ASUS's First Humanoid Robot Must Hit 99.5% Accuracy Packing Motherboards

ASUS unveiled Gallop, a two-armed humanoid robot running on a single Nvidia Jetson AGX Thor chip, now learning to pack motherboard parts on its own factory floor.

Key Takeaways

  • ASUS's humanoid robot Gallop is being trained to pack 15 motherboard parts in about 20 seconds with a 99.5% success target.
  • Gallop runs on a single Nvidia Jetson AGX Thor chip and uses an in-house dexterous hand with tactile sensing.
  • ASUS is testing the robot inside its own electronics manufacturing plant rather than with an outside customer first.
  • The program started in early 2026, with public demos planned for Computex 2027 and market availability targeted for the second half of 2028.
  • Gallop currently has a fixed two-arm base; ASUS says a wheeled, mobile version is already planned.

ASUS's First Humanoid Robot Must Hit 99.5% Accuracy Packing Motherboards

ASUS just confirmed it is building a humanoid robot, and the company skipped the usual trade-show reveal. Instead, it put the robot — called Gallop — straight onto a production line at one of its own electronics manufacturing plants, where it is learning to pack motherboard accessories well enough to hit a 99.5% success rate.

What is Gallop actually doing right now?

Gallop works a single fixed station, picking up 15 small parts per motherboard and placing them in roughly 20 seconds, according to ASUS. That 99.5% target is the bar ASUS says it needs to clear before the robot is worth more on the line than the workers it would stand next to.

The robot has two arms on a fixed base for now, though ASUS says a wheeled version is already planned. Its hands were built in-house with tactile sensing, which the company calls the hardest and scarcest piece of humanoid robot hardware in the world right now.

How does Gallop actually learn the job?

ASUS trains Gallop in stages. It starts from an open-source foundation model, then watches experienced workers perform the task through head-mounted cameras, and finally practices on real parts and equipment while a human operator guides it remotely through VR. ASUS describes this as a continuous feedback loop rather than a single training run that gets frozen in place.

Underneath that sits a four-layer software stack — a planner, a supervisor, a skills layer, and a control layer — built so ASUS can swap in newer AI models as they appear without rebuilding the whole system. The entire stack runs on one Nvidia Jetson AGX Thor chip. "A factory needs a robot that does the job every day," the company said. "That is the standard we have set."

Why would a motherboard maker build its own humanoid robot?

ASUS already runs its own electronics manufacturing services operation, building motherboards and components at scale. That gives it a captive test site most humanoid robot startups don't have — real production lines, real defect data, and a direct financial incentive to make the robot work rather than just demo well for investors.

It also puts ASUS in the same race as Figure, Apptronik and Agility Robotics, all of which have run similar pick-and-place pilots at electronics and auto plants this year. ASUS's advantage is that it doesn't need to convince an outside customer to let a new robot onto their floor first — it can fail and fix Gallop on its own line before anyone else sees it.

When will Gallop actually ship to customers?

ASUS says the program only started in early 2026, and the company plans to show Gallop publicly at Computex in 2027. Customer pilots are slated for 2027, with broader market availability targeted for the second half of 2028 — electronics manufacturing first, then hospitals and other service settings later.

That timeline puts ASUS several years behind humanoid robot makers already running paid pilots with automakers and logistics firms. The company's bet is that owning its own factory floor lets it close the reliability gap through repetition rather than marketing, even if it means arriving later to a crowded field.

Category: Humanoid Robots

Tags: Physical AI Manufacturing humanoid robot ASUS Gallop Nvidia Jetson AGX Thor

Related Articles

Frequently Asked Questions

What chip does ASUS's Gallop robot run on?

Gallop runs its entire control stack on a single Nvidia Jetson AGX Thor computer.

What is Gallop's current job?

It packs motherboard accessories at an ASUS electronics manufacturing plant, placing 15 small parts per board in about 20 seconds.

When will ASUS sell Gallop to other companies?

ASUS targets customer pilots in 2027 and broader market availability in the second half of 2028, starting with electronics manufacturing.

How does Gallop learn new tasks?

It starts from an open-source foundation model, watches humans perform the task via head cameras, then practices with a VR-guided human operator before working unsupervised.