Humanoid shipments jumped 432% in six months — and AgiBot dethroned Unitree

IDC counted close to 25,000 humanoid robots shipped worldwide in the first half of 2026, up 432.1% year on year — and AgiBot took the top spot worldwide at Unitree's expense. Both pieces of news landed in the same month, and they contradict each other less than they appear to. The market is roughly doubling every half-year, and it is still small, young and overwhelmingly demonstrative.

The two China numbers you must not mix

This is the easiest trap in the story, and both metrics are nevertheless quoted side by side.

Both statements say China dominates, but they do not measure the same thing. One is a share of worldwide shipments, the other a replacement rate among suppliers inside a single national market. A country can dominate its own purchasing without exporting the same share of its units, and that is exactly China's position. Treating the two figures as interchangeable amounts to claiming that 95% of humanoids sold worldwide are Chinese, which neither source says.

On the same subject, Forbes reports that nine of the top ten manufacturers worldwide are Chinese. That is information about the competitive structure of the sector, not about where the units are sold.

A change at the top, not a new category

The shift that matters is the one at number one. In the first half of 2026, AgiBot overtook Unitree to lead global humanoid shipments. Several financial outlets repeated the ranking on 30 September, the same day AgiBot announced it had delivered its 20,000th robot.

Read together, the two announcements say more than either alone. Twenty thousand units delivered since the company was founded, and a half-year in which it became the world's largest supplier. The delivery curve and the ranking curve are converging. It is the first time a Chinese manufacturer has passed the quadruped specialist that had dominated trade coverage for two years.

The reason probably lies in the product mix as much as in the technology. AgiBot positioned itself around data-collection centres, exhibitions and institutional deployments, which are precisely the segments that weigh heavily in this half-year's numbers, as our panorama of three new humanoids details.

69% of the market is still a demonstration

This is the number growth press releases skip, and it is the most important of the three.

According to IDC, research and education, shows and exhibitions, and government data-collection centres together accounted for 69% of first-half shipments. Put differently: roughly seven units in ten went to research programmes, to showcase venues, or to building training corpora under public control.

That is both good news and a limit. A data corpus is the precondition for the scale-up we analysed with Helix 2.5: no machines running means no data, and no data means no generalization. But a robot bought to fill a data tank is a robot producing no value for the buyer. The difference between a machine that works and a machine that is being demonstrated sits precisely in that 69% split.

For scale: IDC's 25,000 units is the entire world market for six months, and our 2026 humanoid shortlist barely covers a dozen serious suppliers. The ratio between those two numbers is why the 432% figure draws attention while remaining structurally secondary.

The research houses do not count alike

Smart Analytics Global counted 19,100 units in the first half of 2026, up 272%, against IDC's roughly 25,000 units and 432.1%. The gap is not trivia: it shows that measuring a nascent market is first a definition problem and only afterwards a volume problem.

What counts? A lab prototype? A demonstration unit invoiced to a customer? A research robot sold to a university? How are second attempts and revisions treated? The available reporting does not make clear whether Smart Analytics Global's 19,100 units cover the world or China alone. Two institutes, two definitions, two growth curves, and no way today to settle which one is authoritative.

There is also a documentary limit worth stating: IDC's primary release could not be opened, because its main page is JavaScript-driven. The percentages in this article come from dated, credible press reporting and remain attributed to IDC. We are not presenting them as a direct reading of the source.

The financial backdrop does not calm anyone

On 19 August 2026, Unitree listed on the Shanghai Stock Exchange: roughly 6.1 billion yuan raised, about $905 million, with +460% on day one and a close at 845 yuan, before a 42% slide by 24 September and a valuation of about $29 billion. Our analysis of that listing covers the sequence.

Two weeks after a 460% gain, AgiBot takes the top of the shipment ranking. The comparison is instructive: financial markets are already valuing the runner-up, and the podium just changed its leader. Valuations in Chinese robotics resemble the wider technology sector, where the share price runs well ahead of operating results.

What the half-year actually shows

Three things, in order. A change of leader: AgiBot passes Unitree, and nine of the top ten manufacturers are Chinese. Demand that is still demonstrative: 69% of units went to research, events and data collection. Fragile measurement: two institutes, two numbers, two definitions.

The +432% needs a caveat. On a base of a few thousand units a year earlier, a growth rate mostly measures the denominator. The real volume — 25,000 robots in six months worldwide — still belongs to a prototype industry. What is being decided in 2026 is not mass production: it is the building of training corpora, the standardisation of Chinese platforms and the shift of global leadership. For what comes next, our reporting on China's dark factories will tell you more than any annual growth rate.