An analytical breakdown of John Koetsier’s interview with Dr. Yao Maoqing, president of the embodied AI business unit at AGIBOT — “World’s Biggest Humanoid Robot Maker Says Tipping Point Is Near”, Forbes, May 19, 2026. The full interview is at the link.
Shanghai-based AGIBOT is the world’s largest maker of humanoid robots: 5,100 units shipped in 2025, roughly 39% of the global market, more than 10,000 units cumulatively by early 2026, with operations in 17 countries. In a wide-ranging interview with Forbes, Dr. Yao Maoqing — president of AGIBOT’s embodied AI business unit — explained what he believes the actual competitive advantage of Chinese humanoid companies is. It isn’t the story Western analysts usually tell.
The interview’s sharpest question — and an unexpected answer
Late in the conversation, Koetsier asks what may be the strongest question of the whole interview:
“What’s the question Western journalists should be asking Chinese humanoid robotics CEOs that none of us are actually asking?”
Yao’s answer reads almost like an op-ed in his own voice:
“Outside China, the industry is usually explained through a cost advantage. But the real core of competition is not who can make the most impressive demo — it’s who can put robots into the real world faster.”
He then frames what he considers the real Chinese advantage: a unique combination of four factors — a wide range of real-world deployment scenarios, a complete supply chain, fast engineering iteration, and — crucially — a large number of customers willing to try new technology. When these come together, he says, the transition from lab to deployment speeds up dramatically.
In other words: China isn’t winning on price. China is winning on deployment velocity. That’s a fundamentally different framing — and one that’s almost never used in Western coverage.
“X curve” and “Y curve”: the industry is changing eras
To explain what’s happening to the humanoid robotics industry, Yao offers his own framework — two curves.
The X curve is the technology-exploration phase. The whole industry was busy proving that robots could walk, run, and execute complex motions. The era of demos. The era of backflip clips.
The Y curve is the deployment-growth phase. Here the question becomes different: can the robot actually enter workflows, operate continuously and reliably, and produce measurable productivity?
In Yao’s view, the industry is now crossing from X to Y. For years, he says, the field was waiting for three things to come together: powerful enough AI foundation models, reliable enough robot “bodies,” and a continuous data flywheel that only forms once machines are running in the field. These three conditions developed separately for a long time — and have, for the first time, converged in a single time window.
The shift, Yao argues, isn’t that robots have suddenly learned something qualitatively new. The shift is that the conversation itself has changed: from “what can robots do?” to “can robots actually create productivity?” Customers, he says, have stopped looking at demos and started seriously discussing deployment, replication, and ROI.
Where the robots are actually working today
To ground the conversation, Yao lists eight scenarios where humanoid robots are landing first: industrial manufacturing, logistics and warehousing, commercial services, security and inspection, research and education, data collection, and scenario validation. The home, he emphasizes, is a long-term direction — not a near-term one: domestic environments require dramatically higher levels of safety, reliability, and predictability than the industry can deliver today (the same point made in Agility Robotics’ recent position paper, which we covered).
And about money. AGIBOT is already offering robot rentals through a robots-as-a-service (RaaS) model in more than 17 countries. In the United States, Yao says, the starting price is around USD 2,000 per day — but that covers not only the robot itself, but deployment, operations, maintenance, software, and sometimes even on-site engineering support.
About production numbers, Yao is notably calm: AGIBOT reached its first 1,000 robots in roughly two years, 5,000 in another year, 10,000 in the next. Asked directly whether the next target is 100,000, he answers with surprising restraint: “We’re not particularly focused on a specific number. What matters more is whether robots are truly entering workflows, whether customers come back for repeat purchases, whether successful scenarios can be replicated.”
And here — the answer to Silicon Valley
Tucked into the middle of the interview is another important analytical pivot. Koetsier asks directly: “You have a very holistic approach — hardware, software, cloud. Do you think vertical integration will win in the first few generations of the industry?”
Yao’s answer is unambiguous: yes, at the early stage, vertical integration is a strategic necessity. The industry doesn’t yet have stable divisions of labor between suppliers, nor standardized interfaces between components. A change in the motion model may force a change in body design. Feedback from a real deployment scenario may force a retraining of the model. If these capabilities are fragmented across different companies, iteration becomes too slow.
And then comes a key fork. Yao says: “But in the long run, the industry will not remain completely closed.” As the field matures, specialized platforms will emerge — for operating systems, models, robot bodies, supply chains, and packaged scenario solutions. Vertical integration, he says, is a stage of the industry, not its final form.
This is a direct — though carefully phrased — counterpoint to what we covered in our piece on the 1X World Model Lab. 1X is building a closed vertical stack — from the NEO robot to its own factory to its own AI lab. CEO Bernt Børnich put it plainly: “To reach full autonomy fastest, you must own the entire stack.” Yao seems to answer from the other side: yes, for now — but only for now.
And then Yao places his own bet on the ultimate competitive factor:
“In the global market of the future, what will decide things is not who is more closed or more open. It will be who can build a long-term, transparent deployment system that respects local regulation. Trust will be the ultimate moat.“
Where 1X says “everything will be decided by data,” AGIBOT says “everything will be decided by trust.” In the broader picture, both are right — because they’re describing the same long-distance race from different sides: what separates winners from losers over the long haul.
A closing that ties the whole editorial arc together
Yao closes the interview with a short but programmatic line:
“The more important question for the future is not who can create the most impressive video first. It is who can build real-world deployment capability, operational capability, and a data closed loop the fastest.”
What will ultimately determine the industry, in Yao’s view, is not raw model capability — but who can continuously bring physical AI into real workflows.
This is, in effect, a direct response to nearly everything we’ve published in recent weeks. To Ambrose’s Fortune column, where the former head of NASA robotics argued that the winner won’t be whoever builds the flashiest robot but whoever can deploy. To Agility’s position that viral demos often turn out to be “deceptively disguised theater.” To the work of China’s humanoid training grounds, which is exactly about building a shared data flywheel. And to 1X’s World Model Lab.
When such different sources — a Norwegian-American startup, a former NASA robotics chief, California-based industry leaders, Chinese state-run centers, and the global market leader from China — converge on the same conclusion from different directions, it stops being a news story. It becomes a consensus: the race isn’t running on the stage, it’s running on the shop floor. Not for the loudest clip, but for the longest continuous shift a robot can pull in someone’s warehouse. And by that measure, Yao argues, Western commentary is looking in the wrong place.
Image: AGIBOT