ASUS’s first humanoid robot must pack 15 tiny parts at human speed

ASUS, the company better known for laptops and gaming motherboards, has put its first humanoid robot to work on a real factory line. The robot, called Gallop, is learning to pack the accessory bundle that ships with a motherboard: fifteen small parts, many of them sealed in thin, clear plastic bags, at a station built for human hands. ASUS says the point is not a staged demo but a machine that can hold down that station for a full shift.

That is the gap most humanoid robots have not crossed. Viral clips show them dancing, sorting blocks or folding laundry, but a factory manager wants to know whether one can do a specific job every day to a quality standard. ASUS is betting that a robot with two arms and dexterous hands can slot into stations designed for people, so a plant does not have to be rebuilt around automation.

The bar is concrete. A human worker finishes the packing job in about 20 seconds per board, and ASUS wants Gallop to get it right 99.5% of the time within that same window. Soft, thin, see-through bags are exactly what a simple gripper fumbles, which is why the company picked this station. “If a person still has to stand beside it the whole time, nobody gains anything,” Albert Chang wrote in a post on the ASUS pressroom explaining the project.

Gallop learns the job the way a new hire would, by watching. Experienced workers wear small head-mounted cameras, with their consent, so the robot can see how the task is really done. Simulation software then generates variations of each demonstration, so a shifted tray or a crumpled bag does not throw it off. An operator in a VR headset and gloves also steers the robot through the task with the same parts and cameras as the real station, and once Gallop is deployed, what it records on the line feeds back into its training. ASUS does not train its own giant AI model; it starts from open models that already understand how hands and objects behave.

Inside, the software is split into four layers. A planner reads the scene and breaks the job into steps, a second AI supervises every step and re-plans when something goes wrong, a skills layer handles reaching, grasping and placing, and a control layer drives the arms and hands and stops them the instant anything is off. Because the layers talk through fixed interfaces, ASUS says it can swap in a better model every few months without rebuilding the robot. All of it runs on a single Nvidia Jetson AGX Thor computer inside the machine.

The hand is the part ASUS decided to make itself. It designs Gallop’s body and its touch-sensing dexterous hand in-house and buys motors and other standard components from suppliers. That is where the whole industry is stuck. JP Morgan analysts named hands, actuators, the robot’s brain and production capacity as the binding bottlenecks in a humanoid forecast published this week, and Tesla has pointed to the hand as the main reason its Optimus robot will not reach outside buyers until late 2027.

For now, Gallop is less of a humanoid than the name suggests. It works from a fixed base, with two arms and no legs or wheels, and the in-house hand is not fitted yet. ASUS has shared a target, not a measured success rate, and it has said nothing about price. The first job is also inside a factory ASUS controls, with its own parts and its own quality rules. The company presents that as an advantage most robotics start-ups have to rent, but it also means no outside customer has put the robot through its paces. Interesting Engineering, which reported on the project, noted that hitting those factory targets is still part of the development work.

ASUS insists Gallop is meant for repetitive, physically demanding stations that are hard to staff, not to replace the people on them, and says the workers who teach it will move into setting up, supervising and improving the robots. Co-CEO S.Y. Hsu introduced the robot on CNBC before the company published the detailed explanation.

The Gallop program began in early 2026. Over the next year ASUS plans to fit its own hand, put the robot on wheels and show it to customers at Computex 2027 in Taipei. Customer pilots are planned for 2027 and commercial availability is targeted for the second half of 2028, starting in electronics manufacturing before moving into hospitals and service sites. ASUS has not said which countries or customers will get it first.

Until then, the yardstick ASUS has chosen is the same stopwatch it uses on its own staff: about 20 seconds a board, and almost no mistakes.

Tags: nvidia, robotics, humanoid robots, factory automation, ASUS, Gallop