Li Auto conference call: Li Xiang announces return to entrepreneurial model, betting on embodied intelligence, with delivery of self-developed M100 chip AI system in 2026

The latest financial report shows that Li Auto's revenue in the third quarter fell by 36.2% year-on-year. More seriously, the company has turned from profit to loss, with a net loss of 624 million yuan, compared to a net profit of 2.8 billion yuan in the same period last year. This cliff-like decline not only reflects a significant drop in delivery volume but also exposes the multiple pressures the company faces during its transformation period.

In the conference call, CEO Li Xiang announced that Li Auto will firmly return to the "startup management model" for its second decade, abandoning the professional manager system; the product will be redefined as "embodied intelligent" robots, focusing on AI's understanding of the physical world and proactive service capabilities. Li Xiang clearly stated that the self-developed M100 chip AI system will be delivered in 2026, with the core transformation being to shift the car from "passive function stacking" to "proactive automated services."

Li Auto's CTO Xie Yan stated that the company's self-designed M100 AI inference chip is undergoing large-scale system testing and is expected to be commercialized next year, with performance cost ratios reaching more than three times that of today's high-end chips after collaborative design with the foundational model compiler and software system. Li Auto is developing foundational models for physical AI vehicles, focusing on enhancing perception, understanding, and response capabilities, with the development of the next-generation platform and chips already underway.

Regarding the fully self-developed three electric systems, President Ma Donghui stated that in terms of electric drive systems, Li Auto has self-developed silicon carbide power chips, self-manufactured power modules and motor controllers, and established a dedicated drive motor factory, forming a complete self-research chain. In battery technology, the company has built full-stack capabilities around 5C ultra-fast charging, from cell chemistry systems, BMS systems to Pack structural design, and will mass-produce its own brand 5C batteries next year. The supply chain adopts a "purchased + self-developed joint venture" dual model to ensure stability.

For the current pure electric models L6 and L8, President Ma Donghui revealed that by 2026, the Li Auto L series will undergo significant redesign, with the entire series equipped with an 800V high-voltage platform and 5C ultra-fast charging technology, streamlining SKUs and enhancing the entry-level experience, aiming to regain the leading position in range-extended products and significantly increase pure electric penetration.

Transcript of Li Auto's conference call (translated with AI assistance)

Li Auto Management: CEO Li Xiang, CFO Li Tie, President Ma Donghui, CTO Xie Yan

Host:

Thank you, operator. Good evening, good morning everyone. Welcome to Li Auto's third quarter 2025 financial results conference call. The company's financial and operational performance was released earlier today in a press release and is available on the company's investor relations website.

In today's conference call, we will have our Chairman and CEO Mr. Li Xiang, as well as our Chief Financial Officer Mr. Li Tie. After the prepared remarks, our President Mr. Ma Donghui and Chief Technology Officer Mr. Xie Yan will join the discussion for the Q&A session.

Before we continue, please note that today's discussion will contain forward-looking statements made under the safe harbor provisions of the Private Securities Litigation Reform Act of 1995. Forward-looking statements involve inherent risks and uncertainties Therefore, the company's actual results may differ significantly from the views expressed today. More information about risks and uncertainties is contained in certain documents submitted by the company to the U.S. Securities and Exchange Commission. Beyond this, Hong Kong Exchanges and Clearing Limited has no obligation to update any forward-looking statements unless required by applicable law.

Please also note that Li Auto's earnings press release and this conference call include discussions of unaudited GAAP (Generally Accepted Accounting Principles) financial information as well as unaudited non-GAAP financial metrics. Please refer to the disclosure documents of Li Auto in the investor relations section of our website, which contain reconciliations of unaudited non-GAAP metrics to comparable GAAP metrics.

Our CEO will begin his remarks in Chinese, followed by an English translation after he has completed all his remarks. With that, I will now turn the call over to our CEO, Mr. Li Xiang. Please go ahead.

CEO Li Xiang:

Hello everyone, I am Li Xiang. This is the third quarter of 2025, and also the first quarter of Li Auto's second decade. We have faced various challenges brought about by product cycles, public relations issues, supply chain ramp-up, policy changes, and other factors that have impacted our deliveries and operations. However, today I would like to discuss our long-term thinking. What are the three most important key choices that Li Auto should make as we enter our second decade? They are organization, product, and technology.

The first key choice is organization. The choice we face is between the management model of a startup or that of professional managers. In the past decade, Li Auto initially operated under a startup management model for the first seven years. As we scaled up to a revenue level that I had not experienced in my previous ventures, around 2022, many people suggested that we transition to a professional management model. Historically, century-old automotive companies like Mercedes-Benz and BMW, as well as tech giants like Microsoft and Apple, have achieved great success under this model. Over the past three years, we have worked hard to transform ourselves into a governance system of professional managers. After real experiences and implementations, we have realized that the governance systems of startups and professional managers are fundamentally different. It is not related to processes or organizational structures, but rather to differences in management philosophy and elements, each applicable to different stages and industry environments. The management approach of professional managers can be very successful, but it requires three elements: first, a relatively stable industry and technology cycle.

Second, the company must have a leading and stable position in the industry. Third, the founders and startup team must either lack the motivation to continue or no longer be present. If these conditions are met, the professional management model is a very good choice. Both Apple and Microsoft have grown from hundred-billion-dollar companies to trillion-dollar enterprises after transitioning to professional managers. Conversely, the conditions for startups are quite the opposite: first, the industry and technology cycles are undergoing significant changes. Second, the industry landscape is uncertain, and the company is not yet a leader. Third, the founders and startup team are still actively working every day, full of motivation. In today's environment, where artificial intelligence is reshaping various industries, our circumstances and characteristics align more closely with those of a startup. The core of startup management involves four key aspects: first, more dialogue, deep dialogue, rather than more reporting In a highly changing environment, deep conversations are key elements to enhancing cognition and making bold decisions. Second, focus on user value, rather than just delivering what is truly valuable to users; it is worth delivering only when it goes beyond merely completing various delivery tasks.

Third, continuously improve efficiency, rather than occupying more resources. If something cost 10 yuan last year, it should only cost 8 yuan this year, thereby freeing up more resources for long-term investments and capability building that do not generate short-term income. Fourth, identify key issues rather than creating information asymmetry. Only when value increases, efficiency improves, and key issues are resolved can we continuously meet user needs in a highly competitive and rapidly changing environment. Over the past three years, my entrepreneurial team and I have worked hard to learn the management systems of professional managers, forcing ourselves to accept various changes, but we have become increasingly worse versions of ourselves. Meanwhile, NVIDIA and Tesla still manage like startups today. If the strongest companies in the world can use startup management methods, what reason do we have to abandon what we are best at? Since 1998, I have 27 years of management experience in startups and have never worked in any large enterprise as a professional manager. Now, facing a highly competitive industry and a rapidly changing technological environment, I love cars, love products, and love artificial intelligence; work is my greatest hobby. Why not use my best skills and methods to manage Li Auto? This is the most important choice for Li Auto as we enter our second decade. Therefore, starting from the fourth quarter of this year, we are firmly returning to the management model of a startup to face the challenges of the new era and new technologies.

The choice of organization is the foundation of everything. As we enter a new decade, the more critical question is what problems we truly want to solve for users. What is the product? What technology is involved? Where do we go? This is the eternal essence. First, regarding the product, we also need to make a choice. What will our product be for the next decade? Is it electric vehicles? Smart terminals? Or embodied robots? When the product remains as an electric vehicle, the logic of competition will turn into a parameter battle: is the range 20 kilometers more or 20 kilometers less? Is the car length 2 centimeters more or 2 centimeters less? If it is merely a competition of electric vehicles, it will be about larger space, longer range, and cheaper prices. By the way, we might just copy those designs that have already been successfully validated, such as the Li L9.

Any R&D investment beyond this is a waste of R&D costs. Stronger sensors, stronger models, stronger computing power, and stronger active suspension are all wastes of overall vehicle costs. In fact, the energy consumption generated by powerful computing power and active suspension can even have a negative effect on range. If we choose to make smart terminals, we will focus more on the things inside the screen, instinctively turning into a redundant construction of smartphone functions. Most of the innovations in the industry regarding smart terminals in recent years have followed this logic, essentially moving smartphone functions into cars, placing mobile apps into car systems, and even hoping to do coding and deep research inside the car

Are users buying cars for work or for a better life? Clearly, there are better experiences and more natural applications on smartphones and computers, so why insist on having them in cars? The value enhancement from these investments for users is very limited, and it may even be just corporate self-entertainment. We can also turn cars into embodied intelligent products in the physical world. Simply put, they are robots. The animated series and movies of Transformers clearly tell us that the biggest classification of robots is into two forms: one that resembles humans and one that resembles cars. Knight Rider and Cars also clearly tell us that cars are one of the core forms of robots. How to transform cars into robots, giving them the sensory capabilities of eyes and ears, the modeling capabilities of a brain and nervous system, and computational abilities akin to a heart and other organs, is about transforming hardware to form stronger intrinsic capabilities.

Equipping automotive robots with the abilities of top drivers means they can not only drive but also greet you every day, help you park, assist with charging, and open and close doors for you, providing meticulous convenience and care. They can also play the roles of parental assistants and attendants, offering you the most convenient life services and spatial care inside the car, just like first-class service on an airplane, reminiscent of the care and attention from our mothers when we were children.

How do we define a good embodied robot? It evolves from a passive machine to an automatic machine, and then to an active robot. In the next 10 years, the most valuable products of embodied intelligence will certainly possess automatic and proactive capabilities. The competition among automotive products will also hinge on the extent of these automatic and proactive capabilities, and this value will be integrated into high-frequency life and experiences, making it something you can't go back from once you have it.

The choice of electric vehicles is not bad, but it is insufficient. The choice of smart terminals is also not bad, but it is inadequate. Only by choosing the most challenging topic of embodied intelligence can we truly change users' lives and provide embodied intelligent robot products that offer automatic and proactive services. It just so happens that this is the automotive form of Transformers, the automotive form of Cars and Knight Rider. The new era presents the best opportunities and the highest challenges for automotive companies and entrepreneurs.

Next comes the choice of technology, more specifically, the choice of a complete AI system. Should it be oriented towards language intelligence in the digital world or embodied intelligence in the physical world? This means constructing entirely different system capabilities. To excel in embodied intelligence, the most important thing is to build an AI system that differs from language intelligence. It includes sensory capabilities equivalent to eyes and ears, modeling capabilities equivalent to a brain, and an operating system equivalent to a nervous system. The computational power of embodied intelligence is akin to a heart, and its ontology is equivalent to the human body. Currently, no third-party vendor can fully provide this system, and there isn't even one that can supply it.

The biggest characteristic of language intelligence is its focus on good models and computation, with larger-scale models. Larger-scale computation brings stronger capabilities, while embodied intelligence must enhance its perception models of the physical world, which must also be based on an understanding of the physical world. Precision is the primary task, followed by generalization. The operating system needs to ensure optimal integration of software and hardware, improving frame rates, and the overall system must be fast and precise. The computational power provided for the perception models and operating systems of embodied intelligence also needs to be supplied at the edge rather than in the cloud. Finally, hardware must be transformed into the ontology of embodied intelligence, such as three-dimensional neural control, wire-controlled, and active hydraulic suspension systems Improve the efficiency and accuracy of embodiment execution. If you embody intelligent understanding of the entire IT and AI systems, you will find that there are too many things that can be changed and are urgently needed to be changed.

First is perception. Based on the existing perception models and edge computing power, the current 3D Bev OCC occupies the effective perception distance of the network 2D Vit, which I am talking about as effective rather than the theoretical maximum of only over 100 meters, far less than the human eye. If upgraded to human vision, the effective distance of a 3D Vit with a similar working principle can be expanded by 2 to 3 times. Today, more than 50% of common issues in assisted driving can be naturally resolved.

3D Vit is not limited to the field of autonomous driving; proactive interaction and services with the vehicle owner outside the vehicle and proactive interaction and services with family members inside the vehicle also become possible. This requires significant breakthroughs in the research and development of perception models, and it must have chips like M100 that are custom-designed for embodied intelligence, along with a strong compilation team working efficiently together. Next is the model; only with 3D Vit can there be a true understanding of the physical world, and the VL in the VLA model can better perceive and understand the physical world. Human data can be used more efficiently to train visual models, and the generated data can better improve the training of the existing best computing platforms. A 4B MOE model runs at a frame rate of only ten hertz, while the execution system operates at sixty hertz. If the model's frame rate can be increased by 2 to 3 times, some comfort issues in assisted driving and response delay problems can be effectively resolved. This also requires deep modifications and customizations of traditional GPU architectures and computing power, as well as proprietary operating systems.

The M100 chip was developed to address these essential issues of embodied intelligence. Finally, the fastest response speed of human braking and steering is around 450 milliseconds, while the complete link from perception to execution in autonomous driving is around 550 milliseconds. The instinctive reaction from the driver's seat is much slower, similar to elderly drivers. The drive-by-wire system can enhance the entire link's response speed to 350 milliseconds. Don't underestimate this 200-millisecond difference; it could reduce the accident rate by over 50%, and make the driving experience feel better and safer than driving oneself.

It is both rational true safety and emotional security. Based on such demands, the entire control method will become completely different. If we only focus on the scale increase of language intelligence models, a doubling of model size results in only a 5% to 10% performance improvement from increased computing power. However, from the perspective of embodied intelligence as a whole, solving the most critical issues in each link of the entire system should lead to a performance improvement of 5 to 10 times in the next round of autonomous driving.

The ability of embodied intelligence to provide users with automatic and proactive rapid interaction and real services represents a qualitative change from having to not having. Over the past three years, our technical reserves for the complete system of embodied intelligence have given us confidence in our next-generation products. The era of embodied intelligence robots officially begins with automotive robots, and a hundred billion in revenue is just the starting point. The three key choices above determine the foundation for our second decade of development, which is more difficult and challenging than the previous decade We deeply understand that future competition is not about a moment in time, but about long-term directional choices and the determination of continuous investment.

Relying on a solid financial foundation, we will remain focused and use the management model of our beloved startup to create leading embodied intelligent products, ensuring that Li Auto can navigate through cycles, lead technology, and become a company that creates value for users and society in the long term. I also hope to communicate and interact with investors more in this way in the future, rather than through fixed-format reports every quarter. Thank you to our investors for supporting and trusting us even in our most difficult times. We will do our utmost to make Li Auto the best-performing company in the field of embodied intelligence and the company with the highest user value in the next 3 to 5 years. Thank you.

CFO Li Tie:

Hello everyone, I will now introduce some of our financial data for the third quarter. Due to time constraints, my remarks today will be limited to financial highlights, and all figures will be in RMB unless otherwise stated. For more details, we encourage you to refer to our earnings press release.

Total revenue for the third quarter was RMB 27.4 billion, a year-on-year decrease of 36.2% and a quarter-on-quarter decrease of 9.5%, which includes RMB 25.9 billion from vehicle sales, a year-on-year decrease of 37.4% and a quarter-on-quarter decrease of 10.4%, mainly due to a decrease in vehicle deliveries. The quarter-on-quarter decline was offset by a higher average selling price due to different product mix.

The cost of sales for the third quarter was RMB 22.9 billion, a year-on-year decrease of 22% and a quarter-on-quarter decrease of 5.3%. The gross profit for the third quarter was RMB 4.5 billion, a year-on-year decrease of 51.6% and a quarter-on-quarter decrease of 26.3%. The vehicle profit margin for the third quarter was 15.5%, compared to 20.9% in the same period last year and 19.4% in the previous quarter. The year-on-year decline was mainly due to estimated Mega recall costs and higher unit manufacturing costs resulting from reduced production volume. The quarter-on-quarter decline was mainly due to the same recall-related costs. Excluding these recall costs, the vehicle profit margin for the third quarter would have reached 19.8%. The gross margin for the third quarter was 16.3%, compared to 21.5% in the same period last year and 20.1% in the previous quarter. Excluding the aforementioned Li Mega recall costs, the gross margin for the third quarter would have reached 20.4%.

Operating expenses for the third quarter were RMB 5.6 billion, a year-on-year decrease of 2.5% and a quarter-on-quarter increase of 7.8%. Research and development expenses for the third quarter were RMB 3 billion, a year-on-year increase of 15% and a quarter-on-quarter increase of 5.8%. The year-on-year growth was mainly due to the impact of new model project progress, increased investment in expanding our product portfolio and technology, and costs arising from product configuration adjustments. The quarter-on-quarter growth was mainly due to the same product configuration adjustment costs.

Sales and administrative expenses for the third quarter were RMB 2.8 billion, a year-on-year decrease of 17.6% and a quarter-on-quarter increase of 1.9%.

The year-on-year decrease was mainly due to the recognition of equity incentive expenses related to CEO performance bonuses in the third quarter of last year. The operating loss for the third quarter was RMB 1.2 billion, while the operating income in the same period last year was RMB 3.4 billion, and the operating income in the previous quarter was RMB 827 million The operating profit margin for the third quarter was negative 4.3%, compared to 8% in the same period last year and 2.7% in the previous quarter. The net loss for the third quarter was RMB 624.4 million, while the net profit in the same period last year was RMB 2.8 billion, and the net profit in the previous quarter was RMB 1.1 billion. The diluted net loss per ADS attributable to common shareholders for the third quarter was RMB 0.62. In contrast, the diluted net income per ADS in the same period last year was RMB 2.66, and RMB 1.03 in the previous quarter. Regarding our balance sheet and cash flow, our cash position remains strong, with a balance of RMB 98.9 billion at the end of the quarter. The net cash used in operating activities for the third quarter was RMB 7.4 billion, while cash provided in the same period last year was RMB 11 billion, and cash used in the previous quarter was RMB 3 billion. The free cash flow for the third quarter was negative RMB 8.9 billion, compared to RMB 9.1 billion in the same period last year and negative RMB 3.8 billion in the previous quarter.

Now let's talk about our business outlook. For the fourth quarter of 2025, the company expects delivery volumes to be between 100,000 and 110,000 units, with total revenue expected to be between RMB 36.5 billion and RMB 39.2 billion.

This outlook reflects the company's current preliminary view of its business conditions and market conditions, which may change. This concludes our prepared remarks. I will now turn the call over to the operator to begin our Q&A session. Thank you.

Q&A Session

Host:

Thank you. If you would like to ask a question, please press the star key and then 1, and wait for your name to be announced. If you wish to withdraw your request, please press star 2. If you are using a speakerphone, please pick up the handset to ask your question. For the benefit of all participants on today's conference call, please limit your questions to two. If you have additional questions, you can re-enter the queue.

If you are a Mandarin speaker, please ask your question in Chinese first, followed by the English translation. The first question comes from CITIC Securities, please go ahead.

CITIC Securities:

Okay, thank you to the management team. I am from CITIC Securities, and I have two questions. First, we are pleased to hear CEO Li Xiang's reflections on past setbacks, and we are even more pleased to hear that the management plans to return to a startup-like state, along with a ten-year product planning and technology outlook. We also agree that products in the AI era are likely not just simple upgrades, but should be entirely new product designs based on user experiences in the AI era. Cars and embodied intelligence are also important carriers of physical AI, sharing similar underlying technologies and logic. However, we are more concerned about the fact that any generational research and switching of technology and products requires time and is difficult to achieve immediately.

So my first question is, if we only look forward to the next year, what technologies and products can we expect in 2026, and considering the next ten years, how far in the future can investors truly see a leap in the company's products and technologies and feel a real difference? My second question is about the three electric systems; the company's transition from range-extended to pure electric will inevitably face many challenges, and we are already encountering such challenges So, regarding the technology and industrial chain surrounding electric vehicles, what effective technological reserves and supply chain reserves have we made in the field of three electric systems for pure electric vehicles in the future? Can we provide investors with more information and confidence?

Li Xiang:

I am Li Xiang, and I will answer the first question. Yes, we believe that by 2026, when we start delivering products equipped with our M100 chip and AI system as the core, the real value and experience will change. I think the specific changes will still be what I mentioned before, which is that products like cars will transition from passive use to truly automated and proactive services for users. Moreover, this service and experience will not be about conducting deep research, coding, or generating videos, but will be reflected in our daily lives and high-frequency experiences, where we will feel and experience the fundamental changes brought by this value every day.

Yes, I think that’s the first point. The second point I want to make is what clear functions or values can be presented in the next decade. I believe that based on our understanding of artificial intelligence and embodied intelligence, it cannot provide clear functions like traditional programming or hardware features. Instead, when we build the systematic capabilities of the entire artificial intelligence system, as long as some key core issues are resolved or some key performance breakthroughs are made, it will bring a series of changes reflected in our specific lives or product values, and these changes may even exceed our expectations. I think this is our correct understanding of artificial intelligence and embodied intelligence, which can bring transformations to our physical world, products, and daily lives, as well as the driving forces behind them, rather than the traditional programming and functional presentation methods.

Ma Donghui:

I am Ma Donghui from Li Auto, and I will answer the second question regarding the research and supply layout in the field of three electric systems. In fact, Li Auto has always focused on core technologies in the three electric systems, conducting self-research and breakthroughs in electric drive, battery, and electric control technologies. In terms of electric drive systems, we mainly focus on user-valued aspects such as energy efficiency and driving experience. We have self-researched and manufactured silicon carbide power chips, and we have also self-researched and produced power modules and motor controllers. At the same time, Li Auto has built a dedicated factory for drive motors, thus establishing a full-chain self-research capability from silicon carbide power chips to power modules to motors.

Moreover, our electric drive technology is not only used in pure electric vehicles but also includes the entire range of extended-range models, ensuring a quiet and smooth driving experience while effectively optimizing energy consumption levels and improving vehicle range. Regarding battery technology, we focus on charging speed and safety, and we have built full-stack self-research capabilities around 5C ultra-fast charging batteries, including the chemical system of battery cells Then the battery management system, BMS control model and algorithm, as well as the layout and structural design of the battery pack, can achieve the three core advantages of ultra-fast charging, long endurance, and long lifespan.

In terms of battery supply layout, our plan is to adopt a dual model of external procurement combined with self-developed joint ventures. Next year, we will also mass-produce the Li Auto self-owned brand 5C battery, and through the industrialization of self-developed technology, further enhance the safety performance of the battery and upgrade the user’s charging experience. The last point is that in terms of electronic control, we will create an ultimate driving experience through full self-research in both hardware and software. On the software side, we can achieve control of power, energy management, and engine calibration, etc. The full-stack self-research, and on the hardware side, core domain controllers, PCBs, layouts, and underlying software are all independently developed by us. Through the synergy with self-developed chassis technology, we can achieve precise optimization of vehicle drivability and smoothness, creating a driving characteristic that is easy to handle. Overall, through the self-research of the three electric technologies, we hope to bring users a fast charging, long endurance, easy to drive, and safe experience with strong technical capabilities.

Host:

Thank you! The next question comes from Morgan Stanley. Please continue.

Morgan Stanley:

Hello management, I am Tim from Morgan Stanley. First, thank you for sharing the strategic vision for the next 10 years and giving me some insights, and thank you for accepting my questions. I have two brief questions that I would like to return to the short-term operational challenges and progress. The first question is about pure electric vehicles, I would like to understand the current order situation and delivery status of the Li Auto L6 and L8, and how the company is addressing the supply issues? What are the expectations for future sales? The second question is about cash flow. We see that the company's cash outflow from operating activities in Q3 has further expanded, which may lead to a decrease in cash reserves. What is the main reason for this? How do we view the improvement of cash flow in the future?

Ma Donghui:

Hello, I am Ma Donghui from Li Auto. I will answer the first question. This year, we have built a BEV product matrix through the Li Auto L6 and L8, covering the mainstream and high-end family pure electric markets, laying a solid foundation for the long-term stable growth of Li Auto's pure electric business. For us, we have also completed the layout of three energy forms: extended range and pure electric, achieving effective complementarity to meet different user needs. For pure electric models, the core highlight is that the i series products have achieved a significant breakthrough in regional penetration, especially in core markets for pure electric vehicles such as Shanghai, Beijing, and Jiangsu-Zhejiang-Shanghai, where orders in September have significantly increased. Regarding supply, the Li Auto I8 and I6 products are currently experiencing a ramp-up in production capacity, accelerated delivery, and market penetration development Starting from November, in order to address the capacity ramp-up, the supply of I6 batteries will officially adopt a dual-supplier model. However, we will ensure that the performance and quality standards of the batteries from both suppliers are consistent. It is expected that by early next year, the monthly production capacity of the I6 will steadily increase to 20,000 units. We would also like to sincerely apologize to the order owners of the Li Auto I6. Due to the planning of core component supplies and the capacity ramp-up, your vehicle delivery time has been delayed. Thank you for your trust and choice, and we kindly ask for your continued support and time. We will do our utmost to work overtime to increase capacity and accelerate the delivery process. Thank you.

Li Tie:

Regarding the second question, I believe there are two main reasons for the operating cash flow. First, as we pointed out in our last earnings report, we faced significant pressure in deliveries during the third quarter. The decline in delivery volume leads to a decrease in revenue, which ultimately affects the performance of accounts receivable (RPD) in cash flow, along with the impact of shortening the payment cycle to suppliers. As you may know, this is due to the measures implemented nationwide by government authorities starting in June. In fact, we value our relationships with supply chain partners and actively respond to their requests. Currently, our assessment period for all accounts payable is 60 days, and the payment method is online transfer or bank notes, without any commercial notes or any form of proof. No, it is just ordinary bank notes. Thank you.

Host:

Thank you! The next question comes from Bank of America. Please go ahead.

Bank of America Securities:

I have two questions to ask. The first question is how does the company view the changes in the replacement subsidy policy next year and the impact of the purchase tax increasing to 5 percent? If the replacement subsidy declines, what will be the corresponding sales strategy of the company? My second question is regarding the significant facelift of the L series in 2026. What highlights can we expect?

Ma Donghui:

I am Ma Donghui, and I will answer these two questions. The first question is about the replacement subsidy and purchase tax. We believe this is a process where the automotive industry shifts from policy-driven to market-driven, which also highlights the core competitiveness of enterprises. The withdrawal of the purchase tax policy will certainly cause short-term market fluctuations, as users will concentrate on locking in discounts by the end of 2025, leading to a tail effect. Additionally, sales in Q1 2026 will naturally show a quarter-on-quarter decline. However, in the long run, we remain optimistic, and the penetration rate of NVE will steadily increase. We estimate that by 2026, the penetration rate of new energy vehicles in China may reach 55% to 60%, and for the high-end market, the penetration rate of NVE will exceed 60%. As for Li Auto's response strategy, first, during the transition period, we must ensure the interests of users, and our new models developed in 2020 must comply with the new regulatory standards. During the transition period, we have also launched a peace of mind car purchase plan, such as covering the difference in purchase tax for users who lock in orders for the I6 before the end of the year for cross-year delivery Then, all 26 of our products have met the requirements of the new policy regarding fuel consumption and energy consumption, allowing us to enjoy the benefits of the new policy in 2026.

In the long term, we still need to strengthen user value, and there is no shortcut to hedge the impact through technological upgrades. By 2026, we will adopt an 800-volt high-voltage platform and 5C ultra-fast charging batteries across the entire range to improve efficiency and reduce energy consumption. Additionally, we will establish approximately 4,800 ultra-fast charging stations by 2026, with the proportion of high-speed charging piles expected to exceed 35%. Through deepening the localization of the supply chain and achieving economies of scale, stable pricing is crucial for product strength. We must accelerate iteration and further accelerate iteration. In summary, the gradual retreat of this policy will definitely be a watershed for the high-quality development of the industry. We are confident that in 2026, we will achieve a historical breakthrough in delivery volume, and we will cross the cycle with product strength and user value, consolidating our leading position in the high-end market. Thank you.

Regarding the second question, concerning the timing and highlights of the 2026 L series model refresh, we should officially announce the specific release time and details of the product at an appropriate time. However, today I would like to reveal a little. First, the major refresh of the 2026 L series is based on user feedback and core technology accumulation, aiming to create an iterative product that aligns with our overall strategic goal of regaining the leading position in range-extended products in 2026. In terms of vehicle configuration, we will review and return to a streamlined SKU model, balancing market coverage and supply efficiency, hoping to achieve a full core experience and completely eliminate the pain point of a discounted entry-level experience.

In terms of design, we will continue the classic design genes of the L family while enhancing the luxurious feel. We aim to balance the brand's high recognition with the freshness of user experience, creating a product form that better meets family needs. In terms of technology, the entire range will be equipped with 5C ultra-fast charging technology, which will better coordinate with our current pure electric ultra-fast charging network, effectively alleviating energy replenishment anxiety. At the same time, relying on our first-mover advantage and technological accumulation in the range-extended field, we will further strengthen the brand recognition of Li Auto as the leader in range extension. The core of the 2026 L series major refresh is to respond to market uncertainties with certainty in technological upgrades, delivery rhythm, and user value. Thank you.

Host:

Thank you! The next question comes from UBS. Please go ahead.

UBS:

Thank you to the management for accepting my question. I have two questions. The first question is about the Mega recall. As we understand, this recall was announced in the fourth quarter. Why is this cost recorded in the third quarter? How was this amount determined? How much is borne by the supply chain, and how much is borne by us? What impact will it have on the gross margin in the fourth quarter? Additionally, could you provide an update on the current recall situation and share the latest orders for the Mega? This is the first question. Um, the second question is about the AI side. Could the management please share the current progress of VLA and the event model, user feedback, subsequent R&D goals, and iteration progress?

Li Tie:

Regarding this, I think I will (answer) soon. Give me some time. We will answer your questions shortly. First of all, we recognize that this was a test in the third quarter. We regard this event as a follow-up event, so it will be accounted for in the most recent quarter we can record. So this is a basic or coverage standard. Regarding the recall, I believe we have an announcement. I don't want to repeat most of the details that have already been covered. Currently, we are just manufacturing all better battery packs to meet the recall requirements and demands, which means we have reduced our deliveries for 2025 Mega. So this means all better battery packs, most of which we are transferring to replace the recalls for 2024. I think this best serves the interests of our customers.

Ma Donghui:

I am Ma Donghui, and I will answer the second question regarding the current progress of VLA and the speed of iteration. As for Li Auto, the VLA large model was actually pushed to the ad Max models in September, and with the model iteration's migration capability, we quickly covered from the new i series to the 2022 Li L9, allowing all users to simultaneously experience its core capabilities, enabling older vehicles to enjoy new technology and experiences. Through user feedback and data analysis, the upgrade in experience has shown significant results.

Especially for i series owners, the willingness to use intelligent driving is very strong, and user daily activity as well as the average mileage before takeover (MPI) are both showing an upward trend. Users also generally feedback that VLA's vertical control is smoother, and the decision-making for detours is more decisive, with significant improvements in route accuracy at complex intersections. In the future, we will continue to iterate and achieve continuous breakthroughs in functionality.

The OTA 8.0 is the first full push version, which was optimized with safety as the core experience. In early December, we will push OTA 8.1 to further enhance VOA's perception capabilities, making responses more precise. By the end of December, we will conduct an architectural upgrade, focusing on strengthening the interaction of language and behavioral information, optimizing the decision-making process, and adapting to the 2026 self-developed chip M100. In the future, we will also have more innovative features to be launched, introducing the industry's first defensive AES function to upgrade safety protection capabilities. We are also exploring parking from any parking space to any parking space in all scenarios, and combining our self-built supercharging stations to achieve intelligent car-finding functions, truly perfecting the intelligent travel ecosystem. Thank you

Host:

Thank you! The next question comes from Goldman Sachs. Please continue.

Goldman Sachs: Thank you to the management for accepting my question. I am Tina from Goldman Sachs. Due to time constraints, I will ask the last question. The company just mentioned our self-developed chips, so I would like to ask about the progress and future plans regarding the self-developed chips and the open-source operating system.

Li Auto CTO:

Let me answer your question. We believe that AI infrastructure systems are the core foundation of smart vehicles. To achieve this efficiency, the system must be designed as an integrated architecture rather than pieced together as independent parts. Our self-designed controllers, hardware, and operating systems allow us to reduce development time from the industry average of 15 months to 9 months, while also lowering costs by 20%. Many modules in the inference stack still come from suppliers. To innovate faster, we have open-sourced Halo OS, enabling our partners and ecosystem to co-develop. In September, we established the Halo OS Technical Steering Committee and assisted various companies in the smart vehicle value chain in signing the community charter, including vehicle manufacturers, chip manufacturers, software and hardware service providers, and parts suppliers.

At the same time, we are developing our own foundational model for physical AI vehicles. Our focus is on enhancing perception, understanding, and response capabilities. Therefore, this model can see further, understand better, and react faster.

AI inference chips are the computing engine of this system. We are currently conducting large-scale system testing with controllers built using our self-designed chips, the M100. We expect to commercialize it next year. In collaboration with our foundational model compiler and software system, we anticipate that the M100 will achieve at least three times the performance-cost ratio of today's high-end chips in our next-generation VLA-based autonomous driving system, based on efficient AI inference and execution systems. Our next priorities will be faster iterations, continuous performance improvements, and lower costs. Development of our next-generation platform and chips has already begun. Thank you.

Host:

Thank you. As our conference call is now coming to an end, I would like to turn the call back to the company for closing remarks.

Thank you again for joining us today. If you have further questions, please feel free to contact the investor relations team at LEODO. That concludes today's call. Thank you