Nine out of Every Ten Humanoid Robots Worldwide Are Now Made in China

NextFin News -- While Musk was hyping on X that Optimus would “enter mass production next year,” and Silicon Valley startups were racking up awe across the internet with short videos of robots doing parkour or making coffee, a hard-nosed Wall Street research report tore open the industry’s glossy façade.

Morgan Stanley’s latest report released this month laid out an industry reality that is right in front of everyone—yet collectively ignored: of the 13,000 to 16,000 humanoid robots shipped globally in 2025, as many as 90% came from Chinese manufacturers.

From supply-chain spillovers and relentless cost compression to rapidly converging capital, China is using an all-too-familiar playbook to replicate humanoid robotics as “the next electric vehicle.” In this battle for manufacturing clout in humanoid robot hardware, the harsh realities of scaling up have already mercilessly crushed the tech halo surrounding prototype demos.

This set of counterintuitive numbers begs a question: what, exactly, is the industrial logic that makes them possible?

90% of Shipments Not from the U.S.

For a long time, there was a deeply rooted misconception about the global humanoid-robot race: that the United States was still far ahead.

But Morgan Stanley’s in-depth humanoid-robot report released this month poured cold water on that Silicon Valley filter. It delivered a highly disruptive quantitative answer: of the roughly 13,000 to 16,000 humanoid robots shipped worldwide in 2025, about 90% were from Chinese manufacturers. Companies in the United States, Japan, and other countries were still largely stuck in the prototype exploration stage. (Note: methodologies vary slightly across institutions—Omdia put the figure at 13,000 units, Counterpoint at 16,000; this article uses the range cited by Morgan Stanley.)

This isn’t conjecture—it reflects industrial logic grounded in big-data analysis. In the real commercial world, trending topics can’t conjure up production capacity, and prototypes don’t magically become industrial products. The center of gravity in the global humanoid-robot industry has already quietly shifted eastward.

How did China take 90% of shipments? Morgan Stanley’s report pointed to the resonant set of levers behind it.

China's State-Owned Enterprise Orders Serving as the “Icebreaker”

While U.S. peers were still struggling with the gap between “able to walk” and “able to work,” Chinese players had already moved into the hard-driving phase of “cutting costs, scaling output, and finding real scenarios.” In 2025, among early orders for humanoid robots in China, more than RMB 2 billion came from state-owned enterprises (SOEs), mainly for power-plant/data-center maintenance and public services. These early, government-investment-driven orders helped companies make the hardest leap—running through the full closed loop from R&D to commercialization.

The numbers show it most plainly: in 2025, all of the top six vendors by global shipments were in China. Zhiyuan ranked first with 5,100 units (39% of the global total), and Unitree ranked second with 4,200 units (Note: Omdia data; Unitree’s officially announced actual 2025 shipments were 5,500 units—differences in statistical definitions lead to an approximate figure; Omdia’s cutoff date was earlier than Unitree’s full-year disclosure, so the two sets of figures are not directly comparable. Both leading companies shipped more than 4,000 units, far ahead of the pack), while UBTECH ranked third with 1,000 units. By comparison, three U.S. headline names—Figure AI, Tesla, and Agility Robotics—each shipped only about 150 to 500 units for the year, and together accounted for just about 3% to 10% of global volume. In absolute terms, shipments from a single top-tier Chinese company were 36 times those of its U.S. counterparts.

The Supply Chain for the EV Industry Benifits the Robotics Industry

Of the 100 key companies across the global value chain compiled by Morgan Stanley, 56 were from China (Note: this count focuses primarily on the hardware supply chain and differs from the full-industry-chain methodology used in the Humanoid 100 report in early 2025). From reducers made by Leaderdrive, to servo motors from Inovance, to Orbbec’s 3D vision sensors, China had all but taken on half the supply chain.

But the more decisive weapon was “industrial spillover.” China’s first-mover edge in humanoid robots was, in essence, a supply-chain dividend accumulated over the past decade in the new energy vehicle (NEV) industry. From sensors and motors to battery management systems, the supply chains for EVs and humanoid robots overlap heavily. Selina Xu, China AI policy lead at former Google CEO Eric Schmidt’s office, put it bluntly: China has the world’s strongest manufacturing base, enabling iteration speeds far beyond those in the West. This deep integration rapidly drove the price of domestically made full robots down from the million-RMB range to the hundred-thousand-RMB range.

Global Capital Bets on China: 46% of VC Money Flowed into China

Morgan Stanley research report data shows that as of early 2026, global venture capital investment in humanoid robots has already surpassed the total for all of last year, with the Chinese market attracting about 46% of the funding. As Morgan Stanley’s Chief Asia economist Chetan Ahya put it, China is building out the entire supply chain—giving it a first-mover edge over competitors in the U.S., Japan, and South Korea that depend on Chinese components.

The Next EV Script

Once you understand the logic above, looking across the Pacific feels like two parallel universes.

Around the time Morgan Stanley released its report, Tesla’s Austin factory was still repeatedly testing whether Optimus could walk steadily, while Musk kept reiterating his promise of “mass production next year”; Figure AI had just unveiled Figure 03, once again setting off a media frenzy. But most of this remains a “concept show” stuck in the engineering validation stage.

What truly has consumer-product qualities is coming from Chinese companies. Unitree has priced its G1 humanoid robot at RMB 99,000 and is selling directly to end users; comparable U.S. products are either stuck at a steep US$60,000 price tag or haven’t even cleared the threshold for mass production.

This reflects a fundamental divergence in the two countries’ industrial paths: the U.S. approach is “brain first,” attempting to empower hardware with large models and pursue extreme single-point breakthroughs; China’s approach is “body first,” leveraging its full industrial-chain advantages to turn ideas into industrial products at scale—low cost and high efficiency—then iterate through real-world use. The former wins brand buzz; the latter captures the market.

In this report, Morgan Stanley lays out an exceptionally clear framework: China’s humanoid-robot industry is precisely replaying the electric-vehicle script.

In 2025, China’s robot exports were roughly on par with the scale of EV exports in 2019. Analysts project three key milestones:

• 2026 (scenario validation phase): Multiple on-the-ground tests will be carried out across China to demonstrate real applications of robots in logistics, manufacturing, and retail, with annual sales expected to double year over year to around 28,000 units.

• 2030 (the eve of breakout growth): China’s annualized net addition of robots in operation will reach 21 million units, with humanoid robots surging from 12,000 to 260,000. Global export share will rise from the current 15% to 16.5%, and humanoid robots will formally take over from new energy as a hundred-billion-level export pillar.

• 2050 (ultimate endgame): The cumulative global installed base of humanoid robots will reach roughly 1 billion units, with the market expected to hit $5–$7.5 trillion (note: the report projects $7.5 trillion; $5 trillion is a more conservative estimate from other institutions) ——equivalent to 2–3 times the combined revenue of the world’s top 20 automakers in 2024.

However, taking 90% of shipments does not mean China can rest easy. Morgan Stanley’s research also points to two swords of Damocles hanging over the industry.

First is the unfinished quest for the “brain.” World models and VLA (Vision-Language-Action) models have yet to meet the requirements of high-precision industrial applications. Competitive moats are tilting toward a “proprietary data flywheel”: whoever can capture more real-world physical data will be able to train a truly usable robot brain. This has traditionally been a strength of U.S. companies.

Second is the chasm between “cost and expectations.” Morgan Stanley’s survey shows that 62% of Chinese companies have purchase intentions, but only 23% are satisfied with current products; dissatisfaction centers on flexibility, functionality, and price. More importantly, 92% of surveyed companies drew a hard line: only if the unit price drops below RMB 200,000 will they adopt at scale.

“2026 is a pivotal year, and the industry may be on the verge of entering a shakeout period,” warned Zong Sheng, Morgan Stanley’s Chief China Industrials Analyst. Once “who can build it” is no longer the bar to clear, the competitive battleground will shift brutally to “who can make it cheap enough and good enough.”

The biggest contribution of this Morgan Stanley report is that it shows there is another track beyond the AGI brain, reminding the world that a trillion-dollar industry centered on the “robot body” is rapidly taking shape.

From monopolizing 90% of global shipments in 2025 to a near frame-by-frame reenactment of the EV playbook, China has already surged to the front of the pack in mass production. But the real showdown still lies ahead—when the software dividend of foundation models and the hardware momentum of the supply chain are forced to fuse into one, only those who are first to break through the 200,000-yuan cost threshold and make the data flywheel of the physical world truly run will become the “BYD” and the “Tesla” of the robotics era.

As the tailwind shifts from code in the digital world to steel in the physical one, this ultimate endurance race of hardcore manufacturing has only just fired the starting gun.