A humanoid costs $46,000 to build, or $131,000 without China. That is the de-risking bill.

Break the $85,000 down and it has a shape ^[1]. Actuators are 40% to 60% of a humanoid's bill of materials ^[11], which on a $46,000 build is $18,400 to $27,600 ^[4]. Semiconductors are about 10%, roughly $4,600, and the report says most of those are mature 200mm and 28-to-90nm parts almost nobody is short of ^[12][5]. The scarce items sit further down the list: strain-wave and cycloidal reducers, planetary roller screws, frameless torque motors, precision bearings and rare-earth magnets ^[13]. So the increment is not buying chips. It is buying gearing and magnets, and it works out to 1.85 times the entire cost of the robot as it is built today ^[3].

That is why the US position reads worse the closer you look at it. America tops the index on compute at 91.6 and human capital at 78.7, and scores 48.5 on materials and supply chain, fifth overall, against China's 95 ^[8]. The headline gap between China at 82.9 and the US at 69.0 is 13.9 points; on the materials line it is 46.5 ^[7]. Jan Liphardt, OpenMind's founder and a Stanford bioengineering professor ^[3], told the Humanoid Daily podcast that the assumption that the best models win physical AI is, as far as his team can tell, not true ^[17].

The regulatory half of this is where the price tag stops being hypothetical. Forbes reports that the FCC has added foreign-produced advanced robotic devices to its Covered List, covering sidewalk delivery bots, inspection quadrupeds, humanoids and essentially any ground-mobile machine over two kilograms with sensors, autonomy and a real network connection, so foreign-made units can no longer be authorized for the US market ^[9]. IDC's read, per the same piece, is that adoption speed takes a serious hit if the rules stay ^[10]. But the rule as described reaches finished machines, not the reducers inside them, and that distinction is exactly where the $85,000 lives. A robot assembled in the US from Chinese subassemblies and a robot with no Chinese content in it are two different price points, and the source does not say which one the Commission is requiring. Anyone building a 2027 budget is budgeting inside that ambiguity.

Liphardt's summary of the task is that the entire US robotics industry now has to work out how to make advanced robots in the US ^[14]. The one asset the country clearly leads does not appear on a bill of materials at all. Model quality has no line item, and Liphardt argues the edge erodes anyway as China open sources advanced physical AI models ^[15]. Better software does not lower the price of a strain-wave reducer.