Fifteen tiny parts. Many of them sealed in floppy plastic bags. About 20 seconds to get them all into a motherboard box. That’s the first job ASUS has handed to its brand-new robot, the ASUS Gallop humanoid.
You probably know ASUS for motherboards and electronics, not robots. But on November 13, the company showed off Gallop working on one of its own factory lines. And the most interesting part isn’t the robot’s body. It’s how Gallop learns.
So what did ASUS actually build?
Gallop is the first humanoid from ASUS. Its Humanoid Robotics unit built it, inside a business group focused on physical AI. In other words, AI that moves around and touches things in the real world, instead of living in a chat window. The program only started in early 2026, so this is fast work.
Right now, Gallop is fairly simple. It has two arms on a fixed base, so it can’t walk yet. A single NVIDIA Jetson AGX Thor, a compact computer made for robots, runs the whole thing. There’s also a camera on its wrist, which helps it see what its hand is doing.
ASUS designs the body and the hand in-house, while partners supply the motors. The hand has a sense of touch, and ASUS calls it the hardest part of any humanoid. Honestly, that tracks. Plenty of robots can lift a box. Far fewer can pick up a thin, slippery bag without fumbling it.
Co-CEO S.Y. Hsu put it simply in a post shared on launch day: “Meet Gallop at work. Its first job: packing motherboard accessories, 15 small parts, many in soft bags.”
How the ASUS Gallop humanoid learns its job
Here’s the clever bit. Gallop learns from the people who already do the job.

Experienced workers wear small head cameras while they pack, with their consent. Then ASUS uses simulation to create lots of slightly different versions of each recording, so the robot can cope with small changes on the line. Next, an operator wearing a VR headset and gloves steers the robot through the real task. That’s called teleoperation, which just means a person controls the robot remotely so it can learn from the movements.
Once Gallop is on the line, it keeps recording what happens. That data then flows back into training, so the robot should improve as it works. “It learns by watching our own workers, then practicing on the same parts,” Hsu wrote.
We’ve seen this “learn from human video” idea before, for example with NVIDIA’s GR00T 2. ASUS isn’t trying to invent a giant AI brain, though. According to Humanoids Daily, it starts from unnamed “leading open models” and builds on top of them.
The software comes in four layers. First, a planner reads the scene and breaks the job into steps. A supervisor model checks each step and re-plans when something goes wrong. Then a skills layer handles reaching, grasping and placing, while a control layer moves the arms and stops the robot if anything looks off. Because these layers connect through fixed interfaces, ASUS can swap in newer AI models, which it expects every few months, without rebuilding the robot.
Why tiny plastic bags are a big deal
This sounds like a boring job. That’s exactly why it’s a smart test.
The parts come in different shapes, and many sit in thin, bendy bags. Regular robot grippers struggle with that kind of stuff, because soft materials change shape every time you touch them. So if Gallop can handle this station, it can probably handle plenty of other pick-and-place work too. ASUS seems to think so as well.
The target is strict. ASUS wants a 99.5% success rate within the roughly 20 seconds it takes for each board to pass by. The company calls that “the same standard we hold a person to.” Fair enough, since that’s the bar a human packer has to hit every shift.
But here’s the honest caveat. ASUS hasn’t published a measured success rate yet. It only says the full loop now works on the line. So we know the goal. We don’t know the score.
What we still don’t know about the ASUS Gallop humanoid
Quite a lot, actually. ASUS hasn’t shared Gallop’s height, weight, payload, battery life or degrees of freedom (that’s how many separate ways its joints can move). There’s no price either. We also don’t know which open AI models it uses, or which factory it’s working in. Since ASUS builds on other people’s models, a lot depends on how fast those models keep improving.
The timeline is clearer, though:
- Within the next year, ASUS plans to add its own touch-sensitive hand and a wheeled base.
- It wants to show Gallop to customers at Computex 2027 in Taipei.
- Customer pilots are planned for 2027, with a release to partners in the second half of that year.
- A commercial launch is aimed at the second half of 2028.
Electronics factories come first. Later, ASUS is considering hospitals and other healthcare settings. That’s a big jump, so we’d treat it as an idea for now, not a plan.
Okay, so should factory workers worry?
ASUS says no, at least for now. The company pitches Gallop for repetitive, physically demanding tasks, not as a full replacement for people. It also says staff who help train the robot could move into setup, supervision or performance roles. That’s a nicer story than most robot launches tell, and it’s one worth holding them to.
We’ve also watched Xiaomi’s humanoid work a full factory shift and Apptronik’s Apollo head for GXO warehouses. So the pattern is clear: big companies want robots that learn from their own people, on their own lines.
For a tech-hungry country like Armenia, here’s the question we keep coming back to. If the ASUS Gallop humanoid learns from workers’ head cameras, who owns that know-how? The worker, the factory, or the robot maker? We don’t think anyone has a good answer yet.

