Inside a warehouse in California, wooden crates are stacked waist high, each holding a humanoid (人形机器人) built thousands of miles away in eastern China. The crates are not going to a single buyer. They are being split among universities, startups and research labs across the United States.
That flow — from a factory in Hangzhou to a logistics hub in America — is the quiet core of Unitree's overseas business. The company rarely opens its own stores. It ships, and lets local partners do the selling.
A US beachhead built by a partner
Unitree's American presence runs through local resellers rather than company-owned shops. One US partner described moving on the order of a hundred robots a month to roughly a thousand customers, a volume that suggests demand is broad even if the buyers are mostly institutions.
The appeal is price. Where Western humanoids from better-known names still carry six-figure price tags, Unitree's G1 retails for roughly US$16,000 in the United States, and its entry-level R1 model reaches consumers at a still lower cost. For a university lab or a robotics startup, that gap is the difference between one demo unit and a small fleet.
Japan and Korea through agencies and e-commerce
In Japan, Unitree works with two local distributors: TechShare, a Tokyo robotics hardware supplier, and GMO Internet Group, the internet conglomerate. In June 2026, GMO's robotics arm signed a formal agency agreement to sell and support Unitree's humanoids and quadrupeds in Japan, including software and maintenance.
South Korea followed a different route. Unitree began selling its entry-level R1 through the AliExpress marketplace, pricing it at about half of its higher-end G1, and built relationships with three local distributors to reach universities and corporate research teams. The company has not opened a Korean office, leaning instead on platforms and partners.
The money behind the momentum
Unitree's scale is no longer a rumor. According to its IPO prospectus, the company posted 2025 revenue of RMB 1.71 billion (≈ US$241 million / HK$1.88 billion), more than four times the prior year, with net profit of RMB 287.6 million (≈ US$40.5 million / HK$316 million). China accounts for about half of revenue; the rest comes from overseas, with North America its largest foreign market and Japan second.
In July 2026, regulators approved Unitree's initial public offering on Shanghai's STAR Market. Reporting tied to the listing put the offering price at about RMB 150.8 per share, implying a valuation near RMB 61 billion (≈ US$8.6 billion / HK$67 billion) and proceeds of roughly RMB 6.1 billion (≈ US$859 million / HK$6.7 billion) earmarked for R&D, new products and a production facility.
Capacity for a global order book
Unitree says it shipped more than 5,500 humanoid robots in 2025, with total production across models above 6,500. For 2026, a company executive told Nikkei Asia the plan is to at least double shipments, potentially nearing a triple, and to lift annual capacity toward 30,000 units, including a new factory with roughly 20,000 units of capacity funded by the IPO.
Crucially, more than 70% of Unitree's humanoid revenue still comes from research and education, with industrial use only about 9%. The company frames this as a developer-market phase: just as smartphones needed app builders, robots need a software ecosystem that has not yet formed.
Why not build overseas plants yet
All Unitree robots are made in China today. When asked about overseas assembly, the company's regional director said there are no concrete plans, but that if demand keeps rising it may consider assembling abroad — a hedge against tariff and logistics risk rather than a near-term commitment.
That restraint fits the model. Unitree competes on cost and volume, and keeps margins by not duplicating factories. The bet is that demand from researchers and early industrial users will mature before the need to localize production becomes urgent.
What Unitree is really proving
The lesson for observers is not that a Chinese robot is cheap. It is that a company can build a global distribution footprint with almost no owned retail, using local partners who already speak the language, handle warranty claims and understand certification. That is a different kind of moat — one made of relationships, not steel.
For global readers, the signal is that the humanoid race is being won first on the shelves of university labs and the floors of trade shows. Whoever fills those shelves cheapest and keeps them stocked may define the developer base the rest of the industry builds on.
Honest limitations
Overseas sales are real but concentrated in research and education, and Unitree's own figures show industrial applications at only about 9% of humanoid revenue, so "global bestseller" means labs and pilots rather than factories full of robots. The IPO valuation and proceeds are based on the prospectus and listing reporting, and final pricing and market reception at debut could differ. A mid-2026 US move to restrict new authorizations for foreign-made advanced robots, reported by industry outlets, could limit future Unitree models in America even though currently authorized units remain legal. Unitree's cost lead depends on Chinese manufacturing scale, and any shift to overseas assembly would test whether that advantage survives.
What readers can do now
- If you are buying humanoids for research, compare Unitree's total cost of ownership — hardware plus local support — against a US or European rival before assuming the cheaper box wins.
- Track Unitree's 2026 shipment and capacity numbers each quarter; a move past the research niche into industrial volume would be the real signal of maturity.
- For robotics startups, study the "ship from China, sell through locals" model as a template for reaching global users without building a retail footprint.