Tesla Wants to Give Optimus a Skin That Can Feel

The first drawing in Tesla’s patent application shows tactile sensors on a robot hand’s thumb (102A), palm (102B) and index finger (102C). Credit: Tesla patent US20260310299A1 via SETI Park/X

Tesla has applied for a U.S. patent on soft touch sensors that wrap around a robot hand’s thumb, palm and fingers, in an application the U.S. Patent and Trademark Office published Thursday and Seoul-based patent researcher SETI Park flagged on X.

The filing is titled “Three-Dimensional Soft Compliant Array Tactile Sensor with Multi-Modal Sensing and Scalable Manufacturing Methods” (Pub. No. US 2026/0310299 A1). Tesla filed it on September 11, 2025, building on a provisional application from April 4, 2025, and it names 12 inventors based around the San Francisco Bay Area.

According to the abstract, the sensors “can conform to curved geometries and can be used for measuring surface force, touch, and pressure.”

The application’s first drawing, shown above, is a five-fingered robot hand carrying three sensors. One runs along the thumb, a second covers most of the palm and a third sits on the index finger. The filing never mentions Optimus by name, but it says the sensors can go on “fingertips, palms, end-effectors, or other portions of a robot” and can be shaped to act as a tactile “skin” for a humanoid robot.

SETI Park, who tracks Tesla filings, posted the drawings early Thursday and asked whether the technology is headed for Optimus Gen 3.

Tesla’s application frames the problem in plain terms: conventional touch sensors are rigid, flat and two-dimensional, so they can’t follow the curve of a fingertip. The soft sensors that do bend, it says, lack the mechanical stability a working robot needs. Tesla’s answer is a sensor that’s pliable enough to hug a curved part and sturdy enough to survive the job.

The design works like a sandwich, with two thin, flexible plastic sheets, each carrying a grid of printed conductive traces, and a separator layer between them. Wherever a trace on the top sheet crosses one on the bottom sheet, it forms a sensing point. The filing calls these “pixels,” and each one can detect contact on its own, so the hand can tell where it’s being touched as well as how hard.

Pressing a pixel squeezes the middle layer. If that layer is an insulating material, the traces move closer together and the capacitance between them changes. If it’s a piezoresistive material, such as carbon-loaded silicone, a conductive elastomer or conductive foam, its electrical resistance drops instead. Either way, readout circuitry wired to the traces picks up the change.

The sheets can be made from thermoplastic polyurethane (TPU), PET or polyimide, and the traces from silver, carbon or copper-based inks. The finished sensor can be wired up through a flexible circuit board, mounted on either a hard plastic or metal base or a soft foam or silicone backing, and covered with a protective layer of foam, TPU or silicone.

The title promises “multi-modal” sensing, but the text only describes those capacitive and resistive versions measuring force, touch and pressure. It doesn’t describe separate sensing for shear, slip, temperature or proximity.

In a follow-up thread, SETI Park summed up the stack as substrate 1, conductive layer 1, separator, conductive layer 2 and substrate 2. He also noted that the application’s claims lay out two ways of building it.

The first route, covered by claim 1 along with claims 3 and 4, starts flat. Conductive ink is screen-printed through a stencil onto flat sheets, and the sheets are stacked around the separator. The stack is then heated past the point where the plastic softens and pressed into a mold shaped like the finger or palm, a process called thermoforming. Once it cools, it holds the curve. The application also describes a variation where a laser cuts channels into the sheet and the channels are filled with conductive paste.

The second route, covered by claims 10 and 12, flips the order. The base is molded into its final 3D shape first, and a computer-guided dispenser draws the conductive traces directly onto the curved surface, much like a 3D printer. The filing also lists laser activation followed by metal plating, and pad printing, where a soft silicone pad picks up an inked pattern and presses it onto the curved part. The separator can be dispensed on top the same way before the second layer goes on.

SETI Park also turned up an open Tesla job listing that lines up with the filing. The Manufacturing Technology Development Engineer, Optimus Hand role in Palo Alto describes the hand as “a high-dexterity, tendon-driven mechanism with advanced tactile sensing.”

The person hired would “lead process development for advanced tactile sensing hardware, with a focus on thermoforming, dispensing, and lamination processes,” and the listing calls “TPU or similar thermoplastics experience highly relevant.” Those are the same processes and one of the same materials named in the patent application. SETI Park thinks that points to Tesla weighing TPU for the sensors. The listing says the job is about hitting the cost, quality and throughput targets “required for Optimus to ship in volume.”

Touch has been part of the Optimus pitch for a while. When Tesla unveiled Optimus Gen 2 in December 2023, it showed new 11-degree-of-freedom hands with tactile sensing in all fingers, and the demo had the robot passing an egg from one hand to the other. An international Tesla patent published in April detailed a cable-driven hand with 22 degrees of freedom, double the previous design.

Elon Musk has called the hand the hardest part of the robot. “The forearm and hand are more difficult than the entire rest of the robot,” he said on Tesla’s Q3 2025 earnings call. Optimus 3, the first version built for volume production, still hasn’t had its public reveal. Musk pushed that reveal closer to the start of production to keep competitors from copying the design, and Tesla’s own app files already label the robot “Gen 3”.

This is a patent application, not a granted patent, and companies file plenty of designs that never ship. Tesla hasn’t said whether these sensors are in Optimus 3, how many pixels a hand would carry, which of the two manufacturing routes it prefers or when any of it would reach production. The filing also gives no performance figures, such as how sensitive each pixel is or how long the sensors last.

The timing is what makes this one worth watching, since Tesla’s provisional filing goes back to April 2025 and the company is now hiring people to take thermoformed, dispensed touch sensors from prototype to production ramp. That suggests the robot’s sense of touch has moved off the whiteboard and onto the factory planning list.