China is moving humanoid robots from stage demos to assembly lines, and UBTECH’s newly commissioned plant in Liuzhou is the clearest signal yet that “robots building robots” is not a slogan but a manufacturing strategy with scale behind it.
The Short Version
- UBTECH has commissioned a purpose-built humanoid robot factory in Liuzhou, designed for annual capacity above 10,000 units and mixed-model production.
- The site integrates automated logistics and digital management across assembly, testing, warehousing, and shipping, with a stated takt of one unit every 10 minutes.
- Production centers on the Walker S and Cruzr series, aligning with local policy to make Liuzhou a benchmark base for high-end robot manufacturing.
- The build-out fits a broader Chinese push: a deep robot supply chain, the world’s largest installed base of industrial robots, and multiple humanoid mass-production lines coming online.
What UBTECH actually built in Liuzhou
UBTECH’s Liuzhou “super smart factory” is presented as a plant configured for high-volume, mixed-model humanoid assembly rather than a pilot cell. Reporting describes a 14,000-square-meter footprint with a 13.8‑meter clear height, enabling vertical integration of subassembly, end-of-line testing, and intralogistics. The facility is tied to named product families—Walker S (including Walker S2) and Cruzr (including S2 and Y1 variants)—which matters, because general claims about “humanoids at scale” only become credible when the line is grounded in concrete SKUs with known bills of materials and test protocols.
The factory’s control layer is described as “digital management” spanning manufacturing execution through warehousing and shipment. In practical terms, that usually means a manufacturing execution system (MES) coordinating stations, traceability of critical components—actuators, reduction gears, sensor suites—and automated guided vehicles (AGVs) or autonomous mobile robots (AMRs) closing the intralogistics loop. UBTECH and local reporting repeatedly cite a takt target of one industrial humanoid every 10 minutes—an aggressive but legible goal if the line runs multiple shifts and if key bottlenecks (joint assembly, end-of-line calibration, and burn-in) have parallelization designed in.
How “robots building robots” works on the ground
Humanoid production blends automotive-style station discipline with the idiosyncrasies of mechatronic alignment. The core constraint is not fastening shells; it is the precision assembly of actuators and joints, routing of power and signal harnesses, and calibration of perception and control subsystems. A line that claims a 10-minute takt typically decouples rate-determining steps—such as harmonic drive preloading, torque sensor calibration, and IMU/radar fusion checks—into dedicated test stands. Intralogistics robots then ferry subassemblies to final integration where software flashing, safety interlocks, and mobility tests occur before aging tests (hours of gait and manipulation in controlled conditions) validate the unit for shipment.
The Liuzhou site is described as running a closed intralogistics loop—robots moving parts to build other robots. That approach is not theater; it raises utilization of high-cost human labor by offloading non-value-added transport and staging. Done well, it also tightens traceability: material handling events become data points, feeding process control and enabling faster root-cause analysis when a defect appears downstream. This is why the claim of end-to-end digital management is as important as the takt figure; throughput is only one half of manufacturability, while yield and rework rate are the other half that determine whether “10,000 per year” is viable or punitive on cost.
Why Liuzhou, and why now
Locating this factory in Liuzhou is not incidental. Local and national media frame it as a benchmark for the city’s move into high-end robotics manufacturing—an industrial-policy moment that aims to position Guangxi alongside hubs like Shenzhen, Foshan, and Shanghai in the humanoid build-out. The policy logic is clear: humanoids require dense component supply—precision gearboxes, lithium packs, motor windings, cameras, and compute modules—plus the ability to iterate fixtures and test rigs quickly. China’s robot ecosystem already excels at those prerequisites. BBC reporting has documented the country’s unmatched density of industrial robots, an export engine for robot hardware, and a supply chain where parts can be sourced and iterated within hours in regions like Shenzhen.
That base capacity matters because humanoids are not yet commodity goods; they are evolving platforms whose manufacturability improves with each cohort. China’s track record in scaling industrial robots—both installed base and production share—supplies a skilled labor pool and vendor base comfortable with rapid, incremental process improvements. It also supplies demand: automotive, electronics, and logistics sectors already structured around robotized workflows can absorb early humanoid deployments where form factor solves reach, dexterity, or mobility gaps left by traditional arms.
What the factory is claimed to deliver
The Liuzhou plant is positioned as a “10,000-unit-level” humanoid factory that can produce one industrial humanoid every 10 minutes, with robots building robots in a digitally orchestrated flow. The products are named—Walker S and Cruzr series—and the building is physically substantial, with mixed-model lines designed into the layout. Local officials cast the commissioning as the moment Liuzhou “enters” large-scale robot manufacturing, a signal that matters as much for capital and supplier alignment as for immediate output.
The broader context supports the claim’s feasibility rather than undermining it. China leads the world in industrial robot deployment and manufacturing. That base underwrites both the workforce that can run such a site and the vendors who can feed it—motors, drives, batteries, and sensing modules—without long-lead imports. BBC’s reporting underscores the strength of China’s robot hardware advantage and the speed at which factories can operationalize component changes and new fixtures, which is the practical enabler behind a 10‑minute takt target.
How this fits the global humanoid race
UBTECH’s Liuzhou line is not an isolated bet; it is one among several Chinese humanoid production pushes over the past two years. A consistent pattern emerges: capacity announcements framed as “world firsts,” factories configured for five- to ten-thousand-unit annual rates, and visible attempts to close the loop from assembly to data collection and software improvement. Independent of ultimate market size, the manufacturing tactician’s view is straightforward: the country with the fastest iteration cycle—across hardware, fixtures, and validation—will compound learning faster and compress costs sooner. That is the game China is structurally advantaged to play, given its component density and policy focus.
For global manufacturers and policymakers, two implications follow. First, supply-chain optionality for critical mechatronic components will increasingly hinge on Chinese vendors unless alternative ecosystems invest at comparable depth; otherwise, even “domestic” humanoid lines elsewhere will carry Chinese content. Second, integration talent—engineers who can move between robot design, process engineering, and software tooling—will determine which regions can translate pilot cells into reliable, high-yield lines. The Liuzhou site’s emphasis on digital orchestration and intralogistics is a tacit acknowledgment that humanoids are as much a data product as a mechanical one.
The durable takeaway
UBTECH has commissioned a physically credible, purpose-built humanoid robot factory in Liuzhou, configured for mixed-model production of named platforms and run on an end-to-end digital backbone. The headline figures—a 10,000‑unit annual capacity level and a 10‑minute takt—sit within China’s broader robot manufacturing capability and policy aims, not apart from them. Read the plant as an inflection in how humanoids are built: less spectacle, more station discipline; less one-off demo, more repeatable test-and-ship. That is what it takes to turn a humanoid from a research mascot into an industrial product.
Sources:
insiderpaper.com, pandaily.com, en.people.cn, x.com, interestingengineering.com, kantan.news, itiger.com
