ASUS skipped the dance video. Its first humanoid, Gallop, debuts on a motherboard packing line

Most humanoid debuts follow a script: a stage, a soundtrack, a robot doing something photogenic. ASUS went the other way. The Taiwanese company best known for motherboards and laptops introduced Gallop, its first humanoid robot, by pointing at one unglamorous station in one of its own electronics plants, the place where the small accessories that ship in a motherboard box get packed. Co-CEO S.Y. Hsu presented the robot on CNBC on November 13, and ASUS followed with a long post explaining how it is being built.

The job, in numbers

The station Gallop is learning has 15 small parts, many sealed in thin, clear plastic bags, and a board comes through about every 20 seconds. ASUS has set a target of 99.5% success inside that window, which it calls “the same standard we hold a person to,” in a post by ASUS’s Dr. Albert Chang. The company says the full loop now works on the line: a person demonstrates the task, the robot learns it, and it gets tested on real parts. What it hasn’t shared is how close Gallop is to that 99.5% today.

The choice of task is the interesting part. Bagged, floppy, translucent parts are exactly what simple grippers and fixed automation struggle with. Many of the accessories “are very small and put in a plastic bag, so it needs some special training to handle those kinds of parts well,” said ASUS co-CEO S.Y. Hsu. ASUS argues that a robot with two arms and dexterous hands can step into stations already designed for people, which matters most on high-mix lines where the setup changes from run to run.

Not quite the humanoid you’re picturing

Gallop today is two arms on a fixed base, not a walking robot. A wheeled version is planned next. ASUS designs the body and its own dexterous hand, which it says has a sense of touch, and calls the hand the hardest part of any humanoid. Motors and other standard parts come from suppliers it hasn’t named. There is no published height, weight, payload, battery life or price yet.

How it learns

ASUS isn’t training a giant model of its own. Gallop starts from open AI models (ASUS hasn’t said which) and is then taught each job in stages: experienced workers wear small head cameras, with their consent, while they do the task; simulation generates variations of each example; an operator in a VR headset and gloves guides the robot through the work on the real parts; and once deployed, what the robot records feeds back into training.

Inside, the software is stacked in four layers. A planner reads the scene and breaks the job into steps, a second AI acts as a supervisor that checks each step and re-plans when something goes wrong, a skills layer handles reaching, grasping and placing, and a control layer moves the arms and stops the robot if anything is off. The layers connect through fixed interfaces, so ASUS can swap in a newer model without rebuilding the machine. All of it runs on a single NVIDIA Jetson AGX Thor computer.

Where Gallop fits

Jetson Thor keeps turning up as the default brain box. Meta’s Max prototype, built with Fourier, runs on it too, and UK robot maker Humanoid moved to a single Thor in October. What sets ASUS apart is less the computer than the factory. Most robot makers have to borrow a customer’s line to prove themselves. Apptronik needed Mercedes-Benz to show Apollo 2 sorting brackets on live lines, and Agility Robotics proved Digit in GXO’s warehouses. ASUS owns its test floor, with real parts and real quality rules it would otherwise have to rent.

The flip side is that ASUS is starting later than most and is narrow by design. Gallop’s first market is electronics manufacturing, the industry ASUS knows best, with hospitals and service sites named as a longer-term goal because many of those jobs share the same pick-and-place skills.

The timeline

The program only started in early 2026. Over the next year ASUS plans to put its own hand on the robot, add wheels and show Gallop to customers at Computex 2027 in Taipei. Hsu said the robot will go to partners in the second half of 2027, and commercial availability is targeted for the second half of 2028. ASUS also says Gallop is meant for repetitive, physically demanding stations that are hard to staff, with the workers who teach it moving into setup and supervision roles.

Our take

A single station and a published quality bar is a more honest debut than most. The number to wait for is the measured success rate against that 99.5% target, because that is what tells a factory manager whether Gallop can carry a shift. Until then, compare it with the robots already working in plants in our humanoids guide, follow NVIDIA for more Jetson Thor robots, and browse factory robotics roles on our jobs board.