By RWT Robotics Editorial Team Published
DEMO ONLY
ASUS has announced Gallop, its first humanoid robot, and CNBC reports it is being tested on an ASUS group factory line. According to Unite.AI’s reporting of ASUS statements, Gallop is learning to pack 15 motherboard accessories at what ASUS calls human pace. That pace is a training goal. Nobody has independently shown the robot hitting it.
What we know
The core fact is simple. CNBC reports that ASUS has a humanoid called Gallop, that it is the company’s first, and that it is running trials on a factory line inside the ASUS group. That is a real program with a real test environment, which already puts it ahead of a lot of humanoid announcements that exist mainly as renders.
The task is specific, and that specificity is the most interesting part of the story. Unite.AI, reporting ASUS’s own statements, says Gallop is learning to pack 15 motherboard accessories. The stated target is 99.5% success within 20 seconds per board. ASUS describes that as human pace.
Unite.AI is clear on the key point: this is a training goal, not a demonstrated result. The robot is learning the task. It has not been shown completing it at that rate.
The timeline comes from CNBC. ASUS plans to show Gallop at Computex 2027 with a wheeled base and a new hand. Customer pilots are planned for 2027. CNBC attributes the rest of the schedule to ASUS co-CEO S.Y. Hsu: release to partners in the second half of 2027 and a market launch in 2028.
The idea isn’t new inside ASUS. TweakTown reported in 2024 that chairman Jonney Shih said the company was developing a humanoid robot. The Gallop program itself reportedly began in early 2026, which would make it less than a year old as of this writing.
What we don’t know yet
Almost everything that would let an engineer judge the claim is missing.
We have found no independent footage of Gallop packing boards, and no third-party verification of its speed or success rate. Neither the CNBC nor the Unite.AI reporting includes deployment numbers: how many units are on the line, how many hours they have run, what share of boards they handle, or what success rate they hit today. That is why we label this DEMO ONLY.
The target itself needs unpacking. Fifteen parts in 20 seconds works out to roughly 1.3 seconds per part, covering grasp, transport, orientation and placement. Small accessories like screws, standoffs and cables are some of the hardest objects in manipulation. They are tiny, often reflective, sometimes deformable, and they tangle. Hitting that cycle time once on camera is a very different thing from hitting it across a shift.
The 99.5% figure raises its own questions. It means about one failed board in every 200. We don’t know how ASUS defines a failure: a dropped part, a missing part, a misplaced part, or a stoppage that needs a human. We also don’t know whether “within 20 seconds” is an average, a ceiling, or a best case. Those definitions decide whether the number means anything.
“Human pace” does have a stated basis. ASUS’s own press blog says the 99.5% success rate within 20 seconds per board is the standard it holds a person to, so the robot’s target mirrors the company’s human benchmark for the job.
Finally, the hardware is still changing. CNBC reports that the Computex 2027 version will have a wheeled base and a new hand. A new hand matters a lot for this task. Whatever Gallop learns on today’s hand may not carry over cleanly, and any performance numbers from the current build may not describe the robot partners actually get.
What it means
Credit where it is due. ASUS picked a narrow, measurable, unglamorous job instead of promising a general-purpose household helper. Packing motherboard accessories is exactly the kind of repetitive, high-volume work where a robot could earn its keep. A company that builds the boards also controls the line, the parts and the test conditions, which is a reasonable place to start.
A published target with a success rate and a cycle time is also more useful than the usual vague claims. It gives everyone something to check later.
But a target is not a result. Testing in a factory is not the same as being proven in a factory. Until ASUS publishes measured throughput, failure definitions and run time, or until someone outside the company sees the robot work at speed, the human-pace claim belongs in the “goal” column.
The schedule is ambitious too. A program that reportedly started in early 2026 is aiming for customer pilots and a partner release in 2027, with new locomotion and a new hand arriving on the way. Hardware revisions this late tend to reset some of the learning work. If the timeline holds, that will be impressive. If it slips, that would be normal for this field.
Computex 2027 is the next checkpoint. The useful question there won’t be whether Gallop can pack a board on stage. It will be whether ASUS shows sustained runs, error rates and real line data. Those are the numbers that separate a deployable robot from a good demo, and so far ASUS hasn’t shared them.
Follow every development in one place on our humanoid robot tracker.
Keep reading
Robotics NewsNvidia CEO Huang resigns, leaving Thor and GR00T robotics partners with questionsHuang resigned after a personal scandal, per CNBC and CNN. What it means for Thor and GR00T users,…Nov 13, 2026 · 5 min read
Robotics NewsFigure AI night-shift coordinator posting pays $45 an hour, notes limited engineering supportA Figure AI posting offers $40-$45 an hour to lead a night crew for humanoid robots. It's one…Nov 12, 2026 · 4 min read
Robotics NewsAgile Robots’ $8.5B valuation rests on a 2,000-humanoid target the company set for 2026Agile Robots raised $500M at $8.5B, led by SoftBank with Nvidia. Its humanoid output figures are company claims,…Nov 12, 2026 · 4 min readCredits and sources
By RWT Robotics Editorial Team
- Published
