BYD has unveiled the design of its first self-developed humanoid robot, with 31 joints, according to Interesting Engineering. BYD's entry adds another automaker to an already crowded field, and it belongs on our humanoid comparison page. A design reveal is not a verdict.
A joint count and a reveal video describe a design. They do not describe a product that ships, works reliably outside a controlled demo, or can be built in volume. That distinction matters this week, as several reports converged on the same point: the humanoid industry's hardest problems sit in the factory and in the hand, not in the keynote.
Manufacturing at scale is the next wall
As humanoid developers try to move their machines from demonstrations and pilots into factories and other working environments, manufacturing at scale is emerging as the next major challenge, according to Robotics & Automation News. That is a different problem from building a working prototype. A company can demonstrate a robot performing a task convincingly and still be years away from manufacturing at the volume, consistency and cost a real product requires.
That gap matters to anyone pricing a purchase decision on our price tracker. A design reveal says almost nothing about when a unit ships or what it will cost at volume. BYD's car-manufacturing base could be an advantage, since automotive production is itself a scale-manufacturing discipline. But that advantage is speculative until BYD demonstrates it with robots, not cars.
Demos still fail the generalization test
Humanoid robot demos still fail what The Robot Report calls the generalization test: a robot that performs well on a scripted task in a controlled setting often cannot handle the same task with minor variations, in a different environment, or without heavy preparation. The Robot Report says better data and human judgment are needed to help robots move beyond demos toward reliable real-world performance.
That is the gap between a viral clip and a deployable tool. Compare Robots treats polished demo footage as the least informative part of any humanoid robot's story: a robot that works once, on camera, under conditions its own manufacturer chose, tells you little about a robot you can actually deploy. Our best humanoid robots ranking weighs that generalization gap rather than demo polish.
Dexterity is the field's real bottleneck
The Robot Report frames dexterity as physical AI's real bottleneck, with the next major challenge being making the entire robotic system reliable enough to turn capabilities into useful work. Grasping, manipulating and handling varied objects reliably, at the speed and consistency a factory or warehouse needs, remains unsolved even at the field's most capable labs.
Boston Dynamics' own progress illustrates the point. The company gave more detail on its redesigned humanoid hand this week, discussed by Alberto Rodriguez, director of robot behavior for Atlas, according to The Robot Report. That a core piece of hardware, the hand, is still being redesigned on the Atlas program says something useful: dexterity is not a solved problem that newer entrants simply need to catch up on. It is an open problem the whole field is still working through.
Form should follow workflow
A Robot Report commentary raises a more structural question: for physical AI to move from social media demos to sustained commercial use, robotic form should follow workflow. Many current designs start from a general-purpose humanoid shape and then look for tasks to fit it, rather than starting from the specific workflow a customer needs automated and building the robot around that. The commentary asks whether that order needs to reverse.
That cuts against the assumption that a humanoid, general-purpose body is automatically the most commercially useful form factor. It is a genuinely contested point, and this week's reporting does not resolve it. It is also a useful lens for BYD's entry: the question is whether the robot is built around a real deployment workflow or around a general-purpose design in search of one.
What this means for buyers and investors
None of this makes BYD's entry unimportant. A manufacturer with BYD's scale entering the humanoid race is worth tracking, and it could sharpen competition on cost if BYD moves from design reveal to production. But the right test for a 31-joint design announcement is the same one this week's reporting applies to the whole field: how far the hardware is from scaled, consistent manufacturing; whether any demo generalizes beyond a scripted run; and whether dexterity, the bottleneck even Boston Dynamics is still iterating on, is actually solved or only shown once.
BYD's robot is a research preview, not a product. So is most of the field. Readers can track how these questions resolve, robot by robot, as new specs and prices are published.