On the brink of collapse thrice, staging comebacks thrice: A robotics company's 15-year "extreme survival"
How can a robotics company cross the industry's "Valley of Death"?
The office of Li Tong, the CEO of Keenon Robotics, seems completely unrelated to the robotics industry. It is neither intelligent nor cool but features a minimalist style. The desks, chairs, and guest sofas are all in the IKEA style, and the side tables are made of tin. The employees' workspaces also have a "construction - site minimalist" look, with grey tiled floors and no partitions.
As a commercial robotics company established 15 years ago, Keenon Robotics has experienced several lows and highs in the robotics industry. Currently, Keenon Robotics has sold over 100,000 commercial robots, covering scenarios such as delivery, cleaning, and greeting. IDC data shows that in 2024, Keenon Robotics ranked first in the global commercial service robot market with a 22.7% shipment share.
In 2025, with the popularity of DeepSeek and embodied intelligence, the robotics track has once again become the darling of capital, and several robotics companies are also preparing for IPOs.
Photography: Kong Yuexin
Since entering 2025, the most frequently discussed topic between Li Tong and the shareholders has been the IPO wave.
Songhe Capital is one of the investors in Keenon Robotics' Series A round. In the view of Wang Yang, the managing partner of Songhe Capital, there are currently three types of robotics companies. The first type is traditional industrial robotics companies, whose products are directly benchmarked against the "Big Four" (ABB from Switzerland, FANUC from Japan, KUKA from Germany, and YASKAWA from Japan). These companies have already been listed on the domestic A - share market. The second type consists of companies established between 2010 and 2020, which combine artificial - intelligence - related algorithms with robots but have not reached the level of general and embodied intelligence. They can perform some tasks to assist human labor and improve efficiency in specific scenarios. The third type is general - purpose embodied intelligence. "Now is exactly the opportunity for the second - type robotics companies to go public. If they miss this wave of opportunities, it will be very difficult for them to compete with the third - type robotics companies in the primary market," said Wang Yang. Keenon Robotics happens to be an embodied robotics company with both general - purpose and specific - purpose capabilities that has weathered multiple cycles.
"We hope that the robotics boom can last longer and not change too quickly," Li Tong said. "In technology entrepreneurship, especially in the robotics field, a large amount of technology iteration is required. In the past, technology R & D mainly took place in schools, and then companies would carry out industrialization. Now, companies are engaged in cutting - edge technology R & D, and we need more patience from everyone."
Keenon Robotics has also joined this wave of technological updates. In 2024, Keenon Robotics launched its first dual - arm embodied service robot. Since 2025, Keenon Robotics has officially released two humanoid service robots, namely the wheeled XMAN - R1 and the bipedal XMAN - F1.
Li Tong believes that with continuous technological iteration and innovation, even if developing humanoid robots leads to losses, investment must be made at this time. "I think a company needs to continuously innovate. Staying the same for a long time means going backwards, so we will keep changing."
"I dug a huge hole for myself"
In 2010, Li Tong and three friends pooled 200,000 yuan, rented a two - bedroom apartment in Shanghai, and founded Keenon Robotics. Due to financial constraints, they worked in this completely unfinished apartment without even a door for many years before moving to the current industrial park.
This year is widely regarded as the beginning of the mobile - internet era. During this period, companies such as Meituan (2010), Kuaishou (2011), Didi (2012), and ByteDance (2012) were founded one after another. The O2O, local - life - service, and APP - development sectors experienced explosive growth.
In contrast, the robotics track was the most inconspicuous industry at that time. Robotics entrepreneurship seemed neither trendy nor attractive, and few domestic venture capitalists paid attention to it.
In the early days of entrepreneurship, the Keenon team was constantly exploring the development direction in the robotics field, and they successively tried various sub - fields such as educational robots, sweeping robots, and humanoid simulation robots. Without any financing, the team had to take on projects first, "supporting themselves while moving forward."
Source: Respondent
Li Tong described himself as "fearless due to ignorance." "I didn't think too much at that time. After really starting the business, I realized that I had dug a huge hole for myself - without early - stage financing, it's quite difficult for a technology company to survive on its own. The investment was large, and the return was small. Several of my initial business partners 'couldn't hold on,' and the founding team almost completely disbanded."
But Li Tong didn't want to give up. "I didn't have the money to hire the best talents at that time. It was easiest to 'persuade' my classmates." His final solution was to invite alumni to start the business together. In the first few years, Li Tong and several classmates only received a monthly salary of a few thousand yuan to make ends meet.
After a lot of attempts, the Keenon team finally chose to commercialize in the service industry, with the catering industry as the first application scenario. In Li Tong's view, robots are essentially a form of labor. As a labor - intensive industry, the service industry will inevitably face a labor shortage in the future, and the growth space for service robots must far exceed that of other scenarios.
Yunqi Capital was the lead investor in Keenon Robotics' Series A round and also its earliest institutional investor. Chen Yu, a partner at Yunqi Capital, said that around 2016, Dianping had listed 4 million restaurants, there were about 400,000 hotels in the country, and there were about 10,000 large medical facilities including top - tier hospitals. These are three application scenarios of completely different magnitudes. Li Tong chose the scenario with the largest market size and the greatest challenges.
In 2013, Keenon Robotics launched its first - generation catering robot, "Xiaolang." Li Tong still remembers that as soon as "Xiaolang" was launched, the owner of a restaurant came to the company's door to "block" them.
"He was about to open his restaurant but couldn't recruit enough staff. After hearing about us, he immediately rushed over and directly transferred money to us via Alipay. At that time, the transfer limit of Alipay was 5,000 yuan, and he transferred 5,000 yuan several times. After the transfer, he immediately took the robot away to prepare for the opening," Li Tong recalled. This incident even made the news at that time.
This also made Li Tong's team realize that there is a demand for (service robots), and they can replace humans in work.
Compared with the prospects, the challenges are even greater. Compared with the industrial environment, the service - industry environment is more complex. Places like restaurants cannot be changed to accommodate robots like factories. Therefore, robots entering this scenario need to be able to perceive the environment, make autonomous decisions, and flexibly and effectively solve control problems. This is also one of the directions in which Keenon Robotics has been continuously optimizing and improving its technology to this day. The testing area in Keenon Robotics' office building fully reproduces the complex environment of restaurant operations, including simulated tests of various scenarios such as slope passage, passing through 40 - centimeter narrow passages, and moving on oily floors.
In 2015, Li Tong's team received a personal investment of 1 million yuan from his former boss at Microsoft.
During the same period, with the third wave of AI development triggered by breakthroughs in deep - learning technology, "Four AI Dragons" such as SenseTime and Megvii, as well as intelligent enterprises in the field of autonomous driving, emerged one after another. Robots with "AI attributes" also caught the attention of venture capitalists.
In mid - 2016, Wang Yang and Li Tong first met in a Shanghai restaurant that used Keenon Robotics' robots. At that time, most service robots still relied on "tracks" or scanned QR codes on the ground to move along fixed paths, but the Keenon Robotics' robots in the restaurant at that time could move indoors using autonomous navigation. Wang Yang was very interested in this and found Li Tong's team through the product.
On Children's Day of the same year, while shopping in a Shanghai mall, Chen Yu noticed a group of children interacting with a Keenon Robotics' guide robot named "Huasheng." After watching for a while, he thought the robot's movement and interaction were very interesting. After finding the name of Keenon Robotics behind the robot, he searched for Li Tong's contact information online and sent him a text message to arrange a meeting.
At the end of 2016, Keenon Robotics completed a Series A financing of tens of millions of yuan, which was also its first round of institutional financing. The investors were Yunqi Capital and Songhe Capital.
Two desperate situations and turnarounds
After the Series A financing, Li Tong's team quickly concentrated the funds on talent development and product iteration.
But in 2018, when Li Tong decided to start a new round of financing, the capital market changed dramatically. The artificial - intelligence field encountered a serious trust crisis, and the robotics field was also affected.
Keenon Robotics was in a difficult situation of failing to secure financing. "At that time, investors judged technological innovation based on the rhythm of mobile - internet model innovation. So by 2018, they thought all AI - related companies were frauds and none of them could achieve commercialization. The entire capital market didn't trust technology entrepreneurs. It was really like being discarded like a worn - out shoe," Li Tong said.
Chen Yu also admitted that the robot market had not been well - educated and popularized at that time.
An employee of Keenon Robotics told "China Entrepreneur": "In 2019, the company really had no money left in its account. If it couldn't raise more funds, the company would go bankrupt." According to people close to Keenon Robotics, in early 2019, the company's old shareholders made a new round of investment. Li Tong also sighed that he was really lucky that the old shareholders "saved the day."
Chen Yu said that their decision to invest in this round was made at the end of 2018 for two reasons. First, Keenon Robotics had started building a mass - production system more than a year in advance. Second, in October of that year, the smart restaurant of Haidilao in Beijing opened, and there were subsequent procurement orders for hundreds of Keenon Robotics' robots, which verified the company's ability to achieve large - scale commercialization.
"Previously, the penetration rate of robot food delivery in the catering industry was not high. The Haidilao project boosted the entire industry. And Keenon Robotics seized this opportunity and became a paid supplier for Haidilao's smart restaurant. After Haidilao, more and more restaurants began to accept the method of using robots for food delivery," Wang Yang commented. Among them, Keenon Robotics was one of the biggest beneficiaries.
Photography: Kong Yuexin
In 2019, the shipment volume of Keenon Robotics' robots exceeded 1,000 units. At the end of the same year, the company completed a 200 - million - yuan Series B financing led by Source Code Capital.
Since 2020, affected by the pandemic, contactless delivery has become a necessity, and the commercial service - robot market has gradually opened up, making the industry a new investment hotspot. During this period, companies such as Keenon Robotics, Pudu Technology, and Youdi Technology all received large - scale financing. Keenon Robotics received more than 1.5 billion yuan in financing from institutions such as SoftBank Asia and SoftBank Vision between 2020 and 2021.
However, the good times didn't last long. In the second half of 2021, the growth of the service - robot market began to slow down, and in 2022, the growth was even more sluggish. To compete for the market, the service - robot industry carried out promotional activities, which worsened the market environment.
Previously, the industry was overly optimistic about the market. The personnel - scale expansion speed of many robotics companies, including Keenon Robotics, far exceeded the market growth rate, and the operational risks were magnified. Wang Yang pointed out that although Keenon Robotics' situation was not as severe as in 2019, improper handling of the personnel - structure adjustment could lead to a serious organizational crisis. He believed that Li Tong's handling at that time enabled the company to achieve a smooth transition.
Wang Yinzhen, the vice - president of SoftBank Asia, also agreed with Li Tong's strategic choice at that time - Keenon Robotics promptly reduced its personnel scale and focused more on its product lines.
"The complexity of team and financial management in a large company far exceeded our expectations," Li Tong said. "We are now using a lightweight management model to ensure controllable operations." Currently, the company's personnel scale is stable at around 700 people.
Compulsory courses for achieving commercialization
After the explosion of large models in 2022 - 2023, Keenon Robotics restarted the R & D of humanoid robots. The rapid commercialization of OpenAI's GPT model made the Keenon team see an opportunity. "At that time, we thought we could re - consider what can be achieved by combining large models with robots," Li Tong said. He believes that the most fundamental problem in the entire humanoid - robot track at present is the inability to achieve commercialization.
Therefore, from the very first day of developing humanoid robots, Li Tong has focused on one thing: it must be able to work, especially to truly solve problems in service scenarios. Only in this way will customers be willing to pay. "We don't really care how fast it can run or how high it can jump at present," Li Tong emphasized. "The key is whether it is designed for work and can do the work well."
In March 2025, Keenon Robotics officially launched its first humanoid embodied service robot, XMAN - R1, which uses a wheeled solution. In July, Keenon Robotics' bipedal service robot, XMAN - F1, was officially launched and exhibited at WAIC and served as a waiter, performing tasks such as serving popcorn and making iced drinks. Li Tong revealed that the company's humanoid robots are currently undergoing POC (proof of concept) in some customer scenarios and are being adjusted and modified according to customer needs.
"We are not a research institution. The essence of a company is commercialization. If we invest a large amount of money without long - term industrial returns, it's definitely not feasible," Li Tong said.
To accelerate the achievement of commercialization, reducing product production costs is one of the compulsory courses for Keenon Robotics. According to Wang Yinzhen's disclosure to "China Entrepreneur," in 2020, during the pre - investment research, SoftBank Asia found that the manufacturing cost of Keenon Robotics' robots was only one - fifth of that of its American peer, Bear Robotics (also invested in by SoftBank), and its selling price was more competitive.
The lower selling price increased market acceptance and enabled it to start occupying the leading position in the catering service - robot field. Subsequently, Keenon Robotics began to promote the implementation