Dialogue with Ren Geng, Founder of Kunlunxing: For automakers to engage in the embodied intelligence sector, entering the market in the future is a "strategic necessity", while cutting in at the current stage is a "tactical opportunism".
For a long time, there has been a fragmentation phenomenon in the embodied intelligence track: on the one hand, abundant hot money is pouring into the industry, attracting autonomous driving technical talents and academic scholars from universities to join the entrepreneurial camp one after another; on the other hand, the commercialization of embodied intelligence remains unclear, and no enterprise in the industry has yet achieved breakthroughs in large-scale mass production. Therefore, on the other side of the capital carnival, industry insiders such as Zhu Xiaohu, Partner of GSR Ventures, are holding a "full position liquidation" bearish view on the embodied intelligence track.
Kunlunxing Robotics precisely caters to the industry's urgent expectation for mass production and landing capabilities. As a result, before the company even completes its registration process, its valuation has already ranked among unicorns, refreshing the speed of unicorn emergence in the embodied intelligence track.
According to previous reports from 36Kr, Kunlunxing Robotics was co-founded by Ren Geng, former Country CEO of Huawei and former Vice President of Alibaba Group, and Lang Xianpeng, former head of the assisted driving division at Li Auto. Less than 90 days after the enterprise's registration, it has successively completed 3 rounds of financing, with a total financing scale of tens of billions of yuan. Top venture capital firms including GL Ventures and Gaorong Ventures have continuously increased their follow-on investment in Kunlunxing for 3 consecutive rounds.
From the very beginning of its establishment, Kunlunxing has set large-scale mass production rather than running demos as its core goal. Ren Geng, the founder of the enterprise, recently accepted an exclusive interview with 36Kr.
Ren Geng mentioned that embodied intelligence is divided into three categories: full humanoid embodied intelligence, humanoid-like embodied intelligence, and non-humanoid embodied intelligence. Kunlunxing will make layout for all three forms of products. "Because many of their AI and hardware capabilities can be reused", but Ren Geng said that Kunlunxing will "develop full humanoid robots first, and then extend the technology to other forms".
In terms of product and technical capabilities, Kunlunxing's humanoid robot chooses to benchmark against Tesla Optimus, taking the full-stack integration route of software and hardware. There are not many companies in the industry that choose this strategy.
Ren Geng believes that "the upper limit of embodied intelligence is determined by AI, and the lower limit is determined by the ontology, or the hardware". "Without physical AI, the embodied intelligence industry would not exist". "At the same time, without the carrier of hardware ontology, no matter how good the AI is, it cannot be presented".
In terms of landing scenario selection, Ren Geng also put forward five core principles for Kunlunxing: full closed loop, value presentation, relative simplicity, data flywheel and human-robot interaction. The ability to achieve end-to-end closed loop and present commercial value is the core requirement.
"If the only purpose is to do POC, shoot videos and make performance rankings, there is no value at all", Ren Geng said sharply.
He believes that the reason why Kunlunxing has planned a clear commercial landing stage comes from the successful past practices and industry accumulation of the founding team members. Based on his practical experience of fighting tough battles and winning victories at Huawei and Alibaba Cloud, Ren Geng is responsible for grasping the company's strategic direction, building the organizational system, commercial operation and enterprise management.
Lang Xianpeng, the other co-founder, has experience in delivering 1.5 million units of high-level assisted driving products, and is proficient in AI R&D experience of achieving technological counterattack with limited resources.
Not only for entrepreneurial partners, Kunlunxing also emphasizes cognitive consistency when screening the swarms of incoming investors. He said "this is to ensure that we reach a consensus in the shortest time and achieve long-term win-win cooperation".
Facing many car companies and autonomous driving companies that cross-border into the embodied intelligence track, Ren Geng holds a pessimistic view. He told 36Kr: "Embodied intelligence is still in a very early stage. At this time, the so-called entry of car companies or autonomous driving companies has no obvious advantages, while their shortcomings are very prominent".
He commented that car companies will definitely enter the embodied intelligence industry in the future, but not at the current stage. It will happen when the market truly matures and enters the explosive period, and there is another condition: the pattern battle among car companies themselves has been completed.
For this reason, for embodied intelligence startups today, "the biggest opponent is themselves". "Their judgment on strategic direction, choice of technical route, construction of organizational system, commercial hematopoietic capacity, talent selection and appointment" determine whether an embodied intelligence startup can survive in the end.
The following is the full interview of 36Kr Auto with Ren Geng, founder of Kunlunxing, with slight edits:
On car companies entering the track: Car companies or autonomous driving companies doing embodied intelligence now have no prominent advantages, but obvious shortcomings
36Kr Auto: How do you view the embodied intelligence businesses of some car companies or autonomous driving companies today?
Ren Geng: It is an inevitable strategic choice for car companies to cut into the embodied intelligence track in the future; but cutting in at the current stage is a tactical speculative idea. From the tactical perspective, embodied intelligence is still in a very early stage. At this time, the entry of the so-called car companies or autonomous driving companies brings more disadvantages than advantages.
The strategic inevitability means that car companies will definitely cut into embodied intelligence, but two conditions need to be met: First, the embodied intelligence market has fully matured and completely exploded, which will bring at least visible and definite hundreds of billions of yuan in revenue to car companies every year; Second, the final battle among car companies has ended, and they have truly secured a seat at the winner's table of the automotive industry. The life-and-death battle of car companies today is far from over. In industries such as automobiles, mobile phones and chips, only a limited single-digit number of players can truly secure a seat at the table in the final stage.
The so-called tactical speculation lies in that for medium-sized or large factories, their biggest advantage over startups is resource advantage. But if a medium-sized or large factory can only assign a small department to invest and compete in a relatively cutting-edge industry, this small department under the guise of a large factory may not even have richer resources than startups.
Second, in terms of ownership, the leader of this department is not an entrepreneur after all. Although he may be very excellent, he essentially works for his superiors, while entrepreneurs work for themselves. Their different ownership determines that the energy they release is completely different.
Third, current car companies doing embodied intelligence will face various rigid restrictions from existing processes within the group, such as computing power management, human resources policy management, legal affairs management, financial management, all of which are huge constraints.
In addition, people's motivations for doing embodied intelligence are also different: some are for retaining talents, some are for market value management, and some are for going public, etc. Therefore, on the whole, when car companies invest in embodied intelligence at this stage, their strengths are not obvious, while their shortcomings are very prominent.
36Kr Auto: Then some CEOs of car companies also serve as the CEO of their robotics business. Can't this solve the problems of internal resource division and strategic ownership?
Ren Geng: I think the key does not lie in the title, but in practical actions, to vote with your feet. For example, Lei Jun, Mr. Lei, made cars, and his resource investment in the automotive business is exactly voting with his feet.
Voting with your feet means whether the CEO's time and the enterprise's resources are all put into the business. This is the real investment, not just whether you hold a part-time position on the surface. If today's automotive companies announce their entry into embodied intelligence, we need to see how much resources they have invested. If it is only a resource input of 100 or 200 people, there is no essential change. If you invest 5,000 or 10,000 people, that is real voting with your feet, which can reflect your all-in attitude.
36Kr Auto: If car companies spin off their robotics business for financing, how will the capital market view this type of business today?
Ren Geng: I think spinning off the robotics business for financing may solve these constraint problems to a certain extent. But for a large enterprise, if this is a very good business, a golden asset, it will be reluctant to spin it out for financing. If it chooses to spin it out for financing, it proves that it may not be so optimistic about this business, or thinks that the business requires a relatively long cycle of cultivation and development.
In that case, why spend time and energy investing in it? We often say a word in the past: do not consume the strategic resources of large factories on non-strategic opportunity points. Because the advantage of large factories lies in their strategic resources, which means that they can carry out saturated attacks and make high-intensity concentrated investment.
Therefore, spin-off seems to solve the problem to a certain extent. But fundamentally, we need to go back to the core question: whether you should do this business at all.
36Kr Auto: Car companies have reusable advantages in hardware, electric drive and intelligent driving algorithms. Once they enter the track, how long do you think the window period for startups like Kunlunxing will be, and how long will it take to catch up with these advantages?
Ren Geng: As analyzed earlier, the real all-in entry of large factories is far from coming. Because there is not enough market space to support them to invest strategic resources. Xiaomi is a large factory, but its team doing embodied intelligence is a very small department, which should be less than 200 people. I think only when the team size reaches at least 3000 people can it be regarded as the real entry of a large factory.
Because the combat effectiveness of a 200-person team may be far less than that of an external startup, there are many internal frictions, and it also lacks ownership. If the owner of a large factory personally takes charge of the business, he must see huge returns, otherwise it is impossible for him to do so. All these determine that this is not the time for large Chinese enterprises to invest all-in in embodied intelligence.
But they will definitely enter the market after the industry matures, and I think it will most likely be 10 years later.
On founding team: Find suitable entrepreneurial partners, not just familiar ones
36Kr Auto: Lang Xianpeng, like you, was once the top leader of the core business of a leading enterprise. How did you two come together to start a business?
Ren Geng: I think for an entrepreneur, he needs to avoid falling into a trap: all the partners he finds are people he was familiar with in the past. This is a dependency trap and a trap of lack of confidence.
Entrepreneurs should find the most suitable people, because suitability is more important than familiarity.
So when I look for partners to start a business together, I have several principles: first, whether they are competent; second, whether we can get along well; third, whether we can stick together for a long time.
Dr. Lang was the top leader of the assisted driving division at Li Auto, and the assisted driving performance of Li Auto is in the first echelon of the industry. His past success has proved his competence.
Then for the founding team, harmony is the most precious, and the first principle of harmony is that everyone has the same cognition. I am very lucky that Dr. Lang has experienced the complete process from technology to product and then to commercialization at Li Auto. So we have a high degree of consistent cognition on the choice of technical routes, judgment of technology trends, product selection and many other things.
Finally, I also need to consider whether we can stick together for a long time. The dream of the enterprise is to become a large-scale enterprise in 10 or 15 years when physical AI arrives, which means that the founders need to grow all the way from 0 to 10, and then to 100. Dr. Lang has proved at Li Auto that he can grow to a very high stage, so the answer to the question of whether we can stick together for a long time is also very clear.
36Kr Auto: How do you divide the work at present?
Ren Geng: Dr. Lang will give full play to his unique advantages, and he is the leader of our AI division. For other aspects, such as the company's strategic direction, construction of organizational system, commercial operation, enterprise financing and so on, I will be in overall charge.
36Kr Auto: Kunlunxing's financing progress is amazing. It has raised tens of billions of yuan in three months. How did you achieve this? I heard that when you meet institutional investors, you almost only meet their top leaders. Is this one of the secrets?
Ren Geng: This is just a tactical choice. Because time efficiency is very important for entrepreneurs, especially the top leader, so communicating directly with investors who have decision-making power, no matter he is the first, second or third leader of the institution, is to save the time of both sides and ensure that we reach a consensus as soon as possible.
From the perspective of investors, for them, there are no more than three core elements for investment:
First, the track they choose;
Second, the team they choose;
Third, the strategic direction of the founding team they choose.
I think these three points are the most core. Our team does have some uniqueness, and we have comprehensive, end-to-end successful practical experience in the past.
In addition, the strategic direction we formulated is targeted at mass production, with continuous AI innovation as the core. These are our core strategies. I think the founding team and our strategic direction are quite attractive to investors.
36Kr Auto: At present, embodied intelligence enterprises are divided into several factions: some are academic factions, some are technology factions from large factories, and some investors call founders like you and those from Agibot the helm faction. How do you view this kind of faction division in the industry?
Ren Geng: I think there are indeed many excellent colleagues in the embodied intelligence industry. As you said, some are professors or scholars with deep accumulation in the academic field; some are people with AI practice experience in the autonomous driving field; others are like us, whose founders have accumulated a lot of experience in past enterprises.
Our advantages are mainly reflected in two aspects: one is the experience above the iceberg, and the other is the capability under the iceberg.
The experience above the iceberg refers to the comprehensive and rich experience we have accumulated in the past. This experience may come from whether you have successful landing innovation practices in the AI field, whether you have built a R&D system for world-class hardware products with millions of units in shipment, whether you have large-scale commercial success experience, whether you have successful practical experience in supply chain management, and whether you have successful experience in motivating the team and uniting people's hearts. I often say a word in the past: we are strong because we have won, not we win because we are strong. You need to go through many successful things to become stronger. The process of experiencing victory will be accompanied by many failures, but only when you finally go from failure to success, you will become strong. Therefore, for operating an industry with high threshold, great difficulty and long cycle like embodied intelligence, past successful practices are very important.
Under the iceberg, I think our biggest feature lies in cognition. The founder's cognition comes from three dimensions: first, your past experience, including the scale of the AI+ hardware business you have managed in the past, the scale and performance of the enterprise you have operated, and the success you have experienced from 0-1 or 1-100. All these are the cognitive foundation for you to run a high-threshold enterprise. As we often say in the past, if you have never been to the world, you will not have a real world view; if you do not have relevant experience, it is difficult for you to have many essential cognitions.
Another source of cognition comes from the more senior people you have communicated with in the past. Thanks to our past work experience, we can have in-depth exchanges with more excellent people, which also improves our cognition.
The third source is systematic reading, but the most important source is past practices and the work you have engaged in before. So I think cognition is very important for founders.
36Kr Auto: Kunlunxing has chosen a path that takes into account both software and hardware and benchmarks against Tesla Optimus. Why did you choose this direction? Now many leading domestic projects focus on ontology + control algorithms, and AI does not occupy the core position in the enterprise strategy.
Ren Geng: This strategic path is considered from several aspects. First, from the essence of embodied intelligence, I think the upper limit of embodied intelligence is determined by AI, and the lower limit is determined by the ontology, or hardware.
AI is the ceiling of embodied intelligence. Without physical AI, the embodied intelligence industry would not exist. Therefore, the development speed of physical AI determines how fast and how far embodied intelligence can develop. When physical AGI really comes, embodied intelligence will be a market of hundreds of trillions of yuan, which is a revolutionary event that can completely change the course of human history.
At the same time, without the carrier of hardware ontology, no matter how good the AI is, it cannot be presented.
Second, from the perspective of business model, in the AI 1.0 era, the best business model is not to simply develop