Actuation can represent 40–60% of humanoid BOM — the next robotics winner may not be the robot brand

The most expensive humanoid subsystem may be the joints, not the brain

McKinsey estimates actuation accounts for roughly 40–60% of a humanoid robot's bill of materials, making it the largest cost block.

AI compute and perception matter, but the cost and manufacturability of physical joints can become an equally important commercialization constraint.

An actuator is not just a motor

A humanoid joint typically integrates a motor, gearbox, driver, encoder, force or torque sensing, bearings and other mechanical components.

McKinsey estimates the gearbox alone can represent roughly 30–50% of actuator cost, affecting torque density, backlash, durability and joint size.

Dozens of joints create a large supplier value pool

A humanoid needs many joints to walk, rotate its torso, reach, grasp and manipulate. McKinsey cites 28 body joint actuators in Tesla Optimus as one example.

That means one robot sale can require dozens of repeated joint systems, creating a potential market outside the finished robot brand itself.

The industry still lacks an automotive-style standardized actuator ecosystem

Mature automotive OEMs do not design and manufacture every subsystem themselves. Tier-1 suppliers sell modules and common technologies across multiple brands.

Humanoids are earlier in that evolution. Leading projects still rely heavily on custom development, and the supplier structure remains pre-modular.

Vertical Integration ≠ Strategic Preference

An OEM building actuators internally does not prove that vertical integration will always be the best strategy.

Low volume, changing designs, limited standard parts and a shortage of qualified suppliers can force OEMs to internalize work they might later source externally.

The low-volume trap slows supply-chain maturity

OEMs face high component prices, while suppliers hesitate to build dedicated lines before demand is proven.

That can create a loop: Low Robot Volume → Low Supplier Investment → High Component Cost → High Robot Price → Limited Demand → Low Robot Volume. McKinsey places current humanoid BOM broadly around $30,000 to $150,000 and notes that sub-$20,000 is often discussed as a long-term mass-market threshold.

Harmonic drives and roller screws do not scale like software

Compact high-torque humanoid joints depend on highly precise mechanical parts such as strain-wave gearboxes and planetary roller screws.

These components require tooling, metrology, qualification and skilled manufacturing. Software can be copied instantly; precision hardware capacity only grows when factories, equipment and trained people grow.

China's advantage may sit below the robot brand

McKinsey highlights China's deep manufacturing base across magnets, precision bearings, motors and power electronics, all relevant to humanoid systems.

IFR reports China installed 354,000 industrial robots in 2025, 59% of global installations, while domestic suppliers held 55% of China's market. Large domestic deployment can accelerate volume learning across suppliers and OEMs.

Early vertical integration can evolve into specialist platforms

If humanoid volumes rise and joint requirements stabilize, multiple OEMs may converge on similar performance ranges.

A specialist supplier could then aggregate demand across brands, invest at larger scale and drive down cost faster than each OEM building everything alone. Standard joint modules or actuator platforms could emerge.

The most famous robot company may not capture the most value

A robot brand earns from its own units. A component platform used across shoulders, elbows, knees and wrists of many brands can participate in industry-wide volume growth.

The EV industry created large value pools in batteries, inverters, power semiconductors and motors. Humanoids may develop similar downstream winners.

Physical AI talent demand extends far beyond AI software

Scaling humanoids requires not only foundation-model and perception engineers, but precision-motion, motor, reducer, roller-screw, bearing, sensing, power-electronics, thermal, manufacturing and supplier-quality expertise.

Physical AI can therefore increase the value of mature mechanical and electrical engineering capabilities rather than replacing them.

BANSEOG VIEW | The humanoid supply chain is still unfinished

A 40–60% BOM share does not mean actuator suppliers will capture 40–60% of future industry value. Designs and standards can change significantly.

But an immature supply chain creates room for new industrial positions. Low Volume → Custom Parts → Vertical Integration can potentially evolve into High Volume → Standardization → Supplier Investment → Platform Components.

The next Physical AI question is not only who builds the smartest brain, but who can mass-produce the body with the right cost, precision and reliability.