Yin Qi, a new player in intelligent driving, and his goal of achieving a production value of tens of billions with "AI + automotive"
Text by | Zhou Xinyu
Edited by | Su Jianxun
The intelligent driving company "Qianli Technology", which has kept a low profile for a year, officially launched its brand to the public.
On September 28, 2025, the two most important figures behind the establishment of Qianli Technology - Yin Qi, the chairman of Qianli Technology, and Li Shufu, the founder of Geely Automobile, announced the new English brand name "AFARI" and the brand logo of Qianli Technology.
Middle: Li Shufu, the founder of Geely Automobile. Right: Yin Qi, the chairman of Qianli Technology. Photo source: Taken by the author.
At this debut event, Yin Qi summarized Qianli's business progress and plans into two keywords: Commercialization, Internationalization.
In August 2025, the semi - annual report of Qianli Technology for 2025 showed that the company achieved an operating income of 4.184 billion yuan in the first half of the year, a year - on - year increase of 40.04%.
The strong growth actually mainly comes from the automobile and motorcycle businesses of its predecessor "Lifan Technology", and the leverage of the "AI + vehicle" has not been reflected yet.
Regarding the commercialization of "AI + vehicle", Yin Qi's expectation is: To achieve an output value of over 10 billion yuan this year and form large - scale revenue next year.
Currently, Qianli Technology has three major businesses: Qianli Intelligent Driving (L2+ to L4 - level intelligent driving solutions), Qianli Intelligent Cockpit (natural interaction cockpit experience), and Qianli Intelligent Mobility (Robotaxi).
Yin Qi told the media that as Qianli's first AI solution, the intelligent assisted driving solution will be able to generate large - scale commercial revenue next year; Qianli's cockpit products and Robotaxi business will also make clear progress next year.
The second keyword is internationalization.
As early as May 2025, Yin Qi mentioned in an interview that there are two differences between Qianli Technology and Huawei. One is openness, and the other is internationalization: "Combined with Geely's international experience, overseas automakers will be our focus."
At the brand launch event, Wang Jun, the chairman and CEO of Qianli Intelligent Driving and the co - president of Qianli Technology, regarded internationalization as a necessary move to seize market share: "To achieve one - third of the global market share, we cannot only compete in the domestic market. We must go global."
Yin Qi measures intelligent driving solutions by the "model content", that is, the proportion of large models in the "model, rules, map" system.
He told the media that currently, the model content of leading overseas intelligent driving solutions can reach 80% - 90%, while most domestic solutions only have 50%. Qianli's goal is to raise the model content to the leading overseas level in the next six months.
Now, in its international layout, in addition to relying on Geely's overseas resources, Qianli has also introduced capital from global automakers. On September 25, 2025, Qianli Technology announced that Mercedes - Benz acquired 3% of Qianli's shares at a price of approximately 1.339 billion yuan, becoming its fifth - largest shareholder.
Mercedes - Benz's investment means that Qianli's global layout has a solid industrial system as a backing. At the brand meeting, Yin Qi responded to Mercedes - Benz's investment: "In the future, we will conduct close strategic cooperation with Mercedes - Benz in terms of quality and quality control."
Since the establishment of Megvii Technology, Yin Qi has had more than 10 years of experience in the AI field. But for the intelligent driving industry, he claims to be a "newcomer".
However, he believes that in the intelligent driving field, those who take the lead often have the burden of technical routes, and Qianli has a more open attitude towards technical solutions: as long as it can reduce costs and be scaled up, it can be adopted. "For the first time, we feel that we have the late - mover advantage."
In a group interview with the media, four senior executives of Qianli Technology responded to questions about the technical stage of intelligent driving, overseas business layout, the industry pattern of intelligent driving, Qianli's commercialization and global business layout. They are:
- Yin Qi, the chairman of Qianli Technology;
- Wang Jun, the chairman and CEO of Qianli Intelligent Driving and the co - president of Qianli Technology;
- Chen Qi, the co - CEO of Qianli Intelligent Driving and the chief intelligent driving scientist of Geely Holding Group;
- Yang Mu, the CTO of Qianli Intelligent Driving.
"Intelligent Emergence" has sorted out some of the views:
- In the history of intelligent driving, there are two routes: one is the pure vision model solution represented by Tesla, and the other is the lidar system centered on "high - definition maps + rules" represented by early Waymo.
These two routes have converged significantly today. The ultimate intelligent driving system in the future must be centered on visual perception, with a larger large model as the underlying logic, and special sensors are added in different scenarios to ensure the safety boundary. But the underlying must be a large model with extremely high "model content".
- All intelligent driving systems on the market can be summarized by the "hamburger theory": the core is a model, the upper layer is an expert system based on rules, and the upper layer is the map system. The "model content" refers to the proportion of the model, rules, and map in the "hamburger".
The total model content of mainstream intelligent driving solutions on the market may be less than 50%. Abroad, the benchmark model content in the industry can reach 80% - 90%. I am confident that Qianli's model content will reach 80% - 90% in the next six months.
- The team cooperation model in the entire intelligent driving field has gone through many different trends.
In the first stage, only teams with pure AI experience could do it; in the second stage, OEM manufacturers were very confident and insisted on doing it themselves for better control. Now, it has reached a rational and open state.
In the industry, several different solution systems will emerge in the future, and these systems will form relatively in - depth relationships with some OEM manufacturers. Because the soul of all AI systems is data, only by forming in - depth connections, rather than just a client - supplier relationship
- China's intelligent driving has developed for more than 10 years. Ultimately, it may gradually present a relatively leading pattern. We don't need so many intelligent driving solutions because many intelligent driving solutions often reinvent the wheel.
Under a large OEM system, there are basically no more than three suppliers. If suppliers within the same system are excellent, they will develop more personalized product ideas, including cockpits and other products.
- In the field of intelligent assistance and AI for so many years, we have felt the late - mover advantage for the first time. Because those who take the lead have a burden. If they say they do pure vision, they are embarrassed to add a lidar.
I (Yin Qi) have a technical background and am quite decisive in technical judgments. I believe that the solution of "driven by large models and centered on pure vision" must be the ultimate solution for all intelligent assisted driving in the future. This is a certainty.
In the use of sensors, the future must be composite sensors, not only lidar, including 4D millimeter - wave radar, and even more innovative sensors in the future. As long as they can help reduce costs in some extreme cases and can be scaled up, they can be used in future intelligent assisted driving.
There are basic principles, and there are superimposed solutions in the future. This is very clear and a definite judgment that can be verified in the future.
- L2+ to L4 must be within the same technical framework. This is a technical issue because data needs to be reused. Without L2+, it is actually difficult to truly replace on a large scale in L3 and L4.
There must be different focuses from L2+ to L4. For example, in L4, a more efficient map system will also be very important in the future.
- Intelligent driving solutions must provide a complete software + hardware solution, which can help improve future value and experience.
- In terms of chips, we will definitely choose a relatively open chip ecosystem. We will choose chips that are truly competitive and cost - effective.
There are still two major problems now: one is computing power, and the other is data. Although the computing power of in - vehicle terminals can reach 800 Tops or 1000 Tops, it still has great limitations compared with the computing power on the cloud side.
In the future, we will definitely see an increase in the computing power in the field of in - vehicle intelligent assisted driving and more applications of multi - modal large models on the terminal side. This trend is also very certain.
- A core issue in going global is data.
- There is a bit too much attention on chips. AI chips are simpler in design than traditional CPU and GPU chips, which is why there are so many chip systems.
Chips must ultimately seek a sustainable business model and be able to be scaled up. If a manufacturer has 1 million vehicles, this data is very small in the chip field. In the chip field, if there are not 100 million chips produced in a year, it is impossible to truly iterate continuously.
- In the future, the base model will be implemented as industry - specific models in different scenarios. In fact, there won't be so many industry - specific models. We must realize this.
Industry - specific models are essentially the opposite of this wave of large models. The characteristic of this wave of large models is universality. In the future, a lot of industry data and industry know - how will be added, but it will still be a model mainly based on universality.
So Qianli is very clear that it will not develop base models. Instead, it will cooperate with its best base model partner, Jieyue Xingchen, to achieve strategic synergy in base models.
- Why emphasize that L2+, L3, and L4 must be within the same framework? This is the only way for the future business model to work because the R & D costs and data are shared.
- The Robotaxi is still in the early exploration stage and has not reached the commercialization stage. The product and experience have not been closed - looped. We have made a simple calculation. Only when 1000 vehicles are in regular operation can it enter the real large - scale commercialization stage. No company in the world has achieved this yet.
The Internet has developed for 30 years. Wherever there are people and traffic on the Internet, there are countless value - added services and monetization possibilities. There is no need to worry about the commercialization of Robotaxi in the future. What people care about is how to make a safer and better Robotaxi as soon as possible.
- An important issue with Robotaxi is related to employment. How can Robotaxi create more jobs? It's not about AI replacing humans, but about AI creating better services and more jobs. Whether it's taxi or ride - hailing drivers, they can work more easily and earn more in this wave. This is very crucial.
- Currently, the team has nearly 2000 people. Of course, a larger team is not necessarily better for an intelligent driving team. But Qianli is working on an overall solution from L2+ to L3 to L4, and 2000 people can do more things.
- Some time ago, there were three - wave personnel adjustments in Qianli Intelligent Driving, including the integration of Maichi, ZEEKR, and the research institute. This integration was a stroke of genius.
Intelligent assisted driving systems require two types of organizational genes: one is a very strong AI model - native gene. Each model's data iteration and model - driven need an excellent AI team. The other is a very powerful engineering system team. These two teams are indispensable.
From a business model perspective, Qianli Intelligent Driving may be replicable, but from the perspectives of the team and the industrial window period, it is not that easy to replicate. These three teams complement each other very well when combined.
Team integration is actually more difficult than building a new team. There are two core points: First, the goals must be consistent. Second, no matter how many teams there are, there is only one core version, the main version. Then we allocate talents around the main version to truly achieve team integration.
Welcome to communicate!