There is a certain science-fiction ring to the phrase "robots building robots," but in Liuzhou, an industrial city in southern China's Guangxi Zhuang Autonomous Region, it is now a description of an ordinary production shift. On September 12, 2026, Chinese robotics company UBTECH opened what it calls the world's first humanoid-robot manufacturing plant built for a ten-thousand-unit annual capacity. At its designed pace, one industrial humanoid robot leaves the line every ten minutes.
Whether the world actually wants ten thousand humanoid robots a year is the open question. But there is a more immediate story worth understanding: China's humanoid-robot industry has crossed from hand-finished prototypes into something that looks a lot like genuine mass production — and the speed of that move has caught even close observers off guard.
What the New Factory Actually Is
Let us start with the confirmed facts, drawn from local government releases, the Guangxi Daily, and company disclosures.
The Liuzhou "super smart factory" covers about 14,000 square meters of floor space, with a ceiling height of 13.8 meters. It builds UBTECH's Walker S series and Cruzr series — embodied-AI industrial humanoids designed to work alongside people rather than entertain them. The headline numbers:
- A designed production beat of one robot completed every ten minutes
- A planned annual capacity of more than 10,000 units
- A digital "smart brain" that schedules the whole plant, using industrial simulation to plan workflows
The "robots building robots" part is literal. UBTECH has placed several of its own humanoid models into the production process. In the materials-handling section, Cruzr Y1 and Cruzr S2 robots take on de-palletizing and palletizing — loading and unloading stacks of parts. For storage, UBTECH built what it says is the industry's first fully automated humanoid-robot warehouse, a compact 65-square-meter structure that can hold 112 robots and, by the company's account, lifts storage-space utilization by more than half.
UBTECH did not build the digital backbone alone. It worked with the industrial-software arm of Siemens to create a customized manufacturing system meant to solve a specific set of problems: humanoid robots have a very large number of components, require frequent movement of parts between stations, demand high assembly precision, and are redesigned quickly. A rigid old-style car line would choke on that combination; a flexible, simulation-driven plant is built around it.
How Fast This Actually Happened
The compressed timeline is the part that says the most about how China's robotics industry now moves.
UBTECH first set up in Liuzhou in 2022. In June 2025, the first industrial humanoid robot built in Guangxi came off the line, and the company says it reached "thousand-unit production" that same year. By September 12, 2026, it had opened a plant sized for ten thousand. The company's own summary of the arc runs from "first unit breakthrough" to "thousand-unit mass production" to "ten-thousand-unit capacity" in a little over a year.
The sales figures, from UBTECH's 2026 half-year report, point in the same direction but deserve careful reading. The company reported humanoid-robot sales of 16,123 units in the first half of 2026, up 268.3 percent year over year. It also said it had recently won overseas orders worth more than 50 million yuan, with what it described as breakthroughs in Europe, Japan, and South Korea. Those are company-reported commercial figures, not independently audited results, and the precise breakdown between paying customers, pilot deployments, and internal use is not public — so they should be read as evidence of rapid scaling, not of proven profitability.
China Is Not the Only Runner
To understand what is distinctive here, compare the most-watched Western rival: Tesla and its Optimus humanoid.
According to reporting from September 2025 — since updated by outlets including Cailian Press and Jiemian — Tesla raised Optimus output roughly tenfold within a few months, going from dozens of units per week in the second quarter to several hundred a week by late September. That is real progress, but the company is still short of its goal of producing a thousand robots a week by the end of the year. The bottleneck, people familiar with the project said, is the robot's complex hand, along with limits at suppliers and breakdowns in automated equipment. Early customer machines may be leased rather than sold, so Tesla can take them back, upgrade the hardware, and keep collecting real-world operating data.
This frames the UBTECH milestone honestly. China's advantage in humanoid robots is not, for the moment, that its machines are clearly more intelligent or dexterous than the best Western designs. It is that the entire component chain — motors, reducers, sensors, batteries, controllers — sits close at hand in a mature manufacturing ecosystem, alongside a home market full of factories willing to test a robot on a real shift. A company can design, source, build, and iterate quickly without waiting on parts from another continent. The Xataka tech outlet, covering the Liuzhou opening, put it plainly: China's edge is less the technology itself than having all the necessary pieces in one place.
The Rest of the Field Is Scaling Too
UBTECH is the clearest factory story of the month, but it is not alone — which is itself part of the picture.
On September 24, rival Chinese firm AgiBot (智元) delivered its 20,000th general-purpose embodied robot, a Yuanzheng A3 Ultra, to the Chimelong Group. The climb from 10,000 to 20,000 units took about six months. The same day, more than 300 AgiBot robots began working at what the partners call the world's first large-scale embodied-AI theme park, at Chimelong's spaceship park in Hengqin — guiding visitors, performing, handling retail and hotel tasks, and running science activities. The International Federation of Robotics added a broader data point: in 2025, China installed 354,000 industrial robots, about 59 percent of the world total, and it expects the Chinese industrial-robot market to keep growing 5 to 10 percent a year through 2029.
That context matters. UBTECH is not a single company hitting an isolated target; it is one of several Chinese firms scaling at once, inside the world's largest and densest robotics market.
The Honest Caveats
No account of this factory is complete without the questions that decide whether it turns into a durable business.
Capacity is not demand. A plant designed to make more than ten thousand robots a year is a statement about what UBTECH can build, not proof that ten thousand buyers will show up. Humanoid robots are still expensive, and for many factory jobs a simpler, purpose-built machine or an ordinary robotic arm does the work more cheaply and reliably. The industry's central risk is building the supply before the paying market fully arrives.
The hard problem is physical precision, not factory space. AgiBot founder Wang Xingxing, speaking at the Global Digital Trade Expo in Hangzhou, argued that the field's biggest remaining problem is the few-millimeter gap between what an AI model intends and what a robot actually does in the physical world. Solve that alignment error, he said, and the core problem is largely settled. A giant factory can scale assembly; it cannot by itself close that gap.
Many deployments are still pilots. Robots on a factory line or in a theme park are real working machines, but a good number remain at the stage of trials, demonstrations, or lease arrangements rather than steady, profitable deployments. The line between "deployed" and "being tested in public" is not always clear from the outside, and the large unit counts should not all be read the same way.
The economics are unproven. Neither UBTECH's humanoid business nor Tesla's Optimus has yet demonstrated that humanoid robots are profitable at scale. Rapidly rising output can just as easily mean rapidly rising costs until a robot reliably replaces enough human labor that a customer is glad to pay the price.
What to Take Away
Strip away the science-fiction framing, and the Liuzhou opening is a genuine marker. A humanoid robot now leaves a Chinese production line every ten minutes, assembled partly by other humanoid robots, inside a ten-thousand-unit plant built with a global industrial-software partner. A product that, a few years ago, existed mainly as a hand-built showcase is now being made with the rhythms, scheduling, and language of real manufacturing.
That is an achievement in scale and organization, not yet proof of a market. The questions that follow are the harder ones: whether enough factories, warehouses, and service businesses will pay for these machines; whether the few-millimeter alignment problem gets solved; and whether the unit economics ever turn positive. China has built the production capacity faster than anyone expected. Whether it has built it ahead of the demand — or exactly in time for it — is the answer the next few years will deliver.