Gallop: ASUS’s new humanoid robot learns to pack tiny motherboard parts in factory

Taiwanese technology company ASUS is developing a humanoid robot called Gallop to automate repetitive tasks in electronics manufacturing. The robot’s customer pilot programs are planned for 2027, and commercial availability is targeted for the second half of 2028.

The robot is currently being tested at an ASUS group electronics manufacturing services (EMS) factory, where it is learning to pack motherboard accessories at a workstation designed for human workers.

According to ASUS, Gallop is being developed to show that humanoid robots can consistently perform real factory operations rather than simply complete staged demonstrations.

Humanoid learns factory

Gallop’s initial assignment involves packing 15 different motherboard accessories, many of which are small, thin, and enclosed in transparent plastic bags. The task presents a challenge for conventional robotic grippers because of the different shapes and flexible materials involved.

ASUS said it selected the workstation to test whether Gallop could take responsibility for most of an operation without requiring a human worker to remain alongside it throughout the process.

Meet Gallop at work. Its first job: packing motherboard accessories, 15 small parts, many in soft bags. It learns by watching our own workers, then practicing on the same parts. Albert Chang explains how:https://t.co/8gtKPDuey4 #ASUS #AI #PhysicalAI pic.twitter.com/CF3iCBmblh

— S.Y. Hsu (@SYHsuASUS) November 13, 2026

The company has developed a training process that combines human demonstrations, simulation, and hands-on practice. With their consent, experienced factory workers wear small head-mounted cameras to record how they perform specific operations. Gallop uses this information to learn the movements and handling techniques required for the job. Simulation software then generates variations of the demonstrations to help the robot handle minor differences in working conditions.

The training process also includes virtual reality (VR), with an operator wearing a headset and gloves to guide Gallop through tasks using the same components and cameras found at the actual workstation. Once deployed, the robot’s recorded experiences will also feed back into its training process.

The company has set a target of achieving 99.5 percent task success within 20 seconds, matching the time required for a human worker to complete the same operation.

Gallop gets smarter

ASUS has divided Gallop’s AI architecture into four layers: planning, supervision, skills, and control. The planning layer interprets the surrounding environment and breaks an operation into individual steps. A separate supervisory AI monitors these actions and can revise the plan when something goes wrong. The skills layer manages movements such as reaching, grasping, and placing objects, while the control layer operates the robot’s arms and hands.

The architecture uses fixed interfaces between the different layers, allowing ASUS to replace AI models as improved versions become available without redesigning the entire robot. The system runs on a single NVIDIA Jetson AGX Thor computer installed inside Gallop, alongside standard robotics software.

ASUS is developing the robot’s body and dexterous hands in-house. The hands are particularly important because they need to manipulate small components while using tactile sensing to understand physical contact. Other components, including motors, are sourced from established industry partners.

Gallop currently operates from a fixed base equipped with two arms. ASUS plans to introduce a wheeled version and integrate its in-house dexterous hand over the coming year. The company also intends to demonstrate the robot to customers at Computex 2027, ahead of planned pilot deployments.

Electronics manufacturing will be its first target market, although ASUS sees potential applications in hospitals and other service environments. The company believes that skills learned at one workstation could be transferred to other operations involving similar pick-and-place activities.

ASUS said Gallop is intended to handle repetitive and physically demanding tasks rather than replace factory workers. Employees involved in training the robot could instead move into roles involving robot setup, supervision, and performance improvements.

The Gallop program began in early 2026. ASUS is targeting customer pilots in 2027 and market availability in the second half of 2028. However, the robot’s ability to meet its factory performance targets remains an important part of its development.