China’s Humanoid Robot Surge Is a State-Backed Bubble, Its Own Regulators Warn

By 2025, more than 140 Chinese companies had launched over 330 distinct humanoid models . Production volumes tell the story of that scale: between 13,000 and 18,000 humanoid robots were shipped globally in 2025, with the vast majority coming from China . Morgan Stanley projected domestic sales would double to around 28,000 units in 2026 . Leading the pack is AgiBot, which captured an estimated 38% global market share, followed closely by Unitree .

This output is underpinned by a brutal cost advantage. Chinese-made humanoids cost at least 20% less than foreign equivalents, with basic editions starting around ¥180,000 ($25,000) and the cheapest full-featured robot, Unitree's G1, now selling for just $16,000 . Unitree itself is the emblem of this push: its average selling price collapsed from ¥593,000 ($85,000) in 2023 to ¥167,600 ($25,000) in 2025, even as its gross margins rose above 60% by self-developing core components .

In 2025, Unitree posted ¥1.71 billion in revenue, a 335% year-on-year jump, and an adjusted net profit of ¥600 million—an outlier in a field where most of its peers, like UBTECH, are still burning millions in operating cash flow . But even Unitree's shine is dimming: in the first quarter of 2026, as it pushed toward a major Shanghai IPO, its adjusted net profit plunged 52% year-on-year, squeezed by surging R&D and capacity spending .

The Demand Mirage and the Bubble Warnings

The production numbers are real, but the buyers are not. A stark assessment from the field captures the disconnect: "Without the demand and without that scale from the market, these companies are not able to really go into mass production," one expert observed . Most of the robots being shipped are not going to factories or homes, but to other robotics companies for data collection, research labs for experimentation, or onto stages for publicity stunts .

The practical barriers are immense. Even the cheapest full-featured humanoid costs $16,000, while enterprise-grade models from Boston Dynamics and Agility Robotics run to $150,000–$250,000 or more . This makes a return on investment almost impossible to justify for most real-world tasks. More critically, the robots themselves are not ready: "Most humanoid robots are still performative rather than functional, falling short of working in messy, unpredictable environments" . Embodied AI "brains" remain a bottleneck, with scarce training data for fine manipulation tasks . Current deployments are largely confined to factory trials and traffic management—narrow niches that cannot absorb the millions of units being forecasted .

The most extraordinary warning, however, came from Beijing itself. In late November 2025, Li Chao, a spokesperson for the National Development and Reform Commission (NDRC)—China's top economic planning body—did something almost unthinkable: he publicly stated the sector was forming a bubble. He pointed to over 150 companies, most of them startups, churning out "highly similar" products and stressed that "proven, large-scale commercial scenarios remain extremely limited" . His call to prevent "blind expansion" and to guide the industry toward consolidation over output growth hammered home the extent of the state's anxiety . The immediate market reaction was brutal: China's humanoid robotics stock index fell around 20% .

Beijing's warning did not come in a vacuum. Goldman Sachs had already cautioned about overcapacity building as production scaled up without actual orders . On the other side of the world, famed roboticist Rodney Brooks was simultaneously sounding the alarm on a humanoid investment bubble, noting that billions in venture dollars were being poured into companies with limited real-world dexterity solutions .

The Tension Between Tomorrow's Promise and Today's Glut

The cognitive dissonance of this moment is extreme. IDC projects global humanoid shipments exceeding 510,000 units by 2030, and Morgan Stanley has floated the idea of a $5 trillion future market . Goldman Sachs itself hiked its market forecast sixfold between 2022 and 2024 . These long-term visions are grounded in real demographic and automation trends, as labor shortages intensify across manufacturing economies .

But the path to that future is not a straight line. It runs directly through a present defined by a classic hype-cycle pattern: more than 140 nearly identical manufacturers fighting over a tiny pool of real buyers, with a production capacity that has outstripped authentic demand by a wide margin. Beijing's intervention is an acknowledgment that a painful shakeout is likely, one where only a few vertically integrated and genuinely innovative players survive. The robots will eventually find their place, but for now, the industry has built the factory before knowing what the world actually wants to buy from it.