Written by David McMahon
A design patent published by China’s intellectual property administration reveals BYD’s first self-developed humanoid: a 31-degree-of-freedom walking platform the company reportedly intends to put in front of customers at its car dealerships.
The clearest signal of where humanoid robotics is heading is not a launch event. It is a patent office publication. On 9 October, China’s National Intellectual Property Administration published a design patent for an embodied intelligent robot naming BYD as the applicant, revealing the automaker’s first self-developed humanoid.
The design describes a machine built to walk and maintain balance, with 31 degrees of freedom. Reported direction of travel points to customer-facing work at dealerships — greeting and assisting buyers — which would make BYD’s first commercial deployment a service application rather than a factory one.
Why automakers are well placed to build these machines is not a mystery. A humanoid is, to a first approximation, a manufacturing and cost problem wearing a software costume. It needs high-volume assembly, precision actuation, motor and battery expertise, semiconductor supply and a quality system capable of shipping thousands of identical units. BYD has all of that in-house, at a scale few robotics startups can approach. The company is also the obvious first customer for its own product, exactly as Tesla has positioned Optimus inside its own plants.
The strategic logic is a loop: deploy in your own factories, generate operating data and cost reductions, then sell to others. Automotive manufacturing is the closest existing analogue to what humanoid production demands, which is why the sector has become the field’s most credible supply base. XPeng commissioned humanoid production lines in September. Hyundai owns Boston Dynamics. BYD, the largest of them by volume, has now shown its hand.
The supply chain is forming around this. Schaeffler and Hexagon Robotics signed a partnership covering humanoid components including strain wave and planetary gears, alongside a commitment to deploy at least 1,000 Hexagon AEON humanoids across Schaeffler’s global factories by 2032.
That deployment timeline is the sobering part. Announced production capacity is running well ahead of units doing paid work. UBTECH’s Liuzhou facility is designed for more than 10,000 industrial humanoids a year, roughly one every ten minutes. Tesla has been reported as targeting tens of thousands of Optimus units in 2026. Against those figures, a thousand robots spread across a global factory network over seven years is a rounding error — and those robots spent time in a dedicated training gym in Germany being validated before entering production at all.
The gap is not a manufacturing problem. It is an integration problem, and it sits in the layer Brainlayer builds for. A humanoid arriving on a factory floor has to be taught the tasks, supervised while it learns them, and recovered when it fails — and the variables that decide whether that succeeds are human ones: whether an operator’s attention is on the cell, whether they are ready to intervene, how quickly the handoff is made and how the recovery is executed. Brainlayer’s BrainSim framework models exactly those quantities so that intervention and handoff can be predicted and improved rather than rediscovered on every shift.
For business readers, three things will separate the serious programmes from the announcements. Whether robots appear in the manufacturer’s own operations before they appear in a sales pitch. Whether the first commercial task is controlled and back-of-house or customer-facing. And whether the vendor can state an intervention rate — how often a human has to step in — because that number, not the degree-of-freedom count, determines the labour actually replaced.
