs from our August Samsung RX demo post

Our closest archive piece is the August 2026 Samsung RX closed-door humanoid demo. That story was executive proof that an in-house body could operate a camera, signaling finger dexterity and Hand Lab tendon research without naming a tactile supplier or production schedule.

Today's news is a supplier-linked hand stack at a public startup showcase. The central claim is force-aware gripping through conformal electronic skin, with Midas H&T targeting 2027 manufacturing and citing layering with vision sensors. Same company. Different evidence package: from "can it manipulate a camera" to "can it feel how hard it squeezes a tomato."

Checklist before you list or buy a tactile-heavy humanoid hand

If a seller cites Samsung, Midas H&T, or generic "electronic skin" on a humanoid listing, slow down and ask for documents tied to calibration and wear, not press-day photos only:

Which hand revision ships, and whether tactile film is Samsung captive, Midas H&T, or a third-party retrofit.
Logged calibration dates, pressure thresholds used on real tasks, and who performed the last service.
Whether skin replacement kits transfer with the robot, or stay tied to an OEM service account.
Task proof on delicate picks, not only video of firm grasps on rigid totes.
How vision and tactile fusion is versioned in software, and whether policy updates survive resale.

Until Samsung publishes a commercial SKU and Midas H&T ships production film at scale, treat C-Lab reporting as credible direction for RX humanoids, not as proof that every Samsung-linked biped in a warehouse already wears the glove.

What to watch next

Watch whether RX shows the skinned hand on a full humanoid walk sequence, not isolated prosthetic hands at demo tables. Watch if Midas H&T or Korean institutes publish independent benchmarks against rigid fingertip sensors on Figure, Atlas, or Chinese five-finger packages. Watch whether secondary listings start listing tactile layer part numbers the way they already list battery cycles and joint hours.

Humanoid economics still start at uptime per shift. They increasingly end at whether the hand knows when to stop squeezing. For a humanoid marketplace, that distinction belongs in the diligence packet before price, not in the listing title alone.