Los Humanoid robots They have gone in a very short time from being a laboratory experiment to becoming one of the main drivers of the global technological race. Governments and companies around the world place them at the center of their automation strategiesHowever, many doubts remain about its real viability beyond the spectacular videos circulating on social media.
While China steps on the gas with mass training centersWith a vast industrial ecosystem and over 150 companies focused on this technology, Europe—led by Germany—is making moves to secure its place in the future value chain. At the same time, developers and consultancies are beginning to recognize that the Commercial discourse has gone ahead of what these robots are capable of doing today in real environments.
China makes humanoid robots a strategic priority
In China, the bet on the humanoid robotics It has been fully incorporated into the political agenda. The concept of [it] already appears in the planning documents for its next five-year plan. "embedded artificial intelligence"That is, AI systems integrated into hardware such as robots, autonomous vehicles, or drones. For Beijing, these humanoids are a tool for combat the aging of the populationto maintain industrial competitiveness and strengthen its position in the technological competition with the United States.
The country isn't just betting in theory: over the last five years it has recorded around 7.700 patents related to humanoid robotsa figure far higher than that of the United States in the same period. Added to this is a manufacturing capacity that is difficult to match, allowing it to produce at much lower costs and generate a global dependence on its supply chain.
According to estimates cited in financial reports, if a manufacturer like Tesla were to attempt to assemble its Optimus robot without using Chinese suppliers, the unit cost would skyrocket ranging from around $46.000 to over $130.000. This difference illustrates the extent to which the "world's factory" holds all the cards in this new sector.
Furthermore, China already clearly dominates the classic industrial roboticsIt accounts for around 51% of all robots installed in factories worldwide and adds approximately 280.000 new units per year, with the aim of continuing to automate to offset rising wages and maintain its position against competitors such as India or Bangladesh.
This existing ecosystem is now being leveraged to advance humanoid robotics, which authorities see as the next layer of automation: bipedal machines capable of working in warehouses, assembly lines, services and, in a still uncertain future, in homes.
Schools for robots: the Chinese campus where they learn human tasks
One of the most striking examples of this strategy is Phase II of Beijing Humanoid Robot Data Training Centerin the Shijingshan district of Beijing. There, a kind of “school” for humanoid robotswhere various models learn tasks ranging from logistical work to domestic functions, in spaces that quite faithfully mimic real life.
The complex is spread over two floors and It recreates production lines, logistics areas, kitchens, and bedrooms.Each zone functions as a modular cell that can reconfigure quickly to simulate different processes: today a packaging station, tomorrow an industrial kitchen or a hotel room that needs cleaning. The idea is that the robots perform tasks completed such as sorting reels, selecting packages, preparing food, or tidying a room, and not just isolated movements.
Behind this approach is the need to collect large amounts of real dataEvery gesture, every mistake, and every correction is recorded and later used to refine the control algorithms, improve walking stability, hand-eye coordination and the ability to react to changes in the environment.
One of the pillars of the method is the intensive repetition of movementsTo perfect a single action, a robot may repeat it around 1.250 times. This volume of practice translates into enormous databases that feed machine learning models, in a process reminiscent of sports training, but applied to machines.
According to Zhu Kai, director of this center, Each robot is supervised by two human trainers.His goal is for humanoids to develop what he calls "functional intelligence": that they not only stand and walk, but also integrate vision, balance, and manipulation in a coordinated manner in very diverse scenarios, bringing them closer to a minimally useful performance outside the laboratory.
Kuafu and other Chinese humanoids in intensive training
Within these facilities, the figure of KuafuA humanoid robot, approximately 1,65 meters tall, serves as a "model student" for much of the testing. This robot and others like it face constantly changing environmentswhere they must combine bipedal locomotion, manipulation of small objects and tasks that require a certain precision, such as stacking products or handling delicate parts.
The stated aim of the program is to train a new generation of machines capable of integrate into factories, warehouses and even homestaking on highly repetitive, physically demanding, or dangerous tasks. In theory, these humanoids wouldn't replace people all at once, but rather... complement human work where it is less safe or profitable to maintain a manual job.
However, the very scale of the project fuels an inevitable debate about the future of employment and coexistence with these machinesIf robotization advances at the pace pursued by Beijing, it is worth asking what types of jobs will be displaced, what new profiles will emerge, and how the social impact will be managed in a labor market already facing tensions due to automation.
The Chinese strategy also includes an aspect of technological marketingRobots from companies like Unitree and Deep Robotics have been seen at media events: from specialized sporting competitions to dance performances and demonstrations at massive concerts. These displays aim to familiarize the public with the presence of bipedal robots and project an image of a country at the forefront of technology, even though many of the routines shown are carefully choreographed and far from everyday use.
In parallel, other testing grounds are proliferating, such as the so-called "robotics Olympics" held in Beijing, where humanoid robots from various countries share the stage. These events reinforce the narrative that the humanoid will be the next big household appliance, a message that is still very much yet to be verified in practice.
A Chinese ecosystem with over 150 companies and increasing production
Beyond the training centers, the country has woven a very dense business network around this technology. It is estimated that more than 150 companies focused on humanoid robots are already operating in China, although only a few concentrate most of the investment and cutting-edge developments.
Among them is Unitreewhich is preparing its IPO with a potential valuation of nearly $7.000 billion. The firm has developed several models, including the H2 humanoid, designed both to demonstrate advanced movement capabilities—such as dance routines—and for manipulation tasks. Another key piece of the ecosystem is UBTech Robotics, manufacturer of industrial and commercial humanoids, whose Walker S2 robot is designed to operate virtually continuously thanks to an automatic battery changing system.
UBTech, which is listed in Hong Kong, recently raised capital to finance its expansion and increase production. Its plans include delivering approximately hundreds of industrial robots in the short term and scale up to approximately 5.000 humanoid units in 2026 and 10.000 in 2027. For its part, AgiBot It claims to have already rolled its 5.000th humanoid robot off the production line, which gives an idea of the pace some players want to set for the market.
The focus isn't limited to purely robotic companies. Electric vehicle manufacturers like Xpeng They have also joined the race with projects like Iron, a second-generation humanoid they aspire to mass-produce in the coming years. All of this aligns with the objective expressed by various analysts that China could become the main global market for these robotsboth in terms of volume and early deployment.
Financial consultants project a potential market of several trillion dollars By 2050, China is projected to account for more than half of global demand. These kinds of projections fuel investor interest, but also the risk of overheating if the actual capabilities of humanoids fall short of what has been promised.
Along with the strength of its industry, the country has additional advantages: a very deep supply chainLocal subsidy programs targeting robotics companies and prior experience in rapidly scaling sectors like electric vehicles are key advantages. Some manufacturers, such as UBTech, anticipate annual cost reductions of between 20% and 30% as they increase volumes and optimize processes.
Europe responds: German alliance to industrialize humanoids
In Europe, the most significant movement in recent months has come from Germany, where the industrial group Schaeffler and the robotics company NEURA Robotics They have forged a strategic alliance to bring humanoids from the laboratory to the large-scale productionThe joint objective is to accelerate the development and manufacturing of the next generation of cognitive robots capable of working side-by-side with people in demanding environments.
Schaeffler, known for its motion solutions and precision components for the automotive and other industrial sectors, brings decades of experience to scalability and industrialization. NEURA RoboticsFor its part, it has specialized in what it calls "physical AI," an approach that combines advanced perception, planning, and control so that the robot can interact with their environment more autonomously.
The agreement stipulates that NEURA's humanoid robots will be gradually introduced gradually in Schaeffler's own factories Distributed throughout the world, with an eye toward achieving a significant presence on assembly lines by 2035. The idea is that these systems will handle repetitive or heavy-duty tasks, taking advantage of their ability to change tasks more flexibly than a fixed industrial robot.
One of NEURA's star models is the humanoid 4NE1designed to share workspace with human operators. With this alliance, the company expects to significantly increase the production volume of these types of platforms, moving from prototypes and small series to thousands or even millions of units if costs allow.
NEURA's management has even set the manufacture of up to five million humanoid and cognitive robots by 2030These are very ambitious figures that, according to those in charge, can only be achieved with the support of an industrial partner with the scale and know-how of Schaeffler.
Neuraverse and the European commitment to physical AI
Beyond the hardware, the German collaboration focuses on the value of industrial dataNEURA manages a platform called Neuraversedesigned to aggregate the information generated by its robots during daily operations. By deploying them in real Schaeffler plants, the company hopes to feed this ecosystem with authentic production settingsvery different from the controlled environments of laboratories.
This constant flow of data would allow for the training of AI models that improve capabilities such as 3D perception, handling unforeseen events, and safe collaboration with people. In this way, each deployed robot will also act as learning source for the rest of the fleet, accelerating the development of new functions without the need to reprogram each unit from scratch.
For Germany, agreements of this type are interpreted as an attempt to reaffirm its position as an industrial power at a time when global competition is intensifying. Humanoid robotics is thus entering a stage where we are no longer just talking about spectacular prototypes, but about large-scale phased implementation in European factories, although currently in the pilot phase.
This strategy also seeks to send a message to the rest of the continent: if Europe wants to have a say in the next wave of automation, it will have to combine AI research, industrial expertise and proprietary supply chainsinstead of relying solely on technologies imported from the United States or Asia.
Although mass deployments have not yet materialized in Spain or other European countries, logic suggests that, if the pilot programs in Germany are successful, some of these solutions could eventually reach [the UK/European countries]. automotive and logistics plants in the rest of the Union, where there is already a significant base of traditional robotics on which to build.
Teleoperation, security and the risks of physical training
One of the most common techniques for teaching complex movements to humanoids is the teleoperationIn this scheme, a person directly controls the robot using specific controls or even with a suit that captures body movement. The humanoid It imitates those gestures in real time. and the records are then used to learn by imitation and are refined with other learning algorithms.
This method accelerates the development of skills such as walking, hitting an object, grasping tools, or performing coordinated sequences. However, it also involves certain security risks When the machine and the operator share the same physical space, a miscalculated kick, a sharp turn, or a timing error can turn a simple test into an accident, especially when the robots weigh tens of kilograms.
An illustrative example has been the dissemination of training videos of models such as the Unitree G1where the robot is seen replicating martial arts movements. In one of these rehearsals, a change in the trainer's posture causes the robot to throw a kick without discerning what's in front of it, resulting in the operator being hit. The scene has served as a reminder that working so close to powerful machines It requires strict protocols and highly reliable emergency shutdown systems.
Business consulting experts estimate that for every $100 invested in developing robots for the workplace, Up to 80 could be allocated to ensuring security and avoid harming people. From this perspective, the great challenge is not just for the humanoid to perform a task, but for it to perform it consistently without becoming an added risk on the production line.
In parallel, the developers face a recurring technical difficulty: reproducing human movement hands and fingers with sufficient precision. Many robotic hands still lack the degrees of freedom required to manipulate various objects as a person would, which limits the types of jobs they can take on in uncontrolled environments.
Inflated expectations, expensive prototypes, and the risk of a bubble.
As the industrial ecosystem prepares to scale, a segment of the sector has begun to recognize that the media narrative about humanoids It has been several steps ahead of the available technology. At recent conferences in the United States, engineers from companies like Tesla and Agility Robotics have admitted that market enthusiasm has outpaced the actual capabilities of their machines.
Videos that go viral—robots running, dancing, or moving with apparent ease—usually correspond to highly scripted demonstrationswhere the environment is controlled down to the millimeter. In the day-to-day reality of a factory, with long shifts, dust, temperature variations, and changing tasks, the picture is quite different: many humanoids remain in validation phase and they have not yet demonstrated that they can work profitably for hours without continuous intervention from specialized technicians.
Another weak point lies in the current costs of advanced prototypesVarious analyses place the price of leading-edge units between $150.000 and $500.000 per robot, figures that make them uncompetitive compared to human labor or specialized industrial robots. For the economic model to make sense in sectors such as logistics or manufacturing, this cost would have to be lowered to a much more modest range, around $20.000-$50.000.
In addition to the purchase price, reliability also comes into play: platforms like Apptronik's Apollo, to name one example, offer autonomous communities around four hoursfar removed from a full work shift. Added to this is the complexity of maintaining machines with dozens of joints, motors, and sensors, which increases the cost of each effective hour of work.
In China, the National Development and Reform Commission itself has reached warn of the risk of a bubble in the humanoid robot marketGiven the avalanche of new companies and the proliferation of similar products, the fear is that the combination of media hype, investor expectations, and pressure to bring something to market will generate a wave of unhelpful, expensive, and unreliable humanoids, potentially leading to a sharp correction and hindering innovation in the medium term.
The United States and Europe, between vertical integration and caution
In contrast to China's strategy based on volume, cost, and subsidies, US companies are opting for a model of vertical integrationThe idea is to control key parts of the value chain, from the actuators that move the joints to the artificial intelligence software that governs the robot, with the aim of ensuring higher performance, stronger safety arguments, and defensible intellectual property.
This approach has clear advantages, but it also involves developing many components in-house that other countries acquire in a more fragmented market. The result is a perhaps slower deployment pace, but with the hope of to differentiate themselves through quality and technological control against rivals focused on quickly reducing hardware costs.
In parallel, both Washington and Brussels are studying specific regulatory frameworks for advanced robotics. In the United States, there is even talk of executive orders defining the rules of the game, while in Europe the conversation intersects with regulations on AI, machine safety, and civil liability. The key will be finding a balance that protect workers and users without stifling the experimentation necessary for the sector to mature.
Analysts predict that, over time, the Chinese and US markets will end up converging in sizewith widespread adoption in homes expected from 2040 onwards if these platforms can be made more affordable and robust. Until then, most deployments will remain focused on very specific sectors and controlled pilot programs.
Meanwhile, European countries like Spain are watching these developments with interest. Although the local focus is currently more linked to the assistive robotics, exoskeletons and specific automation For generalist humanoids, the evolution of the sector in Germany and China will largely determine future opportunities for industry and research centers on the continent.
The picture that is emerging around the Humanoid robots It is a technology with enormous potential but still fraught with unknowns: China is driving its deployment with massive training centers, a very broad business network and strong political support; Europe and, in particular, Germany are seeking their own niche by betting on industrialization and physical AI; the United States is trying to differentiate itself through vertical integration and control of key components, while experts from all over the world remind us that, to this day, the gap between what demonstrations promise and what these machines can actually do remains large, both in technical, economic and regulatory terms.