I Broke Down the Humanoid Robot Bill of Materials: The Brain Is 6%, the Joints Are Two-Thirds

Start with the bill of materials and the investment map changes

About two-thirds of a humanoid robot's cost lives in the moving parts: the joints and the hands. The AI brain everybody talks about? Roughly 6%.

Putting those two numbers side by side flipped how I think about robotics investing. It stops being a game of guessing which brand wins and becomes a game of finding who makes the parts that every brand has to buy. Optimus, Atlas, Figure, whoever comes out on top still needs the same reducers and the same motors.

The market is growing north of 30% a year and is estimated to pass $100 billion within the decade. And demand hasn't really started yet. We're on the ground floor here. I laid out how the robotics value chain stacks up in The 4-Tier Robotics Investment Framework; today's piece is about the bottom tier, the companies actually machining metal.

One more number worth holding onto. Build a humanoid without Chinese parts and the cost roughly triples, from about $46,000 to $131,000. That's how a machine quietly became a national security problem, and it's why I look at suppliers before I look at brands.

The joint is the most expensive and the scarcest thing in the robot

The heart of a robot joint is the reducer, the part that converts a motor's fast, weak spin into slow, powerful torque. Nearly every advanced robot joint on Earth runs on one of two types: a strain wave reducer or a cycloidal reducer.

A humanoid uses both, in its hips, its shoulders, its knees. On a machine with dozens of joints, every single one of them is a precision-machined component. That's why I think the real bottleneck in robotics is the joint, not the brain. You can add fab capacity for chips; adding precision reducer capacity takes years of tooling and skilled labor.

Timken (TKR): the only US-listed company making both reducers under one roof

Let me lead with the conclusion. On a US exchange, exactly one company makes both gear types in house, and that's Timken.

The 125-year-old bearings maker builds precision gearing through two businesses, Cone Drive and Spinea, and those gears sit at the heart of a robot's joint. They aren't a someday product either. They already ship in real volume into industrial robots, factory floor arms, medical devices, defense systems, and solar trackers. This is a profitable business today, not a promise.

The numbers back it. Revenue has climbed roughly 60% over the last decade to $4.6 billion. The engine behind that is the Industrial Motion division, the side making these robot gears, now running at $1.5 billion a year, about 34% of the whole company.

Here's the part I like most. While the business grew, management shrank the share count by about 28%, from 98 million to 70 million. Every dollar of profit now lands on far fewer shares. Adjusted earnings are set to climb nearly 20% toward $7.21.

The catch is the price. The stock sits near the top of its 10-year range. Great business, price got ahead of it, so this is a watch list name for me rather than a buy.

Harmonic Drive and Nabtesco: the world's best are still Japanese

Being honest about it, the only two companies on the planet that build these gears better than Timken are Japanese, and in the US you can only reach them over the counter.

Harmonic Drive Systems invented the strain wave reducer and still dominates the high end. A single humanoid can use 20 to 30 of them. Nabtesco holds roughly 60% of the cycloidal reducer market and already supplies the two biggest robot arm builders in the world, Fanuc and ABB.

OTC shares trade thinner with wider spreads. Use limit orders, keep position sizes sensible, and you'll need a broker that supports over-the-counter names. Robinhood does not.

Company	Ticker	Key part	Where it trades	My read
Timken	TKR	Both strain wave and cycloidal	Major exchange	Watch list, valuation stretched
Harmonic Drive	OTC	Strain wave (inventor, high-end leader)	Over the counter	Best in class if you accept access risk
Nabtesco	OTC	Cycloidal (~60% share)	Over the counter	Already supplies Fanuc and ABB
Regal Rexnord	RRX	The whole joint (motor + ball screw + gear)	Major exchange	Best price in the basket today

Regal Rexnord (RRX): the company that sells the entire joint

If you want the one name in this basket where the price still works, it's Regal Rexnord.

Across its motion brands, one company makes the servo motor, the micro motor, the ball screw, and the precision gear that together form a robot's joint. Every other pure play sells one piece. Regal can sell a humanoid builder the whole assembly from a single supplier, and consolidating vendors matters more to an assembler than people give it credit for.

That stack already earns its living inside factory automation, aerospace, medical, and packaging. Cash flow, not a promise. And it's the identical content a humanoid joint needs next.

Here's where most people misread the stock. On trailing earnings it looks expensive, close to 49 times profit. That number is an accounting illusion. When Regal bought Altra for about $5 billion, the way that deal gets written down buries the company's real earnings. The scary figure is paper, not cash. On next year's actual profit it trades closer to 15 times.

The growth underneath is what matters. Over five years Regal has more than doubled revenue to $5.9 billion, and free cash flow more than doubled right alongside it to $893 million. It's using that cash to pay down debt fast, and every dollar of interest expense that disappears drops straight into earnings. That's why profit growth is running near 30%, and why the PEG sits close to 1.

Growing earnings against a modest price. In this basket, that combination is the rarest thing on the list.

Roller screws and linear guides: where America is thinnest

There's one hole left in the framework, and it happens to sit right on top of the scarcest, highest-value parts in the machine.

The roller screw can run up to a third of the cost of a linear actuator, making it one of the tightest choke points in the entire robot. The ultra-precision specialists, Rollvis and GSA in Switzerland, are private, so you can't buy them. But two of the giants chasing this market do trade over the counter: Sweden's SKF, building roller screw modules purpose-built for humanoids, and Germany's Schaeffler, which just signed a major deal to supply actuators for a whole fleet of humanoids.

Same story for the rails those actuators slide along. Linear guides are led by Taiwan's Hiwin and Japan's THK, and both are reachable over the counter.

So is there an American answer? Sort of. The name I watch is Moog (MOG.A). It builds world-class actuators today, but they go into fighter jets and spacecraft, and its robot revenue is essentially zero. What it does have is the engineering to break into this space. So it goes on the watch list not because it has a robot story today, but for the day it lands an actual humanoid win. That would be the signal America is finally closing the one gap it can't build at scale.

How I'd summarize the basket

The appeal of a parts basket is that you never have to pick the winning robot. The discipline required is on price, because a lot of these names have already run.

• Priced where I'd own it today: Regal Rexnord (RRX). Sells the whole joint, ~15x next year's earnings, PEG near 1.
• Watch list: Timken (TKR). Terrific business, stock near the top of its 10-year range.
• Best in class, OTC only: Harmonic Drive, Nabtesco, SKF, Schaeffler, Hiwin, THK. Limit orders and modest sizing.
• Waiting for a signal: Moog (MOG.A). The trigger is a first humanoid contract.

If you want more underpriced hardware names in the same spirit, I went through several in Hidden Value Stocks and the Robotics Revolution.

Where this thesis could be wrong

Fair is fair, so here's the other side.

First, if humanoid demand arrives later than expected, these companies' earnings just track the industrial automation cycle. The robot premium baked into their multiples gets cut first.

Second, "all the parts makers win" is also another way of saying nobody earns monopoly margins. As assemblers scale, they squeeze component pricing. Irreplaceable parts like reducers and substitutable parts like motors will not follow the same margin path.

Third, Chinese suppliers are pushing hard to localize reducers and screws. That triple-the-cost math is a snapshot of today, not a permanent truth.

My conclusion still stands, though. Picking the company that makes what every winner has to buy is a much easier calculation than picking the winner. It just now comes with price as an extra variable in the equation.

This is educational analysis, not financial advice.