Li Auto Plans L4 Autonomous Vehicle by 2028, Calling It the Industry’s "iPhone 4 Moment"
Li Auto a leading Chinese new energy vehicle manufacturer, revealed in a recent interview with National Business Daily that it plans to launch its first vehicle defined by Level 4 (L4) autonomous driving capabilities within the next three years, targeting a release around 2028. The company’s Chairman and CEO, Li Xiang, further speculated that by 2030, there is a 50% probability the company will introduce an artificial intelligence-driven supercar.
This strategic roadmap comes as the global capital market experiences a renewed wave of technology investment centered on AI, and follows the Chinese government's concerted push to deepen the implementation of its "AI+" action plan. Li Auto views the realization of L4 autonomous driving as the automotive industry's genuine "iPhone 4 moment," predicting it will trigger a profound industrial revolution that fundamentally reshapes product forms, user experiences, and business models.
The Automobile Will Transform from a Transportation Tool into an AI-Era "Space Robot"
NBD: China is currently vigorously developing "AI+." In the next 5 to 10 years, what will be the biggest change for automobiles and the automotive industry?
Li Auto: We believe that in the next 5 to 10 years, "AI+" will fundamentally reshape the automotive industry. The automobile will transform from an industrial-era transportation tool into an AI-era "Space Robot." This is not just a technological upgrade, but a revolution in product form and industrial paradigm.
For the industry, electrification is the process, while AI is the endgame. The core of competition will shift from traditional hardware manufacturing to comprehensive capabilities driven by software and artificial intelligence technology. Enterprises need to possess a comprehensive system capability covering models, computing power, operating systems, and hardware bodies. These four elements complement each other and are indispensable.
For car owners, the biggest change in experience will come from AI leaping from an "auxiliary tool" to a "production tool."
Specifically, first is the complete liberation of driving. Through the VLA (Vision-Language-Action) driver foundation model, AI will work like a true "driver." Users will be completely liberated from driving tasks and can freely work, rest, or entertain themselves during the journey. The vehicle will become a safe, mobile free space.
Second is natural human-machine interaction. Users can communicate with the vehicle using natural language, as simply and directly as communicating with a human driver. For example, complex commands such as "take the manual toll lane" or "park in area C3 of the parking lot" will be understood and executed by the vehicle.
Finally, there is safety and comfort that surpasses human capabilities. Through technologies such as "Super Alignment" and "World Models," the AI driver will not only obey traffic rules but also align with human driving habits and values. Its goal is to achieve a travel experience that is safer and more comfortable than human driving.
The Realization of L4 Autonomous Driving Will Be the Automotive Industry’s Genuine "iPhone 4 Moment"
NBD: Many believe that L3 is a technological "pseudo-proposition" or a transitional stage, and that automakers should run fully toward L4. What makes you think of this?
Li Auto: We believe that L3 supervised intelligent driving is not a pseudo-proposition, but a necessary path and key precursor to L4. It is an indispensable transitional stage that allows AI systems to learn, verify, and iterate in the real world, accumulating the necessary experience and data for the ultimate realization of fully unmanned driving.
To achieve L4 autonomous driving, we believe critical breakthroughs are needed in the following areas:
regarding core technology breakthroughs, we rely on the maturity of the VLA driver foundation model as the core algorithm. It needs to possess complete visual perception, language understanding, and chain-of-thought reasoning capabilities, enabling it to handle complex traffic environments like a human.
Huge investment in computing power resources is the material basis for achieving AI breakthroughs. Li Auto invests over RMB 6 billion yuan (approx. US$827 million) annually in artificial intelligence models, computing power, and infrastructure, ranking among the highest levels of investment in the automotive industry. Reinforcement training requires a large number of inference cards, and the parameter scale of vehicle-end models also needs continuous upgrading, which places extremely high demands on computing power.
Regarding the scale effect of the data closed loop, a sufficiently large fleet is required to drive on the road, continuously collecting massive amounts of driving data, including extreme cases, for continuous model training and optimization.
NBD: Some market views suggest that cars without a driver's seat will definitely appear within the next 3 to 5 years. How do you view this perspective?
Li Auto: We will launch an L4-defined vehicle within three years. Regarding the market speculation that "cars without a driver's seat will definitely appear within the next 3 to 5 years," we give a clear and positive response. According to the company's plan, Li Auto will launch its first vehicle defined for L4 autonomous driving within three years, around 2028.
The launch of this model will mean a fundamental change in automotive design concepts. Since the vehicle will have fully autonomous driving capabilities, traditional cockpits, steering wheels, pedals, and other components will no longer be necessities, thereby offering infinite possibilities for interior space design.
At that time, the car will truly transform from a "driving machine" into a pure "living space" or "work space."
Li Auto Chairman and CEO Li Xiang has even further envisioned that by 2030, there is a 50% probability the company will launch an AI supercar. This indicates that Li Auto not only has a clear plan for the future trend of "cars without driver's seats" but is also full of confidence and expectation. The company believes that the realization of L4 autonomous driving will be the automotive industry's true "iPhone 4 moment," triggering a profound industrial revolution.
NBD: To realize "cars without a driver's seat," what other difficulties need to be overcome?
Li Auto: Although the prospect of "cars without a driver's seat" is broad, its implementation still faces huge technological challenges and market tests.
At the technical level, the realization of L4 autonomous driving is itself an extremely complex system engineering task, requiring extremely high reliability and safety in all links such as perception, decision-making, and control. In addition, vehicle redundancy design, cybersecurity, and collaborative interaction with the external environment are all difficulties that need to be overcome.
At the market level, consumer acceptance of fully purely unmanned driving, the improvement of relevant laws and regulations, and the definition of insurance liability are all important factors restricting its large-scale commercial implementation.
However, challenges and opportunities coexist. Once these technical and market obstacles are overcome, "cars without a driver's seat" will open up a trillion-dollar new market.
It will completely change people's way of travel, work, and life, and spawn countless new business models and services. To welcome the arrival of this day, we are not only sprinting fully on technology but also actively participating in the discussion and formulation of relevant policies and regulations, hoping to smoothly promote its commercial implementation when the technology matures.
A "Generational Gap" in Computing Power Exists Between Domestic VLA Tech and Top Overseas Tech
NBD: Current AI applications on the production side of the automotive industry are mostly point-to-point, with highly specific models that are difficult to apply if the scenario changes. How should model generalization be promoted so that it can be applied to multiple scenarios and links, avoiding repetitive investment by enterprises?
Li Auto: The key to solving the problem of poor generalization ability of "point-to-point" models lies in building general AI foundation models and capability platforms.
Our practical path is to deeply cultivate the VLA foundation model. We are currently dedicated to developing the VLA model, hoping to build a "World Model" that can deeply understand the physical world. Once this model matures, its core capabilities can be reused in multiple links such as manufacturing and supply chain management. For example, it can automatically detect production defects through visual analysis or optimize logistics scheduling through language understanding.
In addition, we adhere to the synergy of in-house research and open source. We insist on self-research in key areas to build underlying capabilities. At the same time, we actively embrace open source and give back to the community, such as open-sourcing our self-developed "Li Auto Xinghuan OS" (Li Auto Ring OS), with the aim of avoiding the industry "reinventing the wheel" so that everyone can stand on each other's shoulders to innovate together.
NBD: The "VLA Driver Foundation Model" launched by Li Auto can understand human language. what gaps still exist between domestic VLA technology and overseas artificial intelligence technology?
Li Auto: As a critical technical path to high-level intelligent driving and even General Artificial Intelligence (AGI), the VLA model is becoming the focus of competition among global tech giants and automakers. Through our self-developed Mind VLA model, we have already applied it to mass-produced models.
However, although domestic enterprises show strong competitiveness in the application and iteration speed of VLA technology, compared with top overseas AI technology represented by the United States, there are still significant gaps and challenges in multiple dimensions such as computing power infrastructure and data closed-loop construction. These gaps not only determine the ceiling of technological development but also directly affect the safety, generalization ability, and user experience of future autonomous driving products.
First is the gap in computing power infrastructure. Computing power is the "fuel" for training large models, and its scale and efficiency directly determine the rate of model iteration and the upper limit of capability. On this core element, there is a clear "generational gap" between domestic automakers and top overseas players. The overall scale and investment of Chinese automakers in computing power infrastructure still lag significantly.
Thinking also extends to the data closed loop and model iteration. Data is the "nutrient" of AI models. A high-quality, large-scale data closed loop is the key to driving continuous model iteration and improving generalization capability.
Autonomous driving systems, especially VLA models, rely heavily on massive, diverse driving data collected from the real world. This includes regular scenarios and a large amount of long-tail scenario data. Building an efficient data closed loop, that is, the cycle of "collection—labeling—training—deployment—feedback," is the key for top automakers to build core barriers. It is estimated that 20 million clips (video fragments) of data will be needed in 2025 to achieve L3 autonomous driving, while L4 level will require at least 5 million vehicles capable of collecting data.
Although China has advantages in data scale and market application, there is still a gap with top overseas enterprises in the efficiency of data processing and the completeness of closed-loop construction.
Traditional Human Resource Management Cannot Adapt to the AI Era
NBD: The recent integration of the human resources department at Li Auto has also attracted outside attention. Is this related to adapting to the current AI talent competition?
Li Auto: To better adapt to the requirements for talent and organization in the "AI+" era, Li Auto carried out a profound organizational structure adjustment in 2025. The most notable change is that CEO Li Xiang began to directly manage the human resources department.
Behind this adjustment is our profound thinking on organizational development in the AI era. Li Xiang believes that the traditional human resource management mode, which emphasizes process and rhythm, can no longer adapt to the requirements of the AI era for organizational agility and innovation. By merging the human resources department into the Product and Strategy Group and reporting directly to the CEO, Li Auto hopes to combine talent strategy more closely with the overall development strategy of the enterprise, ensuring that human resources can be configured accurately and quickly in key areas.
In addition to the adjustment of the human resources department, we are conducting a series of broader organizational changes to build an agile and efficient organization capable of adapting to the "AI+" era.
We are transforming from a traditional functional department system to a more flexible matrix organizational structure, breaking down departmental barriers and promoting cross-departmental collaboration and innovation. At the same time, the company is constantly optimizing its internal processes and mechanisms, for example, by simplifying approval processes and delegating decision-making power to improve the organization's reaction speed and execution efficiency.
Furthermore, Li Auto is actively creating a corporate culture that encourages innovation and tolerates failure, encouraging employees to be bold in their attempts and brave in exploration, thereby stimulating the innovation vitality of the entire organization.