On 9 December 2025, AgiBot threw a lavish celebration for the launch of its 5000th humanoid robot. The event, streamed live from the Chinese manufacturer's factory in Shanghai, had all the hallmarks of an industrial milestone: mass production, at last. But the detail that was meant to slip by unnoticed tells a rather different story. This 5000th unit did not head off to a factory, a warehouse or a shop. It was handed over, symbolically, to the studio of Chinese actor Huang Xiaoming.

A showcase robot meant to embody the leap into mass production, ending up in show business: the image is too good to let pass. And when you look closely at how the figure itself breaks down, the tale of the "5000th worker" starts to come apart.

The 5000th robot is a stage model

According to Pandaily, unit number 5,000 was a Lingxi X2, delivered symbolically to Huang Xiaoming's studio. The choice is anything but innocent. The X2 sits at the lightweight end of AgiBot's catalogue: a nimble, half-height biped, built for demonstration and entertainment far more than for any kind of heavy lifting.

In other words, to mark a milestone in industrial manufacturing, the company picked its most crowd-pleasing robot and handed it to a showbiz figure. The casting of the event is, in itself, a confession: you parade the robot where it shines – on stage – and you take it well away from where it is, one day, supposed to work.

What AgiBot actually makes

The breakdown of what has rolled off the line, reported by Interesting Engineering, bears this reading out. Across the 5000 units, the split by range speaks volumes:

  • X-Series (nimble half-height biped, entertainment-oriented): 1,846 units — the most-produced range.
  • A-Series (full-size humanoid): 1,742 units.
  • G-Series (wheeled, task-oriented base): 1,412 units.

The conclusion is plain: it is neither the full-size industrial humanoid nor the task-oriented model that tops production, but the entertainment range. The most-built chassis is not the one meant to replace an operator on a line or a shop-floor worker in the aisles. It is the one that can dance.

Building 5000 robots is not deploying 5000 workers

This is where the sleight of hand creeps in. Producing 5000 humanoids is a genuine manufacturing feat – assembling 5000 frames, feeding a line, hitting volume. But this manufacturing success is being dressed up as a labour success: 5000 robots supposedly about to clock on. Yet nothing in the staging actually backs that equivalence. The number that matters is not "5000". It is which of the 5000, and where they end up.

AgiBot's commercial strategy only deepens the doubt. According to the Japanese outlet Robotstart, the company favours building an ecosystem of partners over selling robots directly: it supplies open-source development platforms, data and large language models, and goes looking for local partners to train its machines for cultural and commercial uses. A company that had mastered robots ready to work would sell them turnkey. AgiBot, by contrast, is selling the tools to learn how to put them to work.

The real work is for later

The company's own line, for that matter, owns up to the delay. Again according to Robotstart, AgiBot's marketing director, Qiu Heng, describes a three-phase deployment strategy, ranked by complexity:

  • Simple environment, simple task: factories and warehouses.
  • Complex environment, simple task: supermarkets.
  • Complex environment, complex task: domestic work, billed as a potential future market of hundreds of millions of units.

This last phase, the most ambitious, remains explicitly in the future. Qiu Heng also points to Japan's konbini — those round-the-clock corner shops — as a particular opportunity, on account of high labour costs and the safety risks hanging over night staff. An opportunity, then; not yet a deployment. The executive showed off the A2, X2 and G2 ranges on the company's stand at the iREX2025 show in Tokyo, but the leap from projection to signed contract has yet to be made.

A shortfall dressed up as a triumph

The 5000-robot milestone is real, and it does tell us something true: the sector has learned to churn out humanoids fast. Very fast. But it has not yet learned to put them to real work at the same clip. By picking a stage robot as its symbol and handing it to an actor, AgiBot unwittingly lays bare that gap. The first success – building them – is being used to paper over the second shortfall – employing them. The open question, once the lines are running at full tilt, is how many of these robots will really go to the factory, and how many will keep going to the studio.