The part nobody benchmarks

A humanoid robot carries 25 to 35 powered joints, and every one of them is built around the same recipe: an electric motor, a transmission that trades its speed for torque, and sensing to close the loop. The transmission war – harmonic versus planetary versus roller screw – is fought loudly and in public. The motor war is not. Spec sheets for complete robots almost never name the motor inside, and yet motor selection decides the joint’s mass, its heat, its controllability and a meaningful slice of its cost.

Two motor families do nearly all the work. Frameless torque motors handle the body – hips, knees, shoulders, everything that moves mass. Coreless motors (hollow-cup motors, in the Chinese industry’s phrase) handle the hands. Tesla’s Optimus Gen 2 is the cleanest public example of the split: 28 frameless motors across its 14 linear and 14 rotary body actuators, and six coreless motors in each hand for finger-level control. One robot, two electromagnetic philosophies.

What “frameless” buys you

A frameless torque motor is a brushless permanent-magnet synchronous motor sold as two loose parts: a stator ring of copper windings and a rotor ring of magnets. No shaft, no bearings, no housing, no end caps, no encoder. The robot maker supplies all of that – the joint’s own structure becomes the motor’s frame, the joint’s bearing becomes the motor’s bearing, and the reducer bolts straight onto the rotor. The idea is old: Inland Motor Company introduced frameless motors in 1949 for missile guidance gimbals. What’s new is a customer that wants 28 of them per unit at automotive volumes.

Why go through the trouble of integrating a motor yourself? Three reasons keep coming up in supplier literature.

• Packaging. Deleting the housing and shaft removes dead mass and lets the motor share bearings with the joint. Kollmorgen’s humanoid-focused TBM2G frameless line comes in seven frame sizes with three stack lengths each, sized to disappear inside a joint.
• The hollow shaft. A large through-bore lets cables, encoder lines and even mechanical shafts pass through the middle of the motor – which is how a humanoid arm routes its wiring through its own joints instead of around them. TQ’s ILM-E series and Kollmorgen’s TBM2G both lead with this.
• Torque at low speed. Torque motors are wound with high pole counts for large torque at walking-pace joint speeds, with low torque ripple – which matters when the “gearbox” is a quasi-direct-drive stage of only 5:1 to 10:1 and every ripple reaches the limb.

The 48-volt detail is easy to miss. Industrial frameless motors were wound for 300–600 V servo drives; a humanoid runs on a battery bus at 48 V or below, so humanoid-grade windings are rewound fat and short to make torque at low voltage. It’s one of the quiet ways “robot-ready” motor lines differ from their industrial parents – and one of the reasons robot makers can’t simply lift parts from the factory-automation catalogue.

The motor is not the actuator

Here’s the number that surprises people. A humanoid-grade frameless motor delivers roughly 2 to 5 newton-metres per kilogram. A finished humanoid joint actuator advertises 80 to 160. Both figures are honest – the gap is the reducer, which multiplies the motor’s torque thirty- or hundred-fold while adding less mass than it saves in copper and magnets. CubeMars’ AKH70-48 hollow-shaft joint module makes the point: a motor plus 48:1 planetary stage weighing 1.396 kg that peaks at 222 Nm, or 159 Nm/kg for the assembly. The bare motor inside it could never get near that.

So when a robot maker chooses a motor, the real question is what the transmission will forgive. A high-ratio harmonic joint can use a modest motor and hide its ripple; a quasi-direct-drive joint with a 6:1 or 9:1 stage needs a fat, high-pole-count torque motor because there’s almost nothing between the windings and the world. That’s also why bare-motor torque figures scale so steeply with diameter. Torque grows with the square of the moment arm – Kollmorgen’s engineers call it the D2L rule – so a wider ring beats a longer one, and humanoid joints keep trending toward large-bore pancake shapes.

Radial, axial and what comes next

Nearly every frameless motor shipping in a humanoid today is radial flux – the magnetic field crosses the air gap sideways, from a magnet ring to a winding ring, exactly as in the diagram above. It’s the production benchmark for a boring reason: a century of manufacturing maturity.

Axial flux is the challenger. Turn the geometry ninety degrees – flat disc rotor, flat disc stator, field running along the axis – and the motor gets 20 to 40 percent shorter for the same torque. Patent-landscape analysis puts axial-flux torque density at 15 to 25 Nm/kg against roughly 10 for comparable radial machines, and argues a disc motor mates more naturally with a pancake harmonic or cycloidal reducer, cutting total actuator weight by as much as 40 to 50 percent. Treat those figures as the optimist’s case – they come from vendors and patent filings, not from teardowns of shipping robots – but the direction of travel is real, and specialist suppliers such as China’s SEMOTOR now market axial-flux joint motors specifically at humanoid makers. Transverse-flux machines promise even more torque at low speed and remain, for now, a research topic.

The hands run on a different motor

None of the above fits in a finger. Hand actuation belongs to the coreless motor: a slotless design whose winding is a self-supporting copper cup rotating around a stationary core, with no iron teeth for the magnets to grab. Zero cogging, almost no rotor inertia, and smooth torque at tiny scales – exactly what tendon-driven fingers need. Each Optimus Gen 2 hand carries six of them.

The economics are brutal and getting more so. Swiss and German precision houses – Maxon, Faulhaber – built this category for surgical tools and aerospace, at ¥4,000-plus per motor. Chinese suppliers now sell humanoid-grade coreless motors from ¥1,200–2,300 (Moons’ Industries) down to around ¥100 (Aoyi Technology). A hundred-fold price spread for the same nominal function is not a market at equilibrium; it’s a supply chain in the middle of being rebuilt, the same 70-plus-percent cost collapse that already hit reducers and screws arriving at the fingertip. The catch, per the engineering literature: coreless motors can fail to start under winding faults, and quality separation between a ¥100 and a ¥4,000 unit shows up in lifetime, not in the datasheet.

Who supplies the muscle

The supplier map looks like the reducer map did three years ago: Western and Japanese incumbents holding the performance high ground, Chinese entrants underneath them at a third to half the price, and integrated-actuator makers packaging motor, gear and drive into one bolt-on module. TQ-RoboDrive carries the deepest research pedigree – its technology comes out of the German Aerospace Center (DLR), and PAL Robotics builds its TALOS and TIAGo platforms around RoboDrive motors. Kollmorgen aims its TBM2G line squarely at 48 V humanoid joints. On the Chinese side, Inovance, Step Electric and Dongjie Zhikong lead frameless supply, with Moons’, Aoyi and Topband contesting coreless hands.

For buyers the practical read is this: if you’re evaluating a humanoid, ask what motor sits in the knee and whether it was co-designed with the reducer, because that answer predicts thermal endurance better than the peak-torque number on the brochure. If you’re a component maker, the frameless category is where 28-motors-per-robot volume meets a part that still has no dominant standard – and the coreless hand motor is where the next price war has already started.

FAQ

Sources