Handlike robots for light industrial work
Mike Xia, CEO of Anvil Robotics, on humanoid vs. non-humanoid robots
The core bet is that the winning robot for light industrial work will be built less like a factory arm and more like a careful pair of hands. Traditional automation was tuned for fixed stations, exact coordinates, and high payloads, where success means repeating the same motion all day with very high precision. Anvil is aiming at the opposite problem, messy assembly and packing jobs where objects shift, bags crumple, and the robot has to adapt instead of just replaying a path.
-
ABB and other industrial robot vendors optimize around payload, speed, and repeatability. Current ABB systems span small 9 kg arms with 0.011 mm repeatability up to heavy robots handling 800 kg to 1,000 kg, which fits welding, machine tending, and material handling far better than fiddly plug in and pack out work.
-
Humanoid companies like Figure, Apptronik, and Foundation mostly enter through factory tasks, but their pitch is still shaped by the old automation bottleneck, expensive retrofits and rigid cells. Anvil sits one layer down the stack, selling arms, legs, and actuators into that ecosystem, which reflects a component first strategy rather than a full robot fleet rollout.
-
The software implication is just as important as the hardware one. Older robots depended on engineered environments and hand written motion logic, while newer systems lean on data driven action models, teleoperation, and learning from edge cases. That matters most in variable workflows like electronics assembly, kitting, labeling, and packaging, where people still outperform classic robots.
This points toward a robotics market that splits by task, not by whether a machine looks humanoid. The next winners in light manufacturing will be the companies whose hardware survives contact, whose models improve from messy real deployments, and whose systems can automate human stations without forcing a full factory rebuild.