AgiBot Passes Unitree as China's Humanoid Shipment Race Gets Real

The News: AgiBot Takes the Shipment Lead

Smart Analytics Global says worldwide humanoid robot shipments reached approximately 19,100 units in the first half of 2026, up 272 percent year over year from roughly 5,100 units. Within that still-small but fast-moving market, AgiBot reportedly shipped about 8,400 units, equal to 44 percent global share. Unitree delivered about 5,900 units, or 31 percent share.

That makes AgiBot the top vendor by shipment volume for the period. It also puts the two Chinese companies together at about three quarters of the reported global market. SAG attributes AgiBot's growth to a wider product portfolio, including full-size bipedal A-series robots, compact X-series units, and wheeled G-series robots. Unitree remains the higher-profile company in many Western conversations, but the H1 data suggests that product segmentation and channel depth may be winning as much as brand recognition.

The timing is awkward for any simple Unitree victory narrative. Unitree has priced a Shanghai STAR Market IPO at a roughly $9 billion valuation and is raising about $900 million, according to multiple financial reports. Investor demand has been intense. Yet the same week investors were buying the category's most visible public-market proxy, a fresh shipment report said a rival had already passed it on units.

Why China Owns the Early Shipment Curve

The most important number in the report is not AgiBot's 44 percent share. It is the claim that Chinese vendors shipped more than 97 percent of the world's humanoid robots in H1 2026. China Daily and People's Daily, citing SAG, also reported that China represented more than 85 percent of global demand. That means the early market is not only China-made, it is mostly China-bought.

That domestic absorption matters. Humanoid robots are not smartphones. A supplier does not only need a factory and a distribution partner. It needs test sites, service crews, spare parts, safety procedures, on-site retraining, local procurement relationships, and customers willing to accept limited autonomy while the product improves. A large domestic market gives Chinese companies places to ship machines even before export channels become clean or geopolitically comfortable.

China's advantage also sits in the component base. Low-cost motors, reducers, batteries, castings, sensors, embedded compute integration, and contract manufacturing are easier to coordinate when the supply chain is near the robot maker. That does not solve dexterity, reasoning, battery endurance, or safety certification. It does let companies iterate faster through hardware versions, price lower, and offer a wider product ladder.

The Specs Comparison That Actually Matters

AgiBot and Unitree are not selling one cleanly comparable product. The shipment race blends different sizes, configurations, and use cases. That makes direct platform comparisons messy, but it also reveals why AgiBot's portfolio approach is important. A humanoid company can chase high-end factory deployments, developer labs, education customers, and event demos with different hardware rather than forcing every buyer into one flagship machine.

AI-generated image

The early humanoid market is being shaped by components, supplier loops, and service infrastructure as much as autonomy models.

Shipments Are Not the Same as Useful Work

The robotics market needs a stricter vocabulary here. A shipment can mean a paid industrial system, a research platform, an education unit, a demo robot, a distributor channel fill, or a machine delivered for a pilot that still needs heavy human supervision. Those are not equal signals. A robot that spends six months in a university lab and a robot that moves totes for two shifts per day are both units, but they tell investors and buyers very different things.

For humanoid robotics, the stronger metrics are operating hours, intervention rate, cycle time, injury-free work near humans, task completion rate, service calls per month, battery swaps per shift, and renewals. If the buyer pays again, expands the fleet, or names the task publicly, the evidence becomes more meaningful. If the robot ships into a channel with no disclosed utilization, the number is still useful, but it is an inventory and adoption signal, not deployment proof.

That distinction is especially important because the global shipment base remains tiny by industrial automation standards. A 272 percent year-over-year increase is real growth, but 19,100 units across every vendor and use case is still a small category. The story is the slope and the concentration, not the absolute scale.

The IPO Feedback Loop

Unitree's IPO is still highly relevant. A public listing gives the sector a price, a disclosure cadence, and a benchmark that other humanoid companies can be compared against. MarketWatch and the Financial Times reported extraordinary retail demand for Unitree shares, with the company raising roughly $900 million and carrying a valuation near $9 billion. That capital can fund AI model development, production lines, service teams, and inventory.

It can also intensify pressure on peers. If Unitree trades well, AgiBot, Galbot, LimX Dynamics, Leju Robotics, X Square Robot, and other Chinese firms have a stronger argument for public-market attention. If it stumbles, the category may still attract strategic money, but later listings will face sharper questions about margins, channel quality, and repeatable commercial work.

The United States has its own public-market path through Agility Robotics, which announced a SPAC deal valuing the company at $2.5 billion. That is a different kind of signal. Agility has emphasized warehouse work and named customers, while Chinese vendors have the early shipment curve. The next market phase will compare those evidence types: more units at lower cost versus fewer units with clearer commercial tasks.

What This Means for U.S. and European Robot Builders

Western humanoid companies should not read China's 97 percent shipment share as a final verdict. The category is too early. Software reliability, safety cases, customer integration, security, labor acceptance, and support economics may matter more than early unit volume. But they also should not dismiss it. Hardware learning curves compound. Every shipped machine can produce supplier feedback, repair data, customer objections, and integration lessons.

The U.S. also faces a policy complication. Recent FCC action around foreign-produced advanced robotic devices reframes humanoids as connected infrastructure and supply-chain risk. That may help domestic suppliers in sensitive sites, but it does not automatically create a low-cost American hardware base. If Chinese vendors can iterate faster and cheaper while U.S. companies spend more time on bespoke pilots, the gap may widen even if export restrictions slow direct sales into the United States.

Europe's path is different again. Companies such as NEURA, Humanoid, and Agile Robots are trying to tie physical AI to industrial customers, research networks, and regional manufacturing. Their advantage may be in safety culture and factory integration. Their risk is speed. The H1 data says the volume race is already moving quickly.

AI-generated image

Humanoid competition now depends on the full stack: hardware, data, service, security, and customer economics.

The 12-Month Outlook

The next twelve months should clarify whether AgiBot's shipment lead translates into durable commercial advantage. Watch for three signals. First, named industrial customers beyond launch events and state-media demonstrations. Second, fleet-level operating metrics, especially intervention rates and hours worked. Third, pricing and service terms that show whether the business can survive outside subsidized pilots and research demand.

For Unitree, the public listing raises the bar. The company will have more capital and visibility, but investors will expect product evidence beyond viral videos and developer sales. If IPO proceeds go into robot AI models and production capacity, the question becomes whether Unitree can convert technical popularity into industrial repeatability.

For the broader market, the SAG report turns humanoid robotics into a measurable race. The category still lacks clean standards for what counts as an active deployment. But the old phase, where companies could compete mainly through demos, is ending. The new phase will be argued through shipments, customer names, field data, safety evidence, and margins.

FAQ

Bottom Line

AgiBot's reported H1 shipment lead is the clearest sign yet that China's humanoid sector is no longer only a Unitree story. The market remains early, and shipment counts should not be confused with productive labor. Still, the concentration is striking: more than 97 percent of global shipments from Chinese vendors, roughly 75 percent controlled by AgiBot and Unitree, and a public-market funding cycle arriving just as the leaderboard changes.

The humanoid race is becoming less about who can make the most convincing demo and more about who can manufacture, ship, support, and improve robots in the field. On that score, the H1 data gives China the early volume advantage, while leaving the hardest question open: which of those robots are doing useful work often enough to justify the hype?