インフィニオンの人型ロボット新興企業コンテスト、伊AdapTronicsが優勝　韓国COSMOKEが2位

The challenge shows where a component supplier thinks humanoid value will sit: in grasping, touch and joint control, the subsystems that consume sensors, microcontrollers and power stages in volume. Infineon gave finalists its own sensing, artificial-skin and motor-control kits, so every winning design is already built around Infineon parts. That is a cheap way to make its chips the default inside start-up hardware before robot makers freeze their bills of materials.

For robot makers the signal is that the hand and skin layer is still open. None of the three winners is a robot company; each sells a subsystem, which is the layer where Korean, Chinese and European start-ups can still compete without building a whole humanoid.

COSMOKE's magnetic tactile skin competes with optical and visuo-tactile approaches such as Jodell's sensors launched at CIIF on Oct 16 and Rhoda AI's in-house tactile sensor described on Oct 13, both reported by MEKAWIRE. Magnetic skins are thin and flexible, which suits fingers and palms; optical sensors give richer contact images but are bulkier.

AdapTronics' grasping approach competes with suction and mechanical grippers from established automation suppliers, and with dexterous hands from Tesollo, Wonik Robotics and Chinese makers. The release does not say how its method works physically, so a technical comparison is not yet possible.

No valuation, funding round or prize amount was disclosed for any of the three start-ups. For scale, MEKAWIRE's verified robot-touch records run from Tangent Robotics' $4.5 million pre-seed to Rhoda AI's $450 million Series A, and an Infineon co-innovation slot is best read as access to engineering support and a supply relationship rather than capital.

The kits Infineon supplied (environmental sensing, artificial skin, motor control, visual feedback) correspond to its magnetic sensors, microcontrollers and power semiconductors. A tactile skin such as COSKIN needs many magnetic sensing points per hand, and in-memory motor control sits next to the gate drivers and power stages in every joint. If any winner reaches production, the bill of materials would be Infineon-heavy by design.

1) The Startup Night in Munich in November, where AdapTronics' co-innovation work formally starts. 2) Any second co-innovation agreement with COSMOKE or TECHiFAB, which Infineon said it is discussing. 3) A robot-hand or humanoid customer naming COSMOKE's skin or AdapTronics' gripper.

Thesis: Infineon is buying optionality on the humanoid hand and joint at very low cost. Confidence: medium, because the release gives no money figures and none of the winners has a disclosed customer. What would change our view: a disclosed investment or a production supply agreement, which would make this a supply-chain move rather than a marketing programme.

How many tactile sensing points does COSKIN use per square centimetre, and which Infineon sensor family does it rely on? Will Infineon take equity in AdapTronics through the co-innovation programme? Which robot makers sat on the jury or have tested the finalists' designs?