Loading
Loading
The bill of materials is still heavy, uneven, and slow to compress. Some categories should fall with scale. Others remain stubborn.
$40-80K
Current mid-range humanoid BOM
~$35K
Bank of America estimate for China-sourced builds
$13-17K
Long-run mass-market target
The cheapest-looking line items are not always the ones that slow deployment. Precision and integration still matter most.
Motors, gearboxes, encoders, and controllers still dominate the bill because every joint needs precision, closed-loop control, and reliability.
Cameras, lidar, and onboard compute should keep getting cheaper as AI hardware and sensor supply chains improve.
Frames, shells, and joints depend on part count, manufacturability, and how much complexity can be designed out of the system.
This part of the stack benefits from the EV supply chain and is already more mature than much of robotics.
Tactile, force, torque, inertial, and temperature sensing remain harder to standardize at scale than cameras and compute.
Cables, connectors, PCBs, and assembly are still labor-heavy and less likely to behave like software cost curves.
Why actuators dominate
Each actuator combines a motor, transmission, encoder, controller, and tight tolerances. A humanoid may need dozens of them, and every weak joint shows up in reliability.
This is why a cheaper bill of materials does not automatically mean an easier product. Gains here come from manufacturing discipline, integration, and simpler designs.
Hardware cadence
Software can iterate in hours. Hardware revisions take quarters. Certification, tooling, and field reliability all move slower than the model layer.
Vertical integration may eventually matter, but early teams still have to choose what to own, what to buy, and where complexity can be removed instead of proudly accumulated.