The first wave of shutdowns of humanoid robot companies has arrived.
In 2025, the humanoid robot industry underwent a dual test of ice and fire.
On one hand, 57 billion yuan in financing poured in, nearly 30 companies rushed to go public, and large - scale orders worth hundreds of millions of yuan were frequently finalized. On the other hand, a star startup in Silicon Valley collapsed on the eve of mass production, a pioneer in collaborative robots "died" for the second time, and even the originator of the floor - cleaning robot applied for bankruptcy protection dejectedly.
After nearly three years of development, the gap between the leading and trailing companies among more than 100 Chinese humanoid robot enterprises has become vast. The first - tier companies are preparing for an IPO with concept - validation orders worth billions of yuan in hand. However, for those companies with stalled financing and difficulty in product implementation, as an industry observer put it, "they are actually on the verge of failure and are just maintaining a low - speed operation." Both technical experts and industry analysts are issuing the same warning: in 2026, the industry consolidation will accelerate. This is not a elegy for the losers but a mirror reflecting the true face of the industry: when the tide goes out, who is swimming naked? When capital shifts from "looking at demos" to "looking at mass production", who can truly survive?
The industry is becoming more polarized, and the elimination race has begun
In 2025, the humanoid robot industry experienced a cruel elimination race. Star companies that once stood in the spotlight fell one after another. The startup myths in Silicon Valley were shattered by reality, and even the industry pioneer that created the collaborative robot category was not spared. When the tide receded, the insurmountable gap between the enthusiasm for technology and the coldness of business finally emerged.
The most dramatic collapse occurred in November 2025. Benjamin Bolt, the founder of Silicon Valley humanoid robot startup K - Scale Labs, sent a long farewell letter to investors and customers: disband the team, refund all the money, and open - source all the technologies. This company had been established for less than a year but was once the focus of the industry - it successfully completed three rounds of financing, launched two products successively, and had orders worth more than $2 million. Its customer list even included Caitlin Kalinowski, the head of OpenAI's robotics business. However, on the eve of mass production, this star company collapsed. Bolt admitted in the letter that the financing failed and the cash flow was out of control. By November 2025, there was only $400,000 left in the account, and the funds were completely exhausted.
On the surface, the cause of K - Scale Labs' failure was simple - it ran out of money. However, its collapse revealed the deep - seated dilemma faced by Western humanoid robot startups: the lack of a complete local supply chain. Most startups simply cannot afford the threshold of mass production independently. In his post - mortem analysis, Bolt said bluntly that Chinese humanoid robot companies are rapidly rising with the advantage of the local supply chain. They plan to dominate the humanoid robot ecosystem, just as DJI did in the early days of the drone industry. This is not just a matter of "low - cost manufacturing"; the United States has lost the supply - chain ability to develop complex hardware from scratch.
Even more lamentable is the fate of Rethink Robotics. Founded in 2008 by Rodney Brooks, a legend in the robotics academic circle, this company created the entire collaborative robot category with its two products, Baxter and Sawyer, and was regarded as an industry pioneer. After its first collapse in 2018, its intellectual property was acquired by the German HAHN Group, and it once saw the hope of recovery. In September 2024, Rethink Robotics, which belongs to the United Robotics Group, restarted its US business and launched a new product line. However, its good fortune did not last long. In August 2025, this legendary company declared bankruptcy for the second time. Julia Riemenschneider, the former CEO, revealed that the products were launched in a hurry before they were ready, the sales were far below expectations, and the investors finally decided to withdraw their funds. Aldebaran, a social robot manufacturer also belonging to the same group, closed its doors in February 2025. The core assets of the once - popular Pepper and Nao robots were finally acquired by a Chinese company.
The situation in the companion robot field was also dismal. Embodied, a company focusing on children's social companion robots, announced its closure, and its flagship product, Moxie - an AI robot priced at $799 and designed specifically for children with autism - stopped operating accordingly. This little blue robot capable of emotional interaction had helped countless children in families learn emotional management. However, it completely relied on cloud services, and the company's collapse meant the permanent "death" of the robot. Videos of children crying while holding Moxie circulated online, exposing the vulnerability of the cloud - dependent model - when the company was unable to pay for the operation of the large - scale model, the valuable hardware in users' hands instantly became useless.
Even the once - most successful consumer - grade robot company was not spared. In December 2025, iRobot, the manufacturer of the floor - cleaning robot Roomba, applied for bankruptcy protection. This legendary company, founded by MIT researchers in 1990 and with cumulative sales of more than 50 million robots, will ultimately be acquired by its Chinese supplier, Shenzhen Picea Robotics. iRobot's decline was a long - term process: in 2022, Amazon offered to acquire it for $1.7 billion, but the deal fell through due to the obstruction of the EU's anti - monopoly regulators. After losing this lifeline, iRobot continued to lose money under the double blow of low - priced Chinese competitors and tariff policies. Colin Angle, the founder, called this outcome "a tragedy for consumers, the robotics industry, and the US innovation economy."
Even Boston Dynamics, a company with strong technical strength, was not immune to the pressure. At the end of 2024, this robot company, famous for its viral videos, laid off 5% of its employees. Robert Playter, the CEO, admitted in an internal letter that the company "was burning cash at a rate faster than its business progress."
These cases of collapse and layoffs together painted a real picture of the humanoid robot industry in 2025. IEEE Spectrum pointed out bluntly in its year - end review that compared with the capital influx and hype in this field, the actual progress achieved was unconvincing. The industry is shifting from the "story - telling" mode to the "accounting" mode.
"Unfortunate" companies have their own "misfortunes"
The fallen companies each have their own stories, but their "death certificates" show similar causes.
Reviewing these cases, we can clearly draw a "portrait of the death of robot companies": interrupted financing, failed commercialization, homogeneous products, and insufficient technical reserves - these four nooses have trapped most of the players who failed to get out of the elimination race.
Interrupted financing is the most direct cause of death. When K - Scale Labs collapsed, there was only $400,000 left in its account. After the failure of Amazon's acquisition plan, iRobot struggled to support itself with a $200 - million loan and finally went bankrupt because it was unable to repay nearly $100 million in debts to its suppliers. The humanoid robot industry is extremely capital - intensive. The cost of developing a basic large - scale model is astonishingly high, and the capital scale of small and medium - sized enterprises is far from sufficient to support it. When the financing window closes, these companies are like divers cut off from oxygen, and suffocation is only a matter of time. Ni Xianhao, the head of the industry research center at the Beijing Academy of Artificial Intelligence, observed a cruel rule: if a company's subsequent financing has been decreasing or even stopped in the past three years, it "may actually be on the verge of failure." In 2025, although hot money from the capital market was still flowing in, there was an obvious "leading effect" - Figure AI was seeking a new round of financing with a valuation of $39.5 billion, while second - and third - tier companies were struggling to survive.
Failed commercialization is a deeper - rooted problem. Currently, more than 50% of the so - called "commercial" orders for humanoid robots are essentially for public relations displays and data collection, rather than real productivity substitution. Wang Feili, an analyst at UBS Securities, pointed out that currently, no humanoid robot company has truly passed the concept - validation stage. Industrial enterprises need to observe the real feedback after the robots are put into use, including efficiency improvement, economic considerations, and whether customers will repurchase after a 3 - to 6 - month trial. Many failed companies remained at the stage of "having a cool demo" but could never answer a fatal question: why should customers buy?
Product homogeneity has turned the competition into a war of attrition. Some companies have focused on or produced too many repetitive products, targeting low - threshold scenarios such as guiding and performing, and their future is worrying. In the humanoid robot field in 2025, "being able to walk" is no longer a selling point. The backflips of Unitree's robots and the ability of Zhipu's robots to ride bicycles - these once - amazing capabilities are quickly becoming industry standards. When everyone is struggling in the "cerebellum" level (motion control), where is the real technological moat? The answer points to the "brain" - the AI module that controls the robot's intelligence. The current biggest bottleneck for humanoid robots is the "brain." Only after a breakthrough in this area will the "electric - vehicle moment" for humanoid robots arrive. Ironically, most humanoid robot manufacturers invest very little in the AI field and can only rely on the general large - scale models of technology giants such as Alibaba and ByteDance. As these technology giants enter the humanoid robot manufacturing field themselves, the prospects of these startups that rely on "borrowing brains" are becoming increasingly bleak.
Insufficient technical reserves are the last straw. Rethink Robotics' Baxter and Sawyer robots once created a new category with the concept of "safe collaboration." However, their technical route based on elastic actuators always had shortcomings in terms of accuracy and speed, which could not meet the needs of industrial customers. When competitors such as Universal Robots occupied the market with more mature technical solutions, Rethink's first - mover advantage disappeared. iRobot also fell behind in the technological iteration. When Chinese competitors offered functions such as laser navigation and AI obstacle avoidance at a lower price, the brand premium of Roomba was no longer valid. Although Boston Dynamics has top - notch technical strength, its hydraulic - powered Atlas humanoid robot has been abandoned by the times. In April 2024, it announced its retirement and started over with an all - electric platform. The cruelty of technological iteration lies in the fact that today's leader may be tomorrow's laggard, and the speed of technological evolution in the humanoid robot field far exceeds the R & D rhythm of most startups.
How to "swim" until the sea turns blue
On one hand, some companies are having difficulty in their operations, while on the other hand, the financing situation is still booming.
In 2025, there were 610 financing events in the Chinese robotics field, nearly three times more than the previous year, and the total financing amount exceeded 57 billion yuan. This "two - extreme" situation precisely shows that the industry is undergoing a profound polarization. Leading companies are continuously receiving capital injection, while trailing companies are being eliminated at an accelerated pace.
In 2026, this polarization will only become more intense. To survive, two hard conditions need to be met: obtaining continuous financing support and achieving repeat orders in industrial scenarios. The former is the prerequisite for "surviving," and the latter is the proof of "thriving." In 2025, there was a key change in the customer structure of humanoid robots - from universities and research institutions to industrial enterprises. Previously, universities were the main customers of humanoid robots, and they mainly used the robots for secondary development and in laboratory scenarios. Now, industrial enterprises have a higher recognition of humanoid robots, and automobile manufacturers have placed orders with robot companies one after another. This means that the industry is shifting from "selling to scientific research" to "selling to production," and the latter has much higher requirements for product reliability and economy.
Technological breakthroughs will determine when the industry will truly explode. According to public data, the global shipment of humanoid robots in 2026 is expected to be about 30,000 units, which is a relatively conservative estimate. However, institutions such as TrendForce predict that it will exceed 50,000 units, which is a more optimistic view. The difference between the two predictions lies in the judgment of the speed of technological breakthroughs - currently, the "brain" development of humanoid robots lags far behind.
Surviving companies need to follow several survival rules. First, find a differentiated positioning to avoid a war of attrition in low - threshold areas. Second, establish a real commercial closed - loop to turn orders from "demonstration purchases" into "production purchases." Third, maintain the ability to raise funds and strive to enter the "first - tier list" of investors in the trend of capital concentrating on leading companies. Fourth, accumulate real - scenario data and use the "data flywheel" to feed back the iteration of the AI model to achieve "becoming smarter with more service." Fifth, embrace industrial chain collaboration to reduce costs and improve the security of the supply chain in the wave of localization of key components.
The elimination race in 2025 was just the prologue. In 2026, implementation and consolidation will come at the same time. Those zombie companies that were still "maintaining a low - speed operation" in 2025 will be completely eliminated in the competition in the new year. Those companies that have truly established a foothold in industrial scenarios and built a commercial closed - loop will have the opportunity to weather the cycle and welcome the "electric - vehicle moment" for humanoid robots. As an industry analyst said: this marathon has just completed the first kilometer. The ultimate champion may not have emerged yet, but those exhausted runners have already started to drop out.
This article is from the WeChat official account "Rongzhong Finance" (ID: thecapital), author: Lv Jingzhi, editor: Wu Ren, published by 36Kr with authorization.