Humanoid robots in China are finally shifting from eye-catching demos to machines built for real warehouse work.
At the World Robot Conference in Beijing, UBTECH ran a live demo where about ten humanoid robots worked together, lifting, stacking and sorting along an assembly line. The scene was staged to show these machines doing logistics-style tasks, not just performing on stage.
Reports presented at the conference back that claim. They say China shipped more than 40,000 humanoid units in the first half of 2026, roughly 97% of the global total, while global shipments exceeded 22,000 in the same period and are projected to top 50,000 for the year.
Why it matters: scale changes what teams can do. When robots move from prototypes to units you can buy and pilot, companies can run their own experiments instead of waiting for bespoke projects or renting access. That lowers the barrier to trying automation in real operations.
How the demos work, in plain terms: each robot handles a narrow task on the line, like a single worker in a crew. Think of a new kind of factory team where sensors and decision software let machines sense items, pick or stack them, then pass work to the next robot. That combination of body and software is often called embodied AI, meaning the intelligence is built to act in the physical world.
Practical reality: big shipment numbers do not erase key limits. Reporters and analysts at the conference flagged gaps in capability, robustness and reliability. Most warehouses are not ready to replace human teams this week; these robots still need more testing before running unsupervised in real industrial workflows.
What changes now is economic: Chinese firms are supplying units at scale, giving companies a lower-cost, widely available option for pilots. If enough early customers accept the current tradeoffs, incremental automation in logistics could speed up, especially for repetitive tasks that tolerate occasional errors.
The open question is whether these machines can prove dependable enough to save money over time. If early pilots show real productivity gains, adoption could accelerate quickly. We’ll likely know within months whether this is a steady march or a sudden industry shift.