Frameless Motors

Frameless torque motors are the body-motor spine, one per body joint, about twenty-eight per robot¹, and they are the genuine binder of their batch. The winding line itself is elastic: dozens of makers build frameless torque motors, the stator-and-rotor assembly is a low barrier, and a new line stands up in well under a year, ALVA scaled from one thousand to fifteen thousand motors a year, roughly fifteen-fold, in a single year¹⁰. So robot-grade output can ramp hard, from about six million motors a year today³ toward a modeled hundred and ninety million a year by 2040. And yet it still crosses. Because every robot keeps its twenty-eight motors for life¹, the motors the worldwide humanoid fleet¹⁴ needs accumulate relentlessly: by 2040 a fleet of roughly forty-seven million robots needs about one and a third billion frameless motors⁸, and even the elastic winding ramp delivers only about ninety-five percent of that. The global crossover, where cumulative production first falls behind cumulative need, lands right around 2040. Building from a single bloc is worse: China-only output, the fastest-ramping bloc, falls behind around 2036, and the thin US aerospace-and-defense niche falls behind almost immediately, around 2028. And the deeper truth is that even this picture is optimistic, because the winding line is not the real ceiling: every frameless motor needs high-coercivity NdFeB¹¹, and that magnet supply, China-monopolized and export-gated¹², is the constraint that actually caps the spine.

A frameless (or slotless) torque motor is a brushless motor sold as just its two active rings, a wound stator and a permanent-magnet rotor, with no housing, bearings or shaft, so the robot’s joint structure becomes the motor frame. That makes it compact, light and high-torque at low speed, exactly what a humanoid joint needs. Tesla’s Optimus uses one frameless torque motor in every body actuator: fourteen rotary joints (frameless motor plus a harmonic reducer) and fourteen linear joints (frameless motor plus a planetary roller screw), so the model uses about twenty-eight motors per robot¹². That is the spine of the machine, these body torque motors are distinct from the small coreless hollow-cup motors in the dexterous hands. The robot-grade frameless-torque tier is a real, distinct market: only about four hundred million dollars in 2024, growing at roughly fifteen percent a year⁴, with the humanoid-specific slice tinier still, about forty million dollars in 2024, but growing fastest of all, near forty-four percent a year⁶. It is not the billions-unit general brushless-DC universe it is often confused with⁹.

Because a frameless torque motor is built into a robot for life, the motors the fleet needs are a cumulative stock, the total number of robots ever built worldwide multiplied by twenty-eight¹, not a per-year flow. Integrating the consensus shipment ramp¹⁴, the worldwide installed fleet reaches about 585,000 robots by 2030, ten million by 2035 and forty-seven million by 2040, requiring roughly sixteen million, two hundred and eighty million, and one and a third billion frameless motors respectively⁸. That billion-plus figure is what makes this the spine: twenty-eight per robot is the single largest motor count in the machine. In the chart below, the black line is that worldwide cumulative need, the grey band is cumulative global production, and the red wedge is the shortfall that opens once need outruns supply.

Production is the only thing that differs between the Global, China and US views. The worldwide humanoid fleet and the motors it needs stay identical in all three; only the supply line changes, asking whether one bloc’s factories alone could equip the whole global fleet. This is the robot-grade frameless-torque motor-unit line, not the broad commodity brushless-DC total, and real output follows an S-curve. Global output runs from about six million motors a year today³ to roughly thirty-two million by 2030, a hundred million by 2035 and a modeled hundred and ninety million by 2040, climbing steeply as the humanoid pull draws the broad winding base onto robot-grade frameless lines⁶, the winding step is genuinely elastic, with sub-year line builds like ALVA’s fifteen-fold scale-up¹⁰. The post-2030 figures are a modeled extrapolation, a humanoid-driven step-change well above the roughly nine-to-fifteen-percent baseline torque-motor growth rate⁴⁵, not a reported forecast. China climbs fastest of all, from about three million motors today toward twenty million by 2030 and a hundred and twenty-five million by 2040: it dominates brushless-DC volume manufacturing and is scaling robot-grade frameless aggressively, Inovance alone is targeting two to three million frameless torque motors for humanoids⁷, though only a small group of Chinese makers can yet supply the consistent quality the joint demands². The US holds far flatter, from about 0.6 million motors to perhaps twelve million, a low-volume, high-mix niche (Allied Motion, Moog, Kollmorgen, Portescap, maxon US) serving aerospace, defense and medical¹³, where the Western footprint is high-spec but small in unit terms, with no announced humanoid-grade reshoring at scale.

The crossover, the year cumulative production drops below cumulative need¹⁴, lands right around 2040 globally. That is later than the grinding-gated precision parts, which bind early in the 2030s, because frameless winding is elastic enough to track the fleet through most of the decade. But it still crosses, and that is the point: the durable stock is relentless, the fleet is adding close to ten million robots a year by 2040 and each one demands twenty-eight motors for life¹, so the need climbs toward one and a third billion motors while even a hundred-and-ninety-million-a-year winding base cannot quite follow. By 2040 cumulative global output meets only about ninety-five percent of the fleet’s need⁸, and the gap widens fast afterward as the fleet climbs toward hundreds of millions of robots. Building from a single bloc brings the wall forward. China-only output crosses around 2036, it expands fastest of any region but cannot alone equip the entire global fleet⁷, and US-only output crosses almost immediately, around 2028, since a thin aerospace-and-defense niche¹³ never keeps up with even the early fleet. Frameless is therefore the genuine binder of its batch: not because the winding line is rigid, but because twenty-eight durable motors per robot is simply an enormous, never-decremented demand.

Cobots share the frameless-motor pool: every collaborative-arm joint is a frameless permanent-magnet torque motor, six per six-axis arm versus a humanoid’s twenty-eight, so the ~3.7-million-strong 2040 cobot fleet adds about 22 million motors to the humanoids’ ~1.3 billion need. Quadrupeds draw on it harder still per unit: a four-legged robot’s twelve quasi-direct-drive joints are each a large-diameter frameless BLDC torque motor (the MIT Mini-Cheetah / Unitree build), so the ~5-million-strong quadruped fleet adds about 62 million motors, more than cobots, though the two together are still a rounding error against the humanoids’ ~1.3 billion. Traditional industrial arms, by contrast, do not draw on this link at all: they use framed AC servo motors bolted to a separate gearbox, a different product class from the frameless torque-motor kits, so their ~6-million fleet adds nothing here (their NdFeB appears on the magnet link instead). Drones do not draw on it either, which is the striking one given they are the largest fleet in the report: a drone’s four-to-eight motors are high-KV outrunner BLDC coupled straight to the props, a hobby/UAV-propulsion product class (T-Motor, DJI in-house) with no supplier overlap with robot torque-motor kits, so like industrial arms they pull the shared NdFeB feedstock but never this link. Frameless motors are the last binder to cross (about 2040), and cobots and quadrupeds leave that timing essentially unchanged, the same TQ-RoboDrive ILM and Kollmorgen KBM class of kits serves cobot, quadruped and humanoid joints alike.

The honest depth-chain root is not the winding line at all, it is the magnet. Every frameless torque motor needs a high-coercivity NdFeB permanent-magnet rotor, and a humanoid carries roughly three and a half to four kilograms of that magnet material across its motors¹¹. China refines and separates about ninety percent of the world’s rare-earth magnet feedstock¹¹, and the constraint is not hypothetical: Tesla has confirmed that Optimus production ‘was impacted’ by China’s rare-earth magnet export controls¹². So while the motor-unit winding base is elastic, dozens of makers, low assembly barrier, sub-year line builds¹⁰, the real ceiling on how many frameless motors can actually be built is inherited from NdFeB supply, which is itself gated by neodymium-praseodymium separation and mining far upstream. To avoid double-counting that rare-earth tonnage, this component models only the motor-unit curve the assemblers can wind; the NdFeB mass and its own China-concentration and feedstock crossover are carried by the separate magnets component, which is where the depth-chain bottoms out. Read the two together: frameless motors cross around 2040 on the durable-stock arithmetic alone, and the magnet root means the true squeeze arrives sooner and harder than the winding-only curve suggests¹². That is why the body-motor spine binds, the line that builds the motor is elastic, but the magnet inside it is not.