What this topic means

China installs more industrial robots each year than the rest of the world combined. It is also, at the level that matters most, still a robotics importer — because the components that determine whether a robot arm is precise, durable, and repeatable have largely been Japanese.

That tension defines Chinese robotics, and it produces a two-track strategy that this guide is organised around:

Track one: close the gap in industrial robotics by moving up from assembly into the high-value components — precision reducers, servo systems, and motion controllers — where Japanese and European suppliers have held durable advantages.

Track two: leap ahead in humanoids and embodied AI, a category where no incumbency exists yet, where China's electric vehicle supply chain supplies most of the necessary parts, and where being fifty years behind on industrial arm reliability is irrelevant because nobody is fifty years ahead.

The second track gets the attention. The first determines whether China becomes genuinely independent in robotics. They are best understood as a hedge against each other.

Why it matters

• Scale is not in dispute. China is the largest market, the fastest-growing installer, and increasingly a significant exporter. Whatever else is uncertain, the volume is real.
• The dependency is the interesting part. A Chinese-assembled robot arm has historically contained Japanese reducers and often Japanese servos. This is the same structural position Korea occupies in semiconductor materials — capable at the system level, dependent at the component level, and spending public money to fix it. (See Korea Semiconductor Ecosystem for the parallel.)
• Humanoids are the first robotics category where China may lead outright, and the reasons are structural rather than promotional.
• It is a direct test of the leapfrog thesis. If a new platform genuinely resets competitive position, humanoids are where that shows up. If incumbency in reliability, safety, and integration turns out to carry across, Japanese and European firms will absorb the new form factor and the moment will pass.

Industrial robotics: catching up from the outside in

For most of the last two decades, industrial robotics has been dominated by four firms — Fanuc and Yaskawa from Japan, ABB from Europe, and KUKA from Germany. Chinese factories bought overwhelmingly from them.

Two things have changed.

Domestic manufacturers have taken real share, and crossed the threshold. Chinese brands passed 50% of the domestic market in 2023, with domestic self-sufficiency in industrial robot production subsequently reported around 57%. The absolute numbers are equally striking: China installed roughly 295,000 industrial robots in 2024, more than the rest of the world combined.

The firms behind that are Estun, Siasun, Efort, Inovance, JAKA, and Dobot. Inovance is the most instructive case: it came into robotics from industrial drives and motion control components rather than from robot assembly, which is precisely the direction that builds durable advantage rather than temporary share.

One of the Big Four is now wholly Chinese-owned. The appliance maker Midea acquired roughly 94.5% of KUKA in 2016 and completed a full buyout in November 2022, taking the company private as a wholly owned subsidiary. A German firm that helped define industrial robotics is now entirely inside a Chinese conglomerate — a fact strangely absent from most discussions of Chinese robotics catch-up, and one that supplied engineering depth which would have taken far longer to build.

Chinese firms have been strongest in collaborative robots — lighter arms designed to work near people. AUBO, JAKA, Dobot, and Elite compete credibly here, and the reason is instructive: cobots are a newer category with a lower precision bar and no fifty-year incumbency. Where the entry barrier is technical heritage, Chinese firms have struggled. Where the category is new, they have not.

The components problem

This is the part of Chinese robotics that determines whether the rest is real.

A precision industrial robot's capability is set largely by three subsystems:

Precision reducers. Gearboxes that convert fast, low-torque motor rotation into slow, high-torque joint movement with minimal backlash. Two types matter: RV reducers for heavy joints and harmonic drives for lighter, more precise ones. Japanese firms — Nabtesco in RV, Harmonic Drive Systems in harmonic — have held commanding global positions for decades. The difficulty is not the design, which is public. It is manufacturing tolerances and consistency at volume, which is exactly the kind of accumulated process knowledge that resists reverse engineering.

Chinese challengers exist and are improving — Leaderdrive in harmonic drives, Shuanghuan in RV — and this is the single most important indicator to track in Chinese robotics.

Servo motors and drives. The motors and their control electronics. Japanese suppliers have led here too, but this is where Chinese substitution has advanced furthest, with Inovance the clearest example.

Controllers. The software and real-time motion control that turns a trajectory into coordinated joint movements. This is the least visible and hardest layer, and it is where Fanuc's advantage has been most durable. Control software improves through decades of deployment across millions of installations, which is not a thing capital can shortcut.

The pattern is consistent and worth stating plainly: China has closed the gap fastest where the constraint is electronics and slowest where it is precision mechanics or accumulated software. That is not a Chinese peculiarity — it is what the underlying learning curves look like — but it does mean component independence will arrive unevenly, and reducers will be the last domino.

Humanoids: the leapfrog bet

Humanoid robotics went from a research curiosity to a national priority in China with unusual speed, and the sector now has more funded entrants than any other country's.

The main players

Unitree (Hangzhou) is the most visible, and its defining contribution is price. Its G1 humanoid has been listed at around $13,500 and the smaller R1 at under $5,000 — figures that reframed what a research-grade humanoid should cost. The larger H-series sits an order of magnitude higher, but the entry point is the point: a robot at R1 pricing costs less than many single industrial arms, which moves humanoids out of institutional procurement and into something closer to ordinary equipment purchasing. National television appearances gave Unitree domestic name recognition no competitor has, and its Shanghai STAR Market listing was approved in 2026.

UBTech (Shenzhen), listed in Hong Kong, took the opposite approach — targeting industrial deployment directly. Its Walker S series has entered training and pilot deployment with an unusually broad set of manufacturers, concentrated heavily in vehicles: NIO, BYD, Geely/Zeekr, FAW-Volkswagen, Dongfeng, BAIC, and Foxconn. UBTech's bet is that the first real humanoid market is factory work.

AgiBot / Zhiyuan Robotics, founded by a prominent former Huawei engineer, is the volume story: reported shipments of over 15,000 units by mid-2026, backing from Tencent, Hillhouse, BYD, and LG Electronics, and a Hong Kong listing process begun in 2026 at a multi-billion-dollar valuation target. Fourier Intelligence (Shanghai) came from rehabilitation robotics, giving it real experience with machines operating near human bodies. Galbot, LimX Dynamics, Robot Era, and EngineAI fill out a crowded field.

Large technology and vehicle companies are also present. Xiaomi unveiled CyberOne in 2022 and continues internal development; XPeng's Iron humanoid runs on the company's own Turing chips and its physical-AI software stack, aimed at factory assembly and retail deployment. Their arrival from consumer electronics and electric vehicles is a signal about where the supply chain sits.

The state participates directly through the Beijing National and Local Co-built Embodied AI Robotics Innovation Center, which operates commercially as X-Humanoid and developed the Tiangong platform — a full-size all-electric running humanoid, open-sourced rather than held proprietary. That choice is worth noting: open-sourcing the platform layer is a familiar Chinese competitive move, commoditising the thing incumbents would otherwise charge for.

Why China's advantages here are real

The structural case is stronger than the demo videos suggest, and it rests on one thing above all.

The electric vehicle supply chain is the humanoid supply chain. A humanoid needs batteries, power electronics, electric motors, reduction gearing, sensors, thermal management, and high-volume precision assembly. China built exactly that industrial base over the past fifteen years for a different purpose. No other country can source humanoid components domestically at comparable cost and speed, and this is why Chinese humanoid pricing is not simply subsidy.

BYD's dual role is the cleanest evidence for this. It appears in this guide as a customer of UBTech's humanoids and as an investor in AgiBot. A vehicle manufacturer treating humanoids as both a factory input and a portfolio adjacency is what it looks like when two industries share a component base — and it is not a pattern that could easily occur anywhere else.

Supporting factors: control of rare earth processing and permanent magnet production, which feeds directly into motors; iteration speed from dense component supply chains where a redesigned part arrives in days; manufacturers willing to host deployment pilots; and abundant state and local funding.

Why scepticism is warranted

The gap between what humanoids demonstrate and what they do productively remains very large, and several specific concerns deserve stating.

Locomotion is the easy part. Walking, running, backflips, and dancing are impressive and largely solved. Dexterous manipulation is not — grasping unfamiliar objects, handling deformable materials, recovering from failure. Almost all economically valuable work is manipulation. A robot that walks beautifully and cannot reliably pick things up is not close to useful.

Reliability is the whole product in industry. Fanuc's advantage is not that its arms are cleverer. It is that they run for years with predictable maintenance, meet safety certification, and integrate into existing lines. Humanoid firms have not yet had time to demonstrate any of this, and no amount of capital shortens a mean-time-between-failure measurement.

Deployment figures are hard to verify — but they are no longer trivial. AgiBot's reported five-figure shipment total is the strongest evidence against pure scepticism, and it deserves to be taken seriously rather than waved away. The caveat is that shipped is not in productive use: a large share of humanoid units sold today go to research labs, universities, developers, and demonstration programmes rather than into production work. Prefer evidence of repeat industrial purchasing over cumulative shipment counts.

This is about to get easier to assess. Unitree and AgiBot both moved toward public listings in 2026, which brings audited disclosure to a sector that has been reported almost entirely through company announcements. Within a few reporting cycles there will be verifiable revenue, margin, and customer-concentration data. That is the most useful development in the sector for anyone trying to separate the real business from the promotion.

The sector shows the classic Chinese industrial pattern. State designation as a priority, abundant local government funding, a rush of entrants, rapid price competition, and eventual brutal consolidation — the trajectory of solar, batteries, and electric vehicles. That pattern has produced globally dominant Chinese industries. It has also destroyed most of the companies that participated in it.

The form factor itself is contestable. The argument for humanoids is that the world is built for human bodies, so a human-shaped robot needs no environmental modification. The counterargument is that purpose-built automation beats general-purpose machines at almost every specific task, and that warehouses and factories are quite willing to modify themselves. This debate is unresolved, and the answer determines whether the entire category is a large market or a fascinating detour.

The policy connects the two tracks

It is tempting to read industrial robotics and humanoids as separate programmes. Chinese policy does not.

MIIT's Guiding Opinions on the Innovation and Development of Humanoid Robots (人形机器人创新发展指导意见), issued in November 2023, set targets for 2025 and 2027. The 2025 objectives were to establish an innovation system, achieve breakthroughs in what the document calls the "brain, cerebellum, and limbs," reach initial mass production — and secure domestic supply of core components. The 2027 objectives extend to a safe and reliable domestic industrial supply chain, deeply integrated into the real economy.

That component clause is the tell. The state is not treating humanoids as a moonshot separate from the unglamorous work of learning to make precision reducers. It is treating the humanoid push as demand creation for a domestic component industry — a large, politically visible, state-backed customer base for exactly the parts China currently imports from Japan.

Read that way, the two tracks in this guide are one strategy. Even if humanoids disappoint commercially, a maturing domestic reducer, servo, and actuator industry would leave China materially less dependent in industrial robotics. The humanoid bet has a consolation prize, and it is the thing China has wanted for twenty years.

The targets are also usefully falsifiable. "Initial mass production by 2025" and a "reliable supply chain by 2027" are dated, checkable claims — worth measuring against outcomes rather than restating.

Reading China against Japan

The two robotics guides on this site describe near-inverse positions, and the contrast is the most useful thing in either.

Japan holds incumbency: precision, reliability, the component base, and decades of deployment. Its growth is in service and logistics robotics driven by absolute demographic decline, and its disposition is conservative — Japanese firms ship when a machine works, not when it demonstrates.

China holds volume, cost, supply chain depth, and speed. Its industrial robotics strategy is substitution from the outside in, and its humanoid strategy is a bet that a new platform makes incumbency worthless.

Three observations from putting them side by side:

They are less adversarial than they appear. Japanese component makers sell into Chinese robot production. Growth in Chinese robot manufacturing has been good business for Nabtesco, Harmonic Drive, and Japanese servo suppliers. Japan monetises Chinese volume — until Chinese component substitution succeeds, at which point that relationship inverts sharply. This is the same dynamic as Japanese materials in Korean semiconductor fabs, and it is worth watching for the same reason.

Both are demographic stories, but not the same one. Japan's labour shortage is absolute and long-anticipated. China's is newer, faster, and complicated by wage inflation and younger workers' reluctance to take factory jobs. China is ageing before reaching Japanese income levels, which makes automation less an efficiency programme than a necessity.

They have divided the frontier by temperament. Japan is strongest where reliability compounds; China is strongest where iteration speed compounds. Humanoid robotics currently rewards iteration. Industrial deployment rewards reliability. Which of those the humanoid market eventually rewards is, in effect, the whole question.

What to watch

• Domestic share of precision reducers. The single best indicator of genuine Chinese robotics independence. Everything else can be assembled; this cannot be shortcut.
• The first audited results. Unitree's and AgiBot's listings will produce disclosed revenue, margins, and customer concentration. One set of audited accounts will settle more arguments about this sector than three years of announcements have.
• Repeat orders, not first orders. A manufacturer buying a second batch after running the first is worth more than any pilot announcement. UBTech's vehicle-maker list is the place to look, since those relationships are furthest along.
• Manipulation benchmarks. Watch for unstructured grasping and failure recovery rather than locomotion. That is where the category's economics live.
• The 2027 MIIT supply chain target. Dated and checkable. Whether China has a genuinely domestic core-component supply chain by then is the clearest available verdict on the whole strategy.
• Consolidation. The current number of funded Chinese humanoid firms is not sustainable. Who acquires whom will be clear within a few years.
• Exports and controls. Chinese robot exports are growing, and robotics sits close enough to sensitive technology that trade restrictions in either direction are plausible. Component export controls would be the sharpest available lever, and Japan holds it.
• Whether Japanese incumbents move. If Fanuc, Yaskawa, or their European peers enter humanoids seriously, that is a signal the category is real. Continued absence suggests they have concluded it is not — and they have been right about robotics fashions before.

Related guides

Japan Robotics Landscape is the direct counterpart and the one to read alongside this. China AI Industry Map covers the model and compute layer that embodied AI depends on. China Semiconductor Ecosystem covers the chips inside these machines, and Korea Semiconductor Ecosystem offers the closest structural parallel to China's component dependency problem — a country strong at the system level and reliant on Japanese inputs beneath it.

Last updated: August 2026