110 AgiBot Robots Dance in Record-Setting Formation, on the Strength of Production-Line Consistency

October 24, Jinan Yellow River International Convention and Exhibition Center. One hundred and ten AgiBot Yuanzheng A3 full-size humanoid robots jumped, flipped and shifted formation in sync with the music, earning a Guinness World Record on site for the most humanoid robots dancing simultaneously.

One detail set it apart from most previous humanoid robot demonstrations: all 110 units were mass-production models shipped to customers, with not a single specially tuned demo machine among them.

Wang Chuang, AgiBot partner and senior vice president, put it bluntly at the event: "The hundred robots on stage today are all mass-production models, not one is a specially tuned demo unit. Joint zero positions, torque settings and motion libraries are standardized at the production-line level on every robot. That consistency is what makes a hundred of them on one stage behave like one."

All 110 on Stage, Not One Specially Tuned

In a lab, one success out of a hundred attempts is enough to declare a technology proven. On a mass-production line, a single failure among a hundred robots counts as a production incident.

So the headline number is not really the point. The point is that all of them came off the same line, calibrated the same way.

On core components, AgiBot has tied up with suppliers including Henggong Precision, which has mass-production experience in reducers and other critical parts. AgiBot is now planning a fully automated line at its Jinan base, expected to be visible by the middle of next year, with annual capacity of 8,000 units per line and four lines planned over the longer term.

IDC's latest report puts global humanoid robot shipments at nearly 25,000 units in the first half of 2026, up 432.1% year on year. Within that surge, AgiBot shipped more than 8,600 units in the first half, roughly 35% of the global market and more than 45% of the Chinese market.

Self-Unboxing, Self-Positioning, Automatic Gap-Filling When a Unit Drops Out

One easily overlooked detail of the challenge: after the 110 Yuanzheng A3 units arrived on site, they drove themselves out of their shipping crates, located themselves and moved into their assigned positions without help.

In Wang's framing, that matters more than the dancing itself: "The core of deployment mode is not flashy stunts. It is making robots behave like mature production equipment or service terminals: plug and play, continuously stable, and quick to replicate and scale."

He also drew a distinction between two modes. Development mode is aimed at professional developers who buy one or two units to validate a single function, with no concern for scaled deployment. Deployment mode is aimed at ordinary consumers and business customers, and requires robots to work fully autonomously without remote control. In development mode the robot is a "machine"; in deployment mode it is a "partner" and an "employee."

Underpinning this is AgiBot's in-house LinkBots fleet coordination platform, which uses a hybrid 5G-plus-WiFi communication setup to keep links stable at hundred-unit scale. Positioning matters even more: the Yuanzheng A3 ships with UWB (ultra-wideband) centimeter-level fused positioning as standard, holding its coordinates without drift even in difficult outdoor conditions such as open spaces, sunlight interference and crowd occlusion.

The system also includes fault tolerance. If a single robot drops offline, the others automatically compensate locally so the overall formation does not collapse. On stage that just means the formation stays tidy. In a factory it maps to a different question: whether a single point of failure can drag down an entire fleet operation.

Taken together, self-unboxing, self-positioning and self-correction compress the logistics, labor and on-site commissioning costs of deploying in volume.

From Chimelong Park to Subway Stations to a Tablet Quality-Inspection Line

Technology ultimately has to prove itself in real settings. AgiBot is currently validating commercial paths in three categories.

Tourism and entertainment got going first. Since September 24, more than 300 AgiBot robots have been working regular shifts at Chimelong Spaceship park in Hengqin, covering entertainment performances and science education programs. In Jinan, AgiBot has launched pilots at subway stations including the Olympic Sports Center and Zhuanshan West Road, offering information and wayfinding, crowd flow assistance and city culture commentary.

Retail and hotels are moving from simple reception work toward deeper service. AgiBot has deployed 100 Lingxi X2 units as sales assistants in ASD stores, each working more than 10 hours a day. In hotels, robots have gradually taken on the full service chain: check-in, carrying luggage, riding elevators, restaurant service and even room tidying.

Industrial settings are where penetration runs deepest. The AgiBot Jingling G2 is now in service at Longcheer Technology's Nanchang plant, covering the entire tablet mass-production quality-inspection stage. That marks humanoid robots moving from peripheral services into core manufacturing.

The product line is already broad: the full-size Yuanzheng, the half-size Lingxi, the wheeled Jingling and the quadruped Kutuo. To date, AgiBot has filed more than 600 patent applications, about 75% of them for inventions.

The Real Test of Deployment Mode

The performance proved two things: AgiBot can build consistent products, and it can coordinate a fleet of more than a hundred.

But dancing is a task with clearly defined boundaries. The motion library is standardized on the line, the choreography and formations are planned in advance, and commands come from a single source on site. It tests execution consistency and communication stability, not a robot's judgment when facing an unscripted task.

In other words, "deployment mode" as Wang uses it refers to delivery capability, not work capability. A stage can validate the former; it cannot validate the latter.

The real answers lie in settings nobody has choreographed: what happens when a robot in a subway station meets a tourist speaking a dialect it cannot parse, when a factory inspection line encounters a defect it has never seen, when a child suddenly blocks a hotel corridor. AgiBot has a presence in all of these settings, but all of them are still in pilot stage or the early phase of regular operation.

Beyond that come capacity and global expansion. IDC forecasts global humanoid robot shipments will top 750,000 units by 2030. AgiBot's recent moves run on two tracks: setting up an office in Hong Kong and assembling an R&D team there, with partnerships at the University of Hong Kong and the Hong Kong University of Science and Technology; and launching the "Yuansheng" ecosystem development plan, investing more than 2 billion yuan over five years in an embodied intelligence industry-education platform.

From a hundred robots dancing in Jinan to regular shifts at Chimelong in Zhuhai to quality inspection on a Nanchang production line, the path is clear enough. How far it runs will be decided in the settings that were never choreographed.