
What does it cost to build a humanoid robot? Morgan Stanley’s teardown of Tesla Optimus Gen 2 gives us the answer: approximately $55,000 in hardware alone.
But the real story isn’t the total—it’s where that money goes.
Looking at the robot from head to toe, the cost distribution is revealing. The head, containing FSD chips and cameras, costs just $2,100 (3.8% of the total). The compute brain of a humanoid is almost a rounding error.
The torso and battery come in at $8,600. That 2.3 kWh, 52V battery pack? Only $300. Again, surprisingly cheap.
Now things get interesting. Arms and shoulders cost $16,300—nearly 30% of the entire robot. These 10 rotary actuators with harmonic reducers and torque sensors enable the precise movements needed for real-world tasks.
The legs and feet represent the second-highest cost at $21,300 (38.6%). Eight linear actuators in the thighs and calves handle dynamic loads during walking, while 6D force sensors in the feet provide ground contact feedback.
But the single most expensive component per unit of size? The hands at $9,500 (17.2%). Elon Musk calls them “the single greatest electromechanical challenge” in humanoid development.
What This Tells Us
The $55,000 breakdown reveals something counterintuitive: this isn’t an AI problem—it’s a manufacturing problem.
Compute and batteries together represent less than 5% of the BOM. The challenge isn’t making robots smart. It’s making their bodies work.
The 28 actuators distributed across the robot account for roughly half the total cost. Each one must perform precise movements under variable loads, with tolerances measured in arcminutes.
Three companies—Maxon Motor, Harmonic Drive, and Kollmorgen—dominate the precision actuator market. Individual harmonic drives range from $2,000-$5,000 each, with custom high-torque variants hitting $15,000.
The Path to $20,000
Tesla’s target is a $20,000 production cost at scale—a 63% reduction from current levels. Is it achievable?
The levers are clear: scale economics (costs dropped 40% from 2022-2024), vertical integration (Tesla designing custom actuators), and automotive component reuse (FSD chips, thermal management).
Goldman Sachs projects another 40% BOM reduction with volume. The question is timing, not possibility.
The $55,000 humanoid is the skeleton key to understanding physical AI economics. Whoever solves precision manufacturing at scale—particularly for hands and legs—wins the race.








