The first humanoid robot company died in mass production.
Embodied intelligence might be the hottest field in the past two years. However, no one expected that the first fall in this field would come so soon.
In November 2025, Benjamin Bolte, the CEO of K-Scale Labs (KSL), issued a long farewell letter: disband the team, refund all the money in full, and open-source all the technologies.
KSL is no unknown entity. It secured three rounds of financing right after its establishment last year, launched two products in succession, and received orders worth over $2 million, including an order from Caitlin Kalinowski, the head of OpenAI's robotics business.
Yet, such a star company at the forefront of the trend collapsed on the verge of mass production.
A very important reason is that it ran out of money for mass production. In that year, KSL only produced about 10 K-Bot prototypes, with each costing over $100,000.
KSL's downfall is not just a failed startup case. It reveals a long-existing rift within the European and American robotics industries:
After the lack of a complete local supply chain, the vast majority of startups do not have the ability to independently bear the threshold of mass production, let alone offer a competitive pricing plan.
As Benjamin wrote at the end of his farewell letter:
"Chinese competitors such as Unitree, Accelerated Evolution, Zhongqing Robotics, and Songyan Power have raised the funds needed for large-scale production, while their Western counterparts have hardly achieved real mass production and delivery."
With only $400,000 in cash reserves, how did the money dry up?
On November 5, 2025, Benjamin, the CEO of KSL, sent an email to pre-order users:
The company would lay off all employees, refund all the deposits for the K-Bot robots in full, and open-source all proprietary technologies at once. This email marked the end of this robotics startup that had been established for less than a year.
In the emails sent to investors and customers, Benjamin pointed out the direct cause of death: failed financing and an out-of-control burn rate. Although KSL still had a revenue of $2.9 million in 2024, by November 2025, there was only $400,000 left in the company's account, and the funds were completely exhausted.
▲ The internal letter is from The Humanoid Hub
In fact, KSL was not short of funds. Since this year, KSL has completed three rounds of financing:
February 2025 (Seed round): Approximately $4 million in financing, with a valuation of $50 million, mainly used for prototypes and underlying technologies.
April 2025 (Additional investment): Nat Friedman and Daniel Gross added $250,000 for the software ecosystem.
June 2025 (Strategic investment): A subsidiary of Zhejiang Taotao Vehicle Industry invested $2 million and provided supply chain support.
In total, KSL received $6.25 million in financing. For the field of embodied intelligence, where financing often reaches hundreds of millions of dollars, this amount is just a drop in the bucket. In the crucial third quarter of 2025, KSL attempted to raise another $10 - 15 million but failed completely.
Looking back, KSL's failure was due to both external environmental factors and its own strategic mistakes.
From an external perspective, embodied intelligence is rapidly entering a stage of "winner-takes-all," and capital's patience for small and medium-sized teams is disappearing.
According to Crunchbase statistics, as of September 2025, the total financing of global robotics startups has exceeded $8.5 billion, reaching a new high since 2021. However, the top 10 companies have taken more than 40% of the funds.
Within KSL, strategic misjudgment accelerated the collapse.
KSL originally planned to have two product lines: the K-Bot for the high-end market and the Z-Bot, a smaller and cheaper robot priced at less than $1,000, which was easier to enter the consumer market. Common sense suggests that the Z-Bot would be more likely to generate cash flow and easier to scale. However, in the end, the company stopped developing the Z-Bot and concentrated all resources on the K-Bot.
The turning point came from a mysterious VC whom he "highly admired." The VC told him that as long as the company received orders for 100 K-Bots, it could successfully raise $20 million in Series A financing, "because there are no similar products on the market."
This statement changed KSL's R & D direction, making the team bet on a very difficult closed-loop: using pre-orders to drive financing and financing to drive mass production.
However, the closed-loop was never truly formed.
In that year, KSL only produced about 10 K-Bot prototypes, with each costing over $100,000. The real challenges of mass production were yet to come. Transitioning from 3D printing to die-casting and cold forging is not only expensive but also requires high tooling investment and additional approvals. The participation of the supply chain partner Taotao Vehicle Industry did not fill the funding gap.
On the cost side, the team tried to save as much as possible. For example, they used LX16 actuators available on Amazon to quickly build prototypes. However, the average failure-free time of these devices is not high. Benjamin said, "A robot with 30 actuators may break one within two weeks." The engineering challenges of the mechanical structure, electronic control system, and joint modules were much greater than the team initially expected.
KSL had only about 10 engineers but was responsible for almost the entire stack of R & D - mechanical, electronic control, joint modules, simulation, and control software.
The complex R & D system, tight schedule, and lack of profit sources made the company completely dependent on continuous financing. During this period, Jingxiang Mo, the company's engineering director, left and founded Gradient Robotics, focusing on the physical AI direction in the United States, which also indirectly reflected the huge pressure on the original team in terms of technology and resources.
▲ The new startup dynamics of Jingxiang Mo
Even so, the market's interest in the K-Bot was actually not low. From May to early November this year, the K-Bot received more than 100 pre-orders, with an amount exceeding $2 million, including an order from Caitlin Kalinowski, the head of OpenAI's robotics business.
However, due to the high mass production cost, the company finally chose to refund all the money, which was equivalent to cutting off the last chance of survival.
Ultimately, a company with an excellent product prototype and a passionate team collapsed because it could not find a sustainable business model and financial support in the real world.
Can't compete on cost - effectiveness
On the surface, KSL's downfall seems like an ordinary startup failure: broken financing, difficulty in mass - producing prototypes, and running out of cash flow. However, the deeper reason is that China's manufacturing advantages in the embodied intelligence supply chain are greatly squeezing the living space of small and medium - sized European and American startup teams.
Benjamin mentioned this at the beginning of his post - mortem analysis:
Some Chinese companies (Unitree, Zhongqing Robotics) are rapidly rising and taking the leading position by relying on the local supply chain. They plan to dominate the humanoid robot ecosystem, just as DJI dominated the drone market in the early days.
Behind this is not simply "low - cost manufacturing" but a gap in the industrial chain that the United States cannot replace.
The investor Elad Gil has mentioned many times that the United States no longer has the supply chain ability to develop complex hardware from scratch. All hardware companies that can iterate rapidly have their core manufacturing in China.
The direct problem is that the United States lacks small - batch factories that can cooperate with startups for rapid iteration. There are neither flexible production lines that can start with 300 units nor a supply chain ecosystem that is willing to stockpile goods, advance materials, and share risks for a pre - A startup.
This forces teams like KSL to rely on 3D printing and outsourced workshops for production and prevents them from entering an economically viable mass - production stage.
This gap is not only reflected in the factories but also in more macro manufacturing data. According to the latest report from the IFR, in 2024, about 295,000 industrial robots were installed in China, accounting for 54% of the global new installations. In the same period, the United States added about 34,000 units, only one - tenth of China's.
The installation volume essentially reflects a country's robot usage level, manufacturing ability, and industrial scale, which are the infrastructure that humanoid robots rely on.
The supply chain itself is also constantly evolving. The localization rate of mid - and low - end components has reached 80% - 90%. Many Chinese companies can provide a complete set of supporting components, including structural parts, wiring harnesses, motors, drivers, and joint modules.
China is also narrowing the gap in high - end components, which have long been considered a bottleneck.
Take the harmonic reducer as an example. Transmission error is a key indicator - the smaller the error, the more stable and precise the joint. In the past, Japan's Harmonic Drive maintained a transmission error in the range of ±20 - 30 arc - seconds, which was the industry's ceiling. Green Harmonic Drive, through the "triple - harmonic" approach, has reduced the transmission error to within ±30 arc - seconds, enabling domestic products to enter the mainstream level for the first time.
With the advancement of mass production, the unit price of Green Harmonic Drive's products has dropped from 8,000 yuan in 2019 to about 3,500 yuan in 2024, directly reducing the overall cost of humanoid robot joint modules by 25%.
A similar trend is happening with servo motors. Unitree's self - developed high - torque - density motor costs about one - tenth of Boston Dynamics' hydraulic system. Ubtech has reduced the price of its bipedal robot to around 100,000 yuan through self - developed servo steering gears.
Bank of America estimates in its report that if mainly using Chinese - made components, the BOM cost of a humanoid robot can be about $35,000. If relying mainly on overseas components, it will remain stable at $50,000 - $60,000. With the advancement of scale, the cost is expected to further drop to $13,000 - $17,000 from 2030 to 2035.
This gap will ultimately be reflected in the market. Currently, most humanoid robots in the United States and Europe are still priced at $20,000, $50,000, or even more than $100,000. Even the pre - sale price of the K - Bot, which emphasizes "cost - effectiveness," is as high as $8,999, about 64,000 yuan.
In contrast, Chinese manufacturers have reduced the prices of some products to $1,400 and $5,000. For example, Songyan Power's robot is priced at less than 10,000 yuan, and the starting price of the Unitree R1 is 39,900 yuan.
These products are not completely comparable to full - sized European and American humanoid robots in terms of size, load, and positioning. However, "lower prices than international competitors" is an obvious fact.
Behind this are the differences in the two industrial structures. Chinese manufacturers can push robots into lower price segments because of a complete, dense supply chain that can continuously reduce costs and cycles. In the United States, the lack of a supply chain makes embodied intelligence a high - cost, heavy - asset game that only a few companies can participate in.
In this sense, KSL's failure is not an accident of a single company but a direct reflection of this structural difference.
This article is from the WeChat official account "Silicon - based Observation Pro". Author: Lang Lang. Republished by 36Kr with permission.