What is this problem
This covers the electromechanical components that make up a robot’s joints: motors (including frameless torque motors and integrated servo actuators), reducers and gearboxes (harmonic drives, planetary, cycloidal, and RV designs), and encoders that provide position and torque feedback. Together these parts set the torque density, backlash, precision, speed, and range of motion available at each joint.
Humanoid robots are especially dependent on this layer: a single humanoid needs 20-40+ actuated joints, each requiring a different combination of torque, speed, and size, so the actuator/reducer/encoder stack effectively is the robot’s musculoskeletal system and a large share of its bill of materials.
The bottleneck and pain points
The core tension is torque density versus cost: packing more torque into a smaller, lighter joint generally means tighter tolerances, better materials, and more precise assembly, all of which raise unit cost. Backlash and repeatability requirements for fine manipulation push toward harmonic or cycloidal reducers, which are more expensive and harder to manufacture at volume than simpler planetary gearing.
Components also face real durability constraints (wear on flex-splines and bearings under continuous duty cycles, lubrication degradation, heat buildup, and noise/vibration) that matter enormously for humanoids expected to run all day in commercial settings.
Underneath the engineering tradeoffs, this is fundamentally a precision-manufacturing and supply-chain problem rather than an algorithms problem: winning requires multi-year qualification cycles with OEM customers, high manufacturing yield on tight-tolerance parts, and durable relationships, while margins are constantly pressured by customer concentration (a few large robot makers can dictate price) and by price erosion as more suppliers scale up capacity. This is why Japan’s decades-long lead in harmonic and precision reducers, Germany’s strength in servo motors and encoders, and China’s fast-scaling, lower-cost manufacturing base matter so much: production economics and manufacturing depth, not just component specs, will decide who captures value here.