Global Humanoid Robot Shipments Near 25,000 in H1, with Chinese Vendors Taking Almost 80%

In its Global Humanoid Robot Tracker report released on September 30, IDC offered a figure that is hard to overlook: in the first half of 2026, global humanoid robot shipments approached 25,000 units, up 432.1% year over year. China's market alone shipped more than 19,000 units, up 426.3% year over year, accounting for roughly 77.9% of the global total. The same report contains another line: the global market exceeded $740 million, up 322.7% year over year.

Put the two sets of numbers together, and unit growth outpaces revenue growth. That suggests the products being sold right now do not carry high price tags, and the industry has not yet reached the stage of making money on high-value orders; it is mostly about seeding products into real-world scenarios on a trial basis.

Where AGIBOT's 8,600 Units Came From

The most closely watched company in the report is AGIBOT. It shipped more than 8,600 units in the first half, about 35% of the global market and over 45% of the Chinese market, placing it at the far right of IDC's vendor bubble chart. Its growth came on top of an already sizable base, and IDC's wording is "a runaway lead."

Unit volumes have to be absorbed by scenarios. AGIBOT disclosed deployment projects along two lines. On the cultural tourism line, starting September 24, more than 300 robots entered Hengqin Chimelong Spaceship Park, spread across over 100 interaction points and covering seven categories of scenarios — entertainment performances, science education and study tours, guided tours and shopping assistance, service retail, intelligent companion tours, hotel services, and sports competition — operating 8 hours a day. On the retail line, 100 robots went into Aishida stores as shopping guides, with Lingxi X2 on duty more than 10 hours a day, using a half-size body to fit the spaces originally set aside for people, such as shelves, checkout counters, and experience areas.

What the two lines have in common is that neither places harsh demands on the environment. Parks and stores are spaces with controllable foot traffic and fixed routes, so robots do not need to make long-chain autonomous decisions in a fully open environment, and staff can step in if something goes wrong. As humanoid robots move from technical validation to commercialization, the first batch of orders will most likely look like this.

The Focus of Competition Is Shifting Tracks

Another judgment worth noting from the IDC report is that competition in the industry is shifting from "body performance" to "comprehensive embodied intelligence capability." Translated: what used to be compared were hardware parameters like joints, gait, and payload; what is compared now is whether a robot can understand the physical world, plan tasks, and learn and correct itself while running continuously. Large models, world models, vision-language-action models (VLA), and robot data loops form the infrastructure of this new track.

This shift is not especially friendly to companies that started out in hardware. Being able to build a body is only a ticket to entry; what comes next is a contest over models and data. Chinese vendors have an edge on this track, because complete-machine makers, component suppliers, AI model developers, and system integrators collaborate within the same market, so product costs and mass-production efficiency should improve faster than before.

They Sell Whole Robots, but the Profit May Lie Elsewhere

Strong shipment numbers do not necessarily mean a strong income statement. In the humanoid robot cost structure, joint modules, reducers, dexterous hands, and sensors account for the bulk, and complete-machine makers in the early stage often have to cede gross margin to the supply chain in exchange for stable production capacity. The IDC report notes that China has a relatively complete industrial chain, with complete-machine manufacturers, component firms, AI model companies, and system integrators collaborating in the same market. The upside of that collaboration is fast iteration; the price is fast competition too, because the same supply chain can serve multiple complete-machine makers, and the differences between products get smoothed away quickly.

This is also why AGIBOT repeatedly stresses full-stack in-house development in its public statements. For a complete-machine maker to build a moat, it either has to develop capabilities in models and data that others cannot take away, or lead in the stability of mass-production delivery. Body design alone makes for a very shallow moat.

The 750,000-Unit Forecast and Its Premises

IDC raised its forecast for global humanoid robot shipments in 2030 to more than 750,000 units, about 50% higher than its previous forecast. The reason behind the upgrade is that small humanoid robots aimed at individuals have begun to appear in the consumer market, with home and personal scenarios treated as new growth space.

Policy is a variable too. Since the start of this year, local governments, industrial parks, and state-owned enterprises have kept stepping up, pushing R&D toward deployment by building innovation platforms, publishing application scenarios, and advancing procurement and deployment. That kind of support can shorten the distance from prototype to order, but it also means that "shipments" in the statistical definition include a fair number of demonstration projects.

What really determines whether these robots stay on the job is still the cost per run and the failure rate. Three hundred units running 8 hours in a theme park and 300,000 units working three shifts in factories are two completely different problems. IDC's own timeline has the market moving from "technically feasible" to "large-scale application and commercial value validation" — "validation" is the key word, indicating the conclusion has not landed yet.

The Real Test Comes After the Sale

Stretch out the curve of the past two years and the first half of 2026 looks most like a turning point: shipments reached the 10,000-unit level for the first time, scenarios spread for the first time to places ordinary consumers can see, and vendors began competing for the first time on comprehensive capability rather than specs. But shipping is only handing out the exam paper.

What is worth watching next is not the next shipment report, but the on-duty rate of these robots that have already started work, six months from now. Only products that can survive a full replacement cycle can even begin to talk about holding on to that 77.9% share.