Chinese humanoid robots are getting cheaper. Can Europe compete?
Unitree’s dramatic arrival on the Shanghai stock market has given investors a new way to bet on humanoid robots—and exposed how difficult the industry is to value.
The Chinese robotics company raised approximately 6.1 billion yuan, or $900 million, when it listed on Shanghai’s STAR Market on 19 August. Its shares closed 460% above their issue price on the first day, briefly valuing Unitree at around $50 billion. They subsequently fell by approximately 45% from that closing level as questions resurfaced about inflated valuations, intensifying competition and the limited commercial deployment of humanoid robots.
Behind the stock-market spectacle lies a more consequential development for the global robotics industry: the price of a humanoid robot is beginning to look surprisingly ordinary.
Unitree sells its 1.23-metre-tall R1 Air for $4,900, excluding tax and shipping. The standard R1 starts at $5,900, while the larger G1 humanoid robot is available from $13,500.
These are not machines that can arrive at a factory on Monday and independently join the production line on Tuesday. Yet their prices mark an important turning point. Humanoid hardware that until recently belonged almost exclusively in well-funded laboratories is moving within reach of universities, developers and corporate research teams.
For European robot manufacturers, the uncomfortable question is whether they can compete as China turns humanoids into increasingly affordable, mass-produced machines.
The answer depends on what competing means. Europe is unlikely to beat China in a straightforward humanoid robot price war. But the cheapest robot is not necessarily the one that delivers the lowest cost per completed task.
How much does a Unitree humanoid robot cost?
Unitree’s R1 is an impressive piece of engineering for its price. Depending on the version, it has between 20 and 40 degrees of freedom and weighs approximately 27 to 29 kilograms. Unitree says it integrates a large multimodal AI model that can process voice and images.
However, the headline price describes the most basic R1 Air. That version has 20 degrees of freedom, a monocular camera and no dexterous hands. Its battery life is approximately one hour.
The standard $5,900 R1 increases the number of joints to 26 and adds a binocular camera, but neither of the two lower-priced versions is intended for secondary development. That capability is reserved for the separately priced R1 EDU, which can also be configured with additional computing power and dexterous hands.
The distinction substantially changes what the robot can be used for. A machine that can perform preconfigured movements is not the same product as an open research platform on which developers can build and test new physical AI applications.
Unitree itself says the global humanoid robot industry remains in the “early stages of exploration” and urges individual users to understand the limitations before purchasing one.
That warning is important. Videos of robots dancing, boxing or performing acrobatics show genuine advances in balance and motion control. They do not prove that a robot can complete commercially useful work for eight hours a day.
Unitree shipped more than 5,500 humanoid robots in 2025, according to its IPO prospectus. However, the prospectus also shows that most customers are universities and research institutions. Its limited industry-related revenue has come from uses including corporate reception, tour guiding, intelligent manufacturing and inspection.
In other words: their humanoid robots are selling, but are not (yet) replacing large numbers of workers.
Why are Chinese humanoid robots so cheap?
Unitree’s cost advantage does not rest on one breakthrough component. It comes from an industrial ecosystem that is difficult to recreate elsewhere.
A humanoid robot requires electric motors, gear reducers, bearings, encoders, batteries, cameras, processors, control electronics and precision-machined structures. Many of these technologies are already produced at enormous scale in China for electric vehicles, smartphones, drones and industrial automation.
That overlap gives Chinese robot manufacturers access to experienced suppliers, production equipment and engineering talent without requiring them to establish an entirely new supply chain.
It also reflects a wider policy choice. As MTN reported earlier, China has placed AI-powered robotics at the centre of industrial strategy, combining embodied AI development with support for advanced manufacturing and domestic supply chains.
Bank of America Securities estimated that the average bill of materials for a humanoid built largely with Chinese components could fall from around $35,000 in 2025 to $17,000 by 2030. It estimated that a comparable machine sourcing its major components outside China could cost considerably more. Such forecasts are uncertain, but the difference illustrates the scale of China’s manufacturing advantage.
Actuators give Unitree a manufacturing advantage
Actuators are particularly important. These compact systems combine motors, gears, bearings, encoders and controllers to move a robot’s joints. They are among the most technically demanding and expensive parts of a humanoid robot.
Unitree designs and produces many of its core motors, reducers, controllers and joint modules itself. This vertical integration reduces its dependence on specialist suppliers and allows the robot and its components to be redesigned together.
If a joint needs to become lighter, stronger or cheaper, Unitree can alter the actuator alongside the rest of the machine rather than wait for a supplier to adapt an existing product.
China’s production volume reinforces the advantage. Larger manufacturing runs spread engineering and tooling costs across more machines. They also put more robots into the hands of researchers, generating feedback that can be used to improve subsequent models.
Western humanoid developers have often started with highly capable but expensive prototypes and considered volume manufacturing later. Unitree has treated manufacturability and price as design requirements from the outset.
Unitree is profitable, unlike many humanoid robotics developers, but its results also expose the pressure created by cheaper hardware. The company’s move towards the lower-priced G1 reduced gross margins, while adjusted profit fell by more than half in the first quarter of 2026 as spending increased and domestic competition intensified.
Falling prices can accelerate adoption, but they also force manufacturers to sell more units and invest heavily in the software that turns those units into useful machines.
Humanoid robot hardware still needs an AI brain
Affordable hardware solves only part of the humanoid robot problem.
A robot can be trained to perform a backflip in a controlled environment without being able to pick an unfamiliar object from a cluttered shelf. The second task requires perception, reasoning, force control and the ability to recover when the object slips or another person enters the workspace.
This is the gap between locomotion and useful autonomy.
Unitree founder Wang Xingxing acknowledged in August that the company was lagging in the real-world application of physical AI models. He said humanoids were not yet capable enough for mass deployment and suggested the industry’s equivalent of a “ChatGPT moment” could still be between two and ten years away.
Large AI models have improved robots’ ability to interpret instructions and recognise objects, but physical work is less forgiving than generating text. A language model can produce an imperfect answer without causing physical damage. A robot’s mistake can destroy a product, stop a production line or injure someone.
Industrial humanoid robots must also withstand dust, vibration, temperature changes and thousands of hours of repetitive movement. Today’s low-cost humanoids have not yet demonstrated that level of durability at scale.
The relevant business metric is therefore total cost of ownership. Integration, programming, supervision, maintenance, spare parts, energy consumption and downtime all contribute to the cost of each successfully completed task.
A $5,000 robot that needs constant assistance may offer less value than a $100,000 machine that works safely and reliably.
How European humanoid robot manufacturers can compete
Europe has no direct equivalent to Unitree’s combination of low pricing and manufacturing volume. It does, however, possess a substantial robotics, automotive and industrial-engineering base.
Germany’s NEURA Robotics is developing 4NE1, a full-size humanoid aimed at manufacturing, logistics and eventually domestic assistance. The company says the robot combines full-body sensing, adaptive control and safe human interaction.
In June 2026, NEURA announced Series C financing of up to $1.4 billion to expand its physical AI platform and manufacturing capacity. As MTN reported when the financing was announced, the round brought together investors and industrial partners including Tether, Qualcomm Technologies, Amazon, NVIDIA, Bosch, Schaeffler and the European Investment Bank.
It provided a rare demonstration that a European robotics company can attract capital on a scale approaching that of leading US and Chinese developers. Because NEURA described the financing as being worth “up to” $1.4 billion, however, the figure should not necessarily be interpreted as capital already received in full.
Barcelona-based PAL Robotics has spent more than two decades developing humanoid and service robots. Its TALOS platform is primarily an advanced research machine, while KANGAROO is intended to give physical AI developers access to reliable hardware, open software and simulation tools.
Europe can also draw on suppliers such as Bosch and Schaeffler, as well as specialist expertise in sensors, actuators, machinery and factory integration.
Its weakness (as ever, it seems) is fragmentation. Robotics expertise is distributed across countries, universities, startups and industrial groups, while converting successful research into high-volume manufacturing remains a persistent European problem.
Could European robot safety become a competitive advantage?
European regulation is often portrayed as an obstacle to innovation. For industrial humanoid robots, it could also support a valuable market position.
A robot working beside people must do more than move impressively. Manufacturers need to demonstrate predictable behaviour, emergency stopping, cybersecurity, safe force limits and compliance with machinery rules. Customers will also want clarity about where operational data is stored and how software updates affect the robot’s behaviour.
These requirements increase development costs. They may also produce machines that European factories are more willing to trust.
Geopolitics could reinforce that position. In July, the US Federal Communications Commission added foreign-produced advanced robotic devices to its Covered List. Such devices will generally be unable to obtain the FCC equipment authorisation required for new products unless they receive conditional approval from the US Department of War.
The measure does not automatically revoke authorisations already granted to existing models. Nor does it apply specifically to Unitree. Nevertheless, it demonstrates how connected robots are becoming part of a wider debate over cybersecurity, critical technology and supply-chain dependence.
European customers may eventually place a premium on robots built, maintained and governed within Europe.
Europe can compete without winning the price war
China has already changed the economics of humanoid robotics. Affordable platforms will put more robots in laboratories, generate more operational data and help Chinese manufacturers improve faster. Europe cannot assume that excellent research or strict standards will compensate indefinitely for lower manufacturing scale.
It does not, however, need to produce a €5,000 humanoid simply because Unitree can.
Europe’s opportunity lies in turning robotics expertise into machines that solve expensive industrial problems: labour shortages, hazardous maintenance, repetitive handling and flexible low-volume production. That requires reliable hardware, capable software and close integration with the factories where the robots will work.
Unitree has shown how quickly the price of a humanoid body can fall. The next stage of the competition will be decided by who can make that body perform useful work—and keep doing it after the promotional video ends.
Liked this article? You can support our independent journalism via our page on Buy Me a Coffee. It helps keep MoveTheNeedle.news focused on depth, not clicks.