ASUS Built a Humanoid Robot, Then Gave It a 20 Second Factory Test
ASUS introduced Gallop, its first humanoid robot, on November 13 and put it on a real factory line packing motherboard accessories. The goal is 99.5% success within 20 seconds per board, the same rate ASUS expects from people. Customer pilots are planned for 2027, with sales planned for the second half of 2028.
Gallop packs 15 small motherboard accessories at an ASUS group electronics factory.
ASUS targets 99.5% success within 20 seconds per board, matching a human worker.
A single NVIDIA Jetson AGX Thor computer inside the robot runs its AI.
ASUS designs the body and a touch sensing hand in house.
Customer pilots are planned for 2027, with sales targeted for the second half of 2028.
What did ASUS announce on November 13?
ASUS introduced its first humanoid robot and deployed it on a live production line. Co-CEO S.Y. Hsu presented Gallop on CNBC, and the company released a detailed post about it the same day. Gallop comes from the Humanoid Robotics unit of the ASUS Physical AI Solutions business group.
Albert Chang, an ASUS corporate vice president who co-leads that group and runs its motherboard business, wrote the company's post on Gallop. He framed it around a question factory managers ask when they watch robot videos. Can the robot hold their quality standard at their station for an entire shift?
The project started in early 2026. Today Gallop has a fixed base and two arms, and its strength is pick and place work at a single station. Chang summed up the difference between viral robot videos and factory work in one line:
"A demo shows that a robot can do something once. A factory needs a robot that does the job every day."
Why a motherboard packing station is a hard first job
ASUS chose a station that a simple gripper cannot handle. Every board comes with 15 small accessories, and many of them are packed in thin, clear plastic bags. Small, thin and soft parts are the natural enemy of a basic robot gripper.
The station takes about 20 seconds per board, and Chang laid out the goal in the company's post:
"Our target is 99.5% success within 20 seconds, the same standard we hold a person to."
That is a target, not a measured result, and ASUS has yet to publish a success rate. The company also set a second rule for itself: a robot has to do most of the job. A robot that needs a person beside it all shift does not save anyone any work.
ASUS has an advantage most robot startups lack: it owns electronics manufacturing plants, so it can test on real parts under real quality rules. Chang noted that most robotics companies have to rent that kind of access.
Gallop learns by watching factory workers
ASUS trains Gallop mostly on recordings of its own experienced workers. The company does not build its own giant AI model. It starts from leading open models that already understand how hands and objects behave.
With their consent, skilled workers wear a small head mounted camera while they do the job. Simulation software then generates variations of each example, so small changes on the line do not surprise the robot.
Next comes hands on practice, with an operator wearing a VR headset and gloves guiding the robot through the task. The practice uses the same parts and cameras as the real station.
Once Gallop is deployed, what it records feeds back into training. Chang wrote that the edge is not more data but the right data: a real station, real parts and real quality rules. He gave the credit to the people on the line:
"Our workers are the teachers, and their know-how is what Gallop is built on."
A modular brain lets ASUS swap in better AI
Gallop splits its intelligence into four layers so ASUS can upgrade one without rebuilding the robot. A planner takes in the scene and breaks the job into steps. A supervisor, which is a second AI, checks each step and plans again when something goes wrong.
A skills layer handles reaching, grasping and placing. A control layer moves the arms and hands precisely and stops instantly if anything is wrong. The layers talk to each other through fixed interfaces.
This structure matters because new AI models now arrive every few months. ASUS can drop a better model into one layer and leave the rest of the robot unchanged.
All of it runs on a single NVIDIA Jetson AGX Thor computer inside the robot, with standard robotics software. NVIDIA markets Jetson Thor as a platform for physical AI and robotics. Its developer kit pairs a 14 core Arm processor with 128 GB of memory.
Is the robot hand the hardest part to build?
ASUS says the hand is the hardest part of any humanoid, so it is building its own. The company designs Gallop's body and a dexterous, touch sensing hand in house. Motors and other standard components come from established partners.
Chang wrote that capable robot hands are still scarce worldwide. Touch matters for this job in particular, since the robot has to feel thin plastic bags that cameras can barely see.
Several Asian governments are also treating robot parts as strategic. Earlier this week, a Sparknify report looked at South Korea's plan to fund robot sensors, actuators and humanoid batteries. It found that Seoul wants local suppliers to hold 80 percent of core robot components by 2030. The same report noted that Taiwan's cabinet approved a NT$10 billion smart robotics plan in May 2025.
ASUS also joins a race where other players already post results. On November 10, Google DeepMind said Apptronik's Apollo 2 humanoid reached a 96% success rate sorting metal brackets in a Mercedes-Benz pilot. The pilot ran on active production lines in Hungary and Germany, according to Google DeepMind's Gemini Robotics 2.5 announcement.
What happens when humanoids start clocking in at electronics plants?
The next 18 months will show whether a humanoid can keep up with a person over an entire shift. Over the coming year, ASUS plans to give Gallop its own hand, add wheels and show it to customers at Computex 2027 in Taipei. Customer pilots follow in 2027, and ASUS is aiming for market availability in the second half of 2028.
For Silicon Valley, the move matters because ASUS is a major player in the hardware supply chain. A PC and server maker that sells humanoids would compete with startups such as Apptronik for factory customers. It could also become a customer for founders working on hands, sensors, actuators and robot software. ASUS says electronics manufacturing comes first, with hospitals and service sites to follow as pick and place skills transfer.
The worries start with jobs. ASUS says Gallop is meant for repetitive, physically demanding stations that are hardest to staff. Workers who teach it are supposed to move into setting up, supervising and improving the robots. Chang tackled the question directly in his post:
"People often ask whether Gallop is meant to replace workers. It isn't."
Labor advocates will want to see ASUS keep that promise once the robots meet their goals. Capturing workers' skills on head cameras and passing them to a robot also raises questions about consent and about who owns that know-how.
Skeptics will note that 99.5% is still a target, and that humanoid robots have been demoed far more often than deployed. A two armed robot on a fixed base also looks a lot like automation that already exists, so why make it humanoid? ASUS answers that factories are built around people, so a robot with two arms and hands can use existing stations and tools without retooling. Supporters add that owning its factories lets ASUS test Gallop on real lines before charging customers for it.
Five things will show which way this goes:
A measured success rate for the packing station, published before the 2027 pilots.
The in house touch sensing hand and a wheeled base on Gallop within a year.
Gallop's customer demonstration at Computex 2027 in Taipei.
Which outside factories sign up for Gallop pilots during 2027.
Pricing and a firm launch date ahead of the second half of 2028.
ASUS has set a simple bar for Gallop: do the job every day at human speed, or it is just another demo.
Subscribe to Sparknify for tech news and the people behind it.
Questions & Answers
Q1: What will ASUS need to prove with its Gallop humanoid robot before customer pilots in 2027 and sales in 2028?
Gallop must hit 99.5% success within 20 seconds per board on its packing station, and keep it up shift after shift. ASUS also plans to add a touch sensing hand and a wheeled base. Precision parts will be key to that progress, such as Mountain Gracious Tech's hydrostatic bearings for robot joints.
Q2: Will humanoid robots like ASUS's Gallop replace electronics factory workers, or will they work alongside them?
ASUS says Gallop is designed for repetitive, hard to staff stations, and the workers who train it take on setup, supervision and improvement roles. Skeptics want proof that the promise holds at scale. The other extreme is China's dark factory model, where a Chengdu plant reportedly makes stealth fighter parts with minimal human intervention.
Q3: I'm a Taiwanese founder building robot hands and joints. How can I get in front of Silicon Valley investors as ASUS moves into humanoid robots?
ASUS designs Gallop's hand in house but buys motors and standard components from established partners, so parts suppliers have a role. Taiwanese startups in robotics and deep tech have pitched Silicon Valley investors at Taiwan Venture Day, where TREE backed teams incubated at Stanford and Berkeley SkyDeck present.
Comments