A humanoid robot is a bipedal machine built in the human form factor — two legs, two arms, hands, cameras for eyes — so it can work in spaces designed for people (factories, warehouses, eventually homes) and use the tools people already use. The "brain" is an AI model; the "body" is a bundle of about 40+ small precision motors (actuators), gears, sensors, and a battery. The product sold is a physical robot unit (plus, increasingly, the software/service that makes it useful). The reason to want one: it is a machine you can point at almost any manual task. The whole thesis hinges on the brain getting good enough — that is the AGI bet. With a brain that generalizes, the same body does any physical job; without it, these are expensive demos. This sheet lays out the facts and the arithmetic; it draws no conclusion about whether to buy.
Long-run demand for humanoids could exceed any supply that exists today — but only if the AI brain crosses the usefulness threshold, which is the open question. Today supply is not the binding limit (nobody is selling millions of units); software intelligence is. When demand does turn on, the constraint shifts to precision parts (actuators, and harmonic-drive gears — the high-precision gear units that form a robot's joints) with multi-year lead times. In money terms, almost nothing you can buy today is priced on actual humanoid cash flow: TSLA carries a large slab of humanoid "option value" (value the market assigns to a possible future payoff that produces no cash yet) inside a price mostly justified by cars and energy; ABB and ROK are priced as profitable industrial businesses with humanoid upside thrown in. This describes where the money sits — it is not a view on whether to buy it.
The unit of sale is one humanoid robot. Think of it as a labor-replacement appliance. The companies in this group can earn cash from it in three ways, and which one dominates is still unsettled:
For the diversified names (ABB, ROK), the near-term money is not from humanoids at all — it is from selling the parts and motion-control systems that any robot, humanoid or not, must contain: servo motors, drives, controllers, robot arms. They earn cash today regardless of who wins the humanoid race.
Source: 500-stocks scan, Module 8, Sectors 01 (Humanoids) & 07 (Actuators/Motion Control).
Today (known): Essentially zero at scale. paper The humanoid platform companies are pre-revenue or earning only tiny pilot/demo revenue. There is no installed base, no backlog of real orders to speak of, and no published unit-shipment number that is material. This is the honest starting point: demand for the finished humanoid is currently a forecast, not a fact.
The forward case (forecast, AGI lens): The premise here is that AGI is arriving, which means the missing ingredient — a brain that generalizes across physical tasks — gets solved. Once a humanoid can reliably do, say, the bulk of warehouse, factory, and basic service-labor tasks, the buyer pool becomes "anyone who currently pays a human to do manual work." That is why the scan frames the addressable market as a slice of the global manual-labor wage bill, on the order of est. tens of trillions of dollars per year (approximate, not live-verified). The arithmetic an owner would run: a robot that costs ~$20K to build and replaces even a $30K/yr worker would pay back its build cost in well under a year, then keep producing — if those cost and capability figures hold. That payback math is what would make demand step-change rather than gradual. Both inputs are targets, not contracted facts.
Today (known): Build capacity for finished humanoids is tiny — prototypes and small pilot runs. The scan's point is that supply is not the binding constraint right now, because demand has not turned on. You cannot be "sold out" of a product nobody is mass-ordering yet. The real limit today is the AI software stack — the brain — not the factory.
The bottleneck once volume scales (forecast): When the brain crosses the threshold and orders surge, the constraint moves from software to a short list of precision physical parts:
Who controls supply / market-share structure: for the finished humanoid, no one has share because there is no market yet — TSLA has the combination of manufacturing scale and AI talent, Hyundai (via Boston Dynamics) has, per the scan, the most advanced locomotion, and Agility would be a pure play (a company whose value rests almost entirely on this one product) if/when it lists publicly. For the parts that gate scale, supply is concentrated among a handful of Japanese/German specialists — an oligopoly (a market supplied by only a few producers) — which is why component suppliers are the structurally tightest link.
Source: 500-stocks scan, Module 8, Sectors 01, 07, 11 and Cross-Cutting Notes. Part-maker market shares and lead times are scan figures and are approximate / not live-verified.
This is the unusual case where the gap is latent — not yet visible in prices or lead times. The product is neither short nor oversupplied today because the market hasn't started. The question is what happens at the flip point — the moment AI capability makes the robot broadly useful. At that instant the gap inverts quickly: very large demand (labor replacement) meets a hardware supply chain that, per the scan, physically cannot scale parts in under 2-3 years.
| Layer of the product | Supply state today | What flips it | How fast supply can respond |
|---|---|---|---|
| Finished humanoid (the platform) | Pre-revenue, no real demand | AI brain crosses usefulness threshold (AGI) | Assembly itself: ~2-4 yrs to scale est. |
| Actuators (muscles) | Adequate for today's tiny volume | Mass orders for 40+ per robot | ~16-26 wk lead times; new lines slow est. |
| Harmonic/cycloidal gears (joints) | Already tightest in robotics | 20-40 per robot at million-unit scale | ~12-18 mo lead; 2-3 yrs new capacity est. |
| AI brain (software) | The current binding constraint | Frontier model progress | Copies at near-zero cost once solved |
Evidence today is sparse by design: there is no sold-out capacity, no pricing trend, and no order backlog for finished humanoids to point at — because the market is pre-launch. The only "tightness" signal that exists now is in the upstream parts (long lead times, concentrated share). When could it flip to oversupply? Only well after the demand turn-on, once part-makers and platform assembly both catch up — plausibly years after first mass deployment, given the multi-year capacity-build times above. In the nearer term the structural risk runs the other way: the brain may take longer than hoped, in which case demand simply does not arrive and there is no gap to capture. Which path occurs is unknown; the timing of AI capability is the single variable that decides it.
Source: synthesis of 500-stocks scan Sectors 01/07/11. Response-time figures are scan estimates, not contracted, and not live-verified.
One way to read this group: one platform bet (TSLA) and a set of component/automation industrials (ABB, ROK, plus the upstream specialists) who get paid no matter which platform wins. The sizes below are rough orders of magnitude from general knowledge, not live quotes.
| Company (ticker) | What it makes | Exposure to humanoids | Rough size est. | Position / scan notes |
|---|---|---|---|---|
| Tesla (TSLA) | EVs, energy storage, FSD, Optimus humanoid | Most leveraged to the thesis, but ~0% of revenue today (paper) | ~$1T+ mkt cap est. | Per scan, the player combining mass-manufacturing scale with frontier AI talent; vertically integrating actuators |
| ABB (ABB) | Industrial robot arms, motion/drives, electrification | Small slice; humanoid upside on a profitable base | ~$100B+ mkt cap est. | Large global industrial-robot maker; broad automation portfolio; sells the "muscles" to all comers |
| Rockwell (ROK) | Industrial automation, motion control, servo/drives | Small slice; component-supplier exposure, not a platform | ~$30-40B mkt cap est. | Leading US motion-control name; supplies components to any robot volume |
| Hyundai / Boston Dynamics (HYN) | Autos + advanced legged/humanoid robots | Real platform, tiny % of a large auto parent | Large auto cap est. | Per scan, most advanced locomotion; humanoid is a small fraction of group revenue |
| Agility Robotics (AGIIR) | Digit warehouse humanoid | Pure play (paper) | Private / IPO-pending | Closest thing to a pure humanoid public vehicle if/when it lists |
| Upstream parts: Nabtesco (NBTCY), Harmonic Drive, RBC, TKR, NIDEY | Gears, reducers, bearings, motors | Indirect but structurally tightest link | Mixed (mid-cap to large) est. | Concentrated control of the gating components; paid regardless of platform winner |
Source: 500-stocks scan Sectors 01/07/11 for who-makes-what and positioning; market-cap magnitudes are approximate general-knowledge ranges, not live-verified.
The plain-money question: for each name, how much of what you pay today is actually humanoid, and what is the cash shape of the business you are buying? Translating to "dollars of market value per dollar of this-year revenue" (a price-to-sales sense check) and to who generates owner cash today:
Money-in / money-out shape of the group: it splits cleanly. The platform side (TSLA's humanoid line, pure plays) is capital-hungry and pre-cash — you fund the build now and the return is a forecast. The component/automation side (ABB, ROK, parts) is capital-light relative to that bet and already throws off owner cash. So an owner buying "the group" is blending a venture-style call option (a bet that pays only if a future event happens) with a set of profitable industrials. The split, the lead times, and the cash-generation facts are stated above so the reader can weigh them; this sheet reaches no conclusion on which is the better use of capital — that turns entirely on one's confidence in AGI timing.
Source: structure from 500-stocks scan; cash-shape characterizations are neutral arithmetic on general-knowledge magnitudes, not live multiples, and not live-verified.
Where a company-level deep-dive would teach you the most, by role in the group (a factual pointer to where information is, not a recommendation):
Source: 500-stocks scan Module 8 Cross-Cutting Notes (component-supplier layer vs platform bets; TSLA as primary listed platform vehicle).
What was used:
Hard vs approximate — stated plainly:
Source: as listed above. This document contains no buy/sell recommendation, price target, or valuation verdict — only facts and neutral arithmetic for the reader to judge.