ASUS Unveils Gallop Humanoid Robot Powered by NVIDIA Jetson AGX Thor for Factory Physical AI
ASUS details Gallop, its first humanoid robot built on NVIDIA Jetson AGX Thor for electronics factory lines, targeting 99.5% accuracy packing 15 motherboard parts in 20 seconds.
ASUS has officially detailed Gallop, its first humanoid robot engineered to bring physical AI directly onto high-precision electronics assembly lines. Running a four-layer AI stack on a single NVIDIA Jetson AGX Thor module, the dual-arm system is currently training on real motherboard production floors to achieve a 99.5 percent success rate packing 15 distinct components within a 20-second human-equivalent cycle.
Factory-Proven Physical AI: Matching Human Pace on Production Lines
Rather than debuting a general-purpose prototype in controlled lab demos, ASUS designed Gallop to operate within existing human-centric electronics factories without requiring workstation redesigns. In its initial fixed-base, dual-arm configuration, Gallop is tasked with picking and packing 15 motherboard accessories per board, targeting a 99.5 percent reliability threshold inside 20 seconds.
Four-Layer Software Stack on NVIDIA Jetson AGX Thor
Gallop executes its real-time perception, motion planning, and force-control loops through a four-layer AI software architecture hosted on a single NVIDIA Jetson AGX Thor compute module. To accelerate skill acquisition for delicate cables and small fasteners, ASUS employs VR-guided teleoperation training where human factory workers demonstrate nuanced manipulation workflows directly to the robot.
In-House Dexterous Hands, Computex 2027 Pilots, and 2028 Rollout
ASUS is developing both the robot chassis and a custom touch-sensing dexterous hand internally while sourcing actuators from established robotics suppliers. The company plans to integrate the tactile hand and add a wheeled mobile base ahead of public demonstrations and external customer pilot programs at Computex 2027, targeting broad commercial market availability in the second half of 2028.