China Is Quietly Out-Shipping Everyone in Humanoid Robots: 62,500 in 2026
China Is Quietly Out-Shipping Everyone in Humanoid Robots:
62,500 in 2026
While the West films viral robot demos, China is quietly building the actual robots — by the thousand. Here's the shipment gap nobody is talking about, why China can do it, and the one catch that could spoil the whole story.
Quick experiment. Open YouTube, type "humanoid robot," and look at the top results. Odds are you'll see a Tesla Optimus folding a shirt, a Figure robot jogging across a parking lot, maybe a Boston Dynamics Atlas doing a backflip. All American. All very shareable.
Here's the part the algorithm doesn't show you: while the West was busy producing clips, China was busy producing robots. Actual units. Boxed, shipped, invoiced. And the gap between those two activities turned out to be enormous.
The headline number making the rounds in Chinese industry circles is 62,500 humanoid robots shipped in 2026. That figure deserves a careful look — including a healthy dose of skepticism — so let's start there.
The 62,500 Number — Real, but Read the Fine Print
The 62,500 figure comes from China's Gaogong Industry Research Institute (GGII), which projects that the country's humanoid robot shipments will hit that level in 2026 — a year-over-year jump of more than 650%. It's a striking number, and it's the one driving a lot of headlines.
Now, the fine print. This is a forecast, not a finished tally, and it comes from a domestic Chinese research house that tends to sit on the bullish end. Western analysts are more conservative: Morgan Stanley expects China's humanoid sales to roughly double in 2026 to around 28,000 units. So the "truth" for 2026 probably lives somewhere in a wide band between those two.
What I find genuinely interesting, though, is that even the cautious number is staggering. 28,000 humanoid robots in a single year, from a product category that barely existed commercially eighteen months ago? That would have sounded absurd in 2024. The argument isn't really about whether it's 28,000 or 62,500. It's about who's shipping them — and the answer to that is not ambiguous at all.
Sources: Gaogong Industry Research Institute (GGII) forecast via industry reporting, Feb 2026 · Morgan Stanley estimate via Fortune/AP, June 2026
2025 Is When the Gap Became a Canyon
Forecasts are arguments. Last year's shipments are facts — and the 2025 facts are where this story stops being debatable.
According to market research firm Omdia, more than 13,000 humanoid robots shipped worldwide in 2025. Two Chinese companies, AgiBot and Unitree, each shipped over 5,000 of them. Their American counterparts — Figure AI, Tesla, Agility — shipped a few hundred each, in some cases fewer. Let me draw that, because the scale doesn't land in a sentence:
Source: Omdia 2025 humanoid shipment estimates, via SCMP & Fortune/AP, 2026. AgiBot ranked #1 globally by Omdia.
That's not a lead. That's a different sport. And it scales up: a Barclays research report estimates that Chinese-made humanoids accounted for roughly 85% of the global total in 2025. When a single country builds more than four out of every five humanoid robots on Earth, the "race" framing starts to feel generous.
For the other side of this scoreboard — the American players and what they're actually doing in their factories — we broke that down separately:
Read · The US SideMeet the Companies Doing the Shipping
Three names carry most of the volume. They're worth knowing, because in a few years you'll be hearing them the way you hear "BYD" today.
Barely two years old and already the world's top humanoid shipper by Omdia's count, with around 5,168 units in 2025. AgiBot's strategy is almost aggressively pragmatic: ship robots that are useful-but-imperfect now, then improve them with data collected from real deployments. Its Expedition A2 line reportedly crossed the 10,000-unit cumulative delivery mark in early 2026.
It's the Silicon Valley "ship fast, iterate" philosophy — except applied to walking machines, and executed in Shanghai.
Sources: Omdia & Counterpoint Research rankings · SCMP, Jan 2026
The one you've probably seen — doing kung fu at the CCTV Spring Festival Gala, or sparring in a robot kickboxing match. Behind the viral stunts is a serious operation: Unitree shipped more than 5,500 full-body humanoids in 2025, and here's the part that surprised me most — it says it actually turned a profit. Roughly 1.7 billion yuan (about $250M) in revenue, with profit north of 278 million yuan ($41M).
A profitable humanoid robot company. In a sector where most players are torching cash, that's the real flex — not the backflips. Its weapon is price: the G1 starts around $13,500, and the lighter R1 launched at roughly $5,900. For context, some US humanoids are priced in the six figures.
Sources: Unitree figures via 36Kr & SCMP · revenue/profit via Fortune/AP, June 2026
While Unitree chases consumers and AgiBot chases volume, UBTECH went straight for the factory floor. Its Walker S2 — a 1.76m industrial humanoid that can swap its own battery to run continuously — entered mass production in late 2025, with first deliveries of several hundred units and a stated path toward thousands per year.
The customer list is the tell: BYD, Geely, FAW-Volkswagen, Foxconn, SF Express. These aren't pilot-program logos for a press release. They're the companies that will decide whether humanoids actually earn their keep in manufacturing.
Sources: UBTECH press releases · People's Daily & AI Business, Nov 2025 – Jan 2026
And those are just the leaders. China reportedly had more than 140 humanoid robot manufacturers and over 330 models by 2025, per its Ministry of Industry and Information Technology. Newer entrants like Matrix Robotics (a $99,000 dexterity-focused humanoid, founded by an ex-Tesla engineer) and EngineAI (a $26,600 base model) keep arriving. It's less a field of competitors than a swarm.
Why China — and Almost Nobody Else — Can Do This
So how does one country out-build the rest of the planet combined? Three reasons, and none of them is "they got lucky."
1. They already own the parts bin. A humanoid robot is essentially a basket of actuators, reducers, sensors, and batteries bolted to a frame. China controls an estimated 63% of the global humanoid robot supply chain, with domestic vendors supplying high-torque actuators and sensors that everyone else has to import. When the parts are made next door, you build faster and cheaper. Morgan Stanley estimates locally sourced components make Chinese robots 20%+ cheaper than foreign equivalents on average.
2. Price as a weapon. This is the BYD playbook, copy-pasted into robotics. A $5,900 robot and a $99,000 robot are not competing in the same universe of buyers. Some Chinese humanoids now sell below $6,000 — a price point that turns "maybe someday" into "let's order a few and see."
3. The state has its thumb on the scale. Humanoid robots sit squarely inside the Communist Party's 2026–2030 five-year plan as a frontier technology, which translates into subsidies, procurement, and patient capital. A meaningful chunk of 2025's roughly $295M in humanoid orders came from state-owned enterprises — power plants, data centers, public venues. When your first big customer is the government, the demand problem solves itself. At least at first.
My read: The West tends to frame this as an AI race, and on AI "brains" the US genuinely leads. But humanoids are a hardware game first, and hardware at scale is the one race China has been training for since the 1990s. We may be watching the iPhone-vs-Shenzhen story play out again — except this time the bodies walk.
The Catch: Building Them Was Never the Hard Part
Now for the twist that keeps this from being a simple "China wins" story. Because here's the uncomfortable question buried under all those shipment charts: who is actually buying these robots, and what are they doing with them?
This is where it gets surprising. According to reporting from Fortune and the Associated Press in June 2026, a chorus of analysts thinks demand is lagging well behind China's ability to build. Most humanoids today are still — their word — performative rather than functional: great at dancing and boxing on stage, clumsy in the messy, unpredictable environments of real homes and warehouses. A New America researcher put it bluntly: the genuinely useful use cases are still narrow.
The economics are brutal too. Humanoids remain expensive to produce, fragile in operation, and reliant on highly structured settings to work at all. A lot of 2025's orders came from research labs and state-owned buyers — demand that's real but not exactly a self-sustaining consumer market. The Chinese government itself publicly warned about bubble risk in the sector, citing the gap between flashy demos and actual commercialization.
The most honest framing I've seen: China has solved the supply side of a market whose demand side doesn't fully exist yet. That's a remarkable position to be in — and also a precarious one. Factories that can make 62,500 robots a year need 62,500 buyers a year, eventually, or the whole flywheel stalls.
Bodies vs. Brains — the Split That Decides Everything
If you zoom all the way out, the global humanoid map resolves into a clean and slightly unnerving division of labor. China builds the bodies — the mass production, the supply chain, the data harvesting. The US leads on the brains — the high-level AI and the chips that run it.
Whoever you think wins depends on which half you believe is harder to replicate. Bodies are a manufacturing problem, and manufacturing problems tend to fall to whoever scales first. Brains are a research problem, and research advantages can evaporate overnight when the next model drops. Honestly? I wouldn't bet confidently on either side. But I'd watch the brains race very closely — because a cheap, capable body is only as valuable as the intelligence you can put inside it.
Read · The Brains SideThe 2025 Scorecard: China vs. the US
One table, no hedging — here's where the two sides actually stand:
| Category | China | United States |
|---|---|---|
| 2025 shipments | ~11,000+ from top 2 alone | ~150 each (Figure, Tesla) |
| Global share (2025) | ~85% (Barclays) | Single digits by volume |
| Supply chain control | ~63% of components | Largely import-dependent |
| Pricing | $5,900–$13,500 (entry) | $20,000–$99,000+ |
| AI "brains" | Catching up fast | Clear lead |
| State support | Five-year plan priority | Mostly private capital |
| Biggest open question | Who buys them all? | Can it scale production? |
Compiled from Omdia, Barclays, Morgan Stanley, and company disclosures, 2025–2026. Figures are best-available estimates.
The Demos Were Never the Story
For three years we watched humanoid robots through the lens of viral video — the backflips, the dance routines, the carefully lit demos. The whole time, a quieter number was compounding in the background: units shipped. And on that number, the contest isn't close.
The open question for 2026 isn't whether China can build 62,500 humanoids. It's whether the world has 62,500 jobs for them to do yet.
"China didn't win the demo. It won the assembly line — and it's still waiting to see if that's the same thing."
Frequently Asked Questions
Estimates vary widely. China's Gaogong Industry Research Institute forecasts 62,500 units — a 650%+ year-over-year jump — while Morgan Stanley projects a more conservative ~28,000. Both figures are forecasts; 2025's confirmed global total was about 13,000 units, most of them Chinese.
By Omdia's 2025 ranking, Shanghai-based AgiBot led globally with around 5,168 units. Unitree shipped more than 5,500 full-body humanoids and reported a profit, while UBTECH leads on industrial deployments with its Walker S2.
China controls an estimated 63% of the humanoid supply chain, so key parts like actuators and sensors are sourced domestically. Morgan Stanley estimates that makes Chinese robots 20%+ cheaper on average. Entry models from Unitree start near $5,900–$13,500, versus $20,000 to $99,000+ for many Western units.
In manufacturing and shipments, decisively yes — China built roughly 85% of the world's humanoids in 2025 (Barclays). But the US retains a clear lead in the AI "brains" and advanced chips that control the robots. It's a bodies-vs-brains split, and which matters more is still an open question.
Demand. Analysts cited by Fortune and the AP warn that China can build humanoids faster than buyers can use them. Most are still better at performing than working, many 2025 orders came from state-owned enterprises, and China's own government has warned of bubble risk.
Comments
Post a Comment