Chinese robots can do backflips, but industrial customers are buying something else entirely

Sanjiv Shah, who works on industrial robotics deployment, draws a sharp line between hardware performance and commercial relevance. Chinese robots, he says, “can walk beautifully. Great engineering. Love it. Do back flips.” His customers, he adds, do not compare that hardware to what his company offers, because the comparison is not between two robot platforms. It is between a robot platform and a full integrated solution.

That distinction matters more than it might appear. The argument that hardware is the moat in robotics is intuitive: manufacturing scale is hard to replicate, precision engineering takes years to accumulate, and cost structures in Chinese manufacturing are genuinely difficult to match. But Shah’s framing, and the evidence behind it, suggests the moat argument gets the competitive terrain wrong. Industrial buyers are not purchasing a walking machine. They are purchasing an autonomous inspection workflow, data they can trust, and integration with their existing systems. On those dimensions, impressive locomotion is close to irrelevant.

The safety constraint Caitlin Kalinowski describes makes that concrete. Current commercial Chinese humanoid robots, she notes, carry a booklet requirement that no human can be within three feet of the operating robot. That single constraint forecloses the deployment scenarios that make humanoid robotics economically interesting: working alongside assembly line workers, operating in shared spaces, integrating into facilities designed around human movement. A robot that requires a cleared perimeter is not a collaborator. It is an obstacle to the workflows it was meant to support.

Data sovereignty adds a second, structural barrier. Shah frames it plainly: industrial customers do not want cameras they do not control inside critical infrastructure. This is not a price objection. It is a trust objection, and it operates independently of how capable the underlying hardware is. A robot from a vendor a customer cannot fully audit is, from a security architecture standpoint, a liability placed at the center of sensitive operations. The software gap reinforces both problems. Hardware that outpaces its own control and reasoning systems creates a ceiling that locomotion quality cannot lift.

Ramin Hasani, who leads Liquid AI, offers a different angle on the same underlying difficulty. His firm made a deliberate decision to pause robotics entirely. The reason he gives is not hardware competition from China. It is the go-to-market complexity of production-grade robotics itself: the verification and validation requirements, the regulatory overhead for deployment, the gap between a working demonstration and a commercially viable product. The difficulty of building a robotics business, in his telling, is sufficient to set the opportunity aside regardless of what the hardware can do. His reasoning is a useful corrective to any assumption that the challenge is primarily technical.

Taken together, these observations point to a structural mismatch between what the robotics hardware race is producing and what industrial deployment actually demands. The competition that matters for industrial customers is not which robot walks best. It is which vendor can deliver autonomy that works reliably in messy real-world conditions, keep proprietary operational data inside the customer’s control, integrate with existing workflows without requiring special exclusion zones, and clear the verification bar for production use. Chinese manufacturers have built a strong position on the hardware dimension of that list. The remaining dimensions are where the viable businesses will be built, and hardware scale does not transfer to them.