Li Auto announced its financial results for the fourth quarter and full year of 2025. The report shows that in Q4 2025, Li Auto's total revenue was 28.8 billion yuan, a decrease of 35% year-over-year but an increase of 5.2% quarter-over-quarter.
Following the earnings release, Li Auto's management held a conference call. Chairman and CEO Li Xiang, CFO Li Tie, President Ma Donghui, and Chief Technology Officer Xie Yan participated and answered analyst questions.
The following are the key points from the Q&A session:
Tim Hsiao, Morgan Stanley Analyst: I have two questions. The first concerns the channel reform: A news report in late February suggested Li Auto plans to deepen channel reforms and close 100 stores. Could you share the specific plans and progress of the store optimization? Also, regarding the store partnership plan mentioned earlier, besides the store manager incentive mechanism, are there other implementation details? When can we expect to see concrete results? The second question is about products: Could management share the launch timeline, pricing strategy, product competitiveness, and profitability of the new-generation Li L9 and L9 Ultra models?
Ma Donghui: First, the rumor about closing 100 stores is inaccurate. We routinely close and replace a small number of underperforming stores that fail to meet sales targets, primarily to address issues like poor initial location selection or declining foot traffic in commercial areas. This is part of our normal operational optimization.
Our core strategy for channel layout this year is "quality over quantity." We will still open new stores this year, prioritizing locations in top-tier shopping malls and premium auto hubs to enhance brand influence and attract high-quality customer traffic. In terms of city coverage, our network in lower-tier cities is already relatively well-developed. Moving forward, we will focus on increasing density in higher-tier cities, aligned with the growing sales of our pure electric models. We are also continuously improving the delivery experience, encompassing the entire process from store reception and test drives to final delivery, including dedicated staffing at highway supercharging stations during holidays. User satisfaction and positive feedback rates are steadily increasing.
Let me also take this opportunity to introduce the store partnership mechanism officially launched in early March. The core idea is to treat each store as a fundamental business unit, creating a direct sales model with Li Auto's unique characteristics. Firstly, we will adhere to the direct sales model to ensure consistent service experiences and pricing policies nationwide. Simultaneously, by delegating operational decision-making authority and profit-sharing rights to store managers, we aim to stimulate their business acumen and initiative. This decision-making authority includes autonomy in customer acquisition, daily operations, and team management.
Our evaluation of store managers has also shifted from a singular focus on sales volume to assessing the overall business results of the store. We hope managers will treat their stores as their own ventures. This mechanism fundamentally addresses past issues like reckless store expansion and uncontrolled external displays. For future new store locations, store managers will be involved throughout the process, with evaluation responsibilities directly tied to individuals, improving store quality from the outset.
The company will also fully empower frontline teams through financial support and digital tools, aiming for significant improvements in sales operations by the third quarter of 2026. Finally, we firmly believe that the health and efficiency of our sales and service system are core to maintaining our sales volume and market position. Starting last August, we spent nearly seven months systematically restructuring the management system of our direct sales model, including high-quality store openings, refined operations, store manager incentives, frontline training, and building functional mechanisms, all to solidify the long-term competitiveness of our sales and service system.
Li Xiang: I will answer the second question. The new-generation Li L9, equipped with our self-developed MACH 100 chip, will be officially launched in the second quarter of this year. This model is defined as an Embodied Intelligence product because we have fundamentally restructured the entire technical system across three dimensions: perception, brain, and embodiment.
The most critical aspect is the perception layer. Whether it's intelligent driving, autonomous driving, or the current embodied intelligence, the industry's biggest technical challenge is essentially using a "watch video" approach for model training, learning, and even vehicle control. This widely adopted industry system does not truly understand the physical world; it merely imitates operations. At the perception level, we will break through the current 2D camera + LiDAR solution and upgrade to a 3D visual solution, enabling the vehicle to perceive and understand the physical world like a human, rather than simply recognizing through video.
This breakthrough requires joint design and innovation across the entire chain, from video encoders and model development to the chips providing computational support for the encoders and models. I believe this will be the most significant technological breakthrough in autonomous driving and embodied intelligence globally this year. Consequently, visual large models will be truly implemented, moving beyond simple object recognition to understanding and reasoning about the physical world like humans—a crucial change.
On the other hand, as the machine's "embodiment," the vehicle itself will undergo major changes. The Li L9 will feature a comprehensive by-wire system, including steer-by-wire, brake-by-wire, four-wheel steering, and an 800V fully active suspension controlled by four independent motors. This makes the vehicle's response speed and safety far superior to traditional cars. More importantly, vehicle control will no longer rely on traditional MCU controllers transmitting commands step-by-step; instead, large models will directly output instructions to the control systems.
We believe this technical system will become the core standard for future automotive embodied intelligence, and all these technologies will be implemented in the new-generation Li L9.
Paul Gong, UBS Analyst: My first question is about the 2026 sales target: Based on current market conditions, what is the sales target for 2026? More importantly, while balancing volume and profit, how does the company view the importance of the sales target? The second question concerns the impact of rising raw material costs: Prices for batteries, memory chips, copper, aluminum, and other precious metals have recently increased. What measures is the company taking? Will costs be absorbed through the supply chain, or will some be passed on to vehicle prices?
Li Xiang: 2026 is the year we deliver products from our third-generation technology platform. We are confident in our product and technological competitiveness this year. However, we also recognize that competition in the entire new energy vehicle market is intensifying. In the mid-to-high-end market segment above 200,000 yuan, the number of new models launched this year exceeds the total from the past few years, while overall market growth is very limited.
Considering these two factors, our sales target for 2026 is to achieve a year-over-year growth of over 20%. To reach this goal, we have formulated a "3+2" core strategy. The three main pillars are: First, effectively manage the sales system. We firmly believe that the efficient management of our direct sales system will be a long-term core competency. Second, ensure the successful generational update of the L-series, with the L9 at its core. Every step, from product launch and production ramp-up to delivery service, must be executed flawlessly. Third, drive stable volume increases for pure electric models. Whether it's the Li L6, L8, Mega, or the pure electric flagship L9 launching in the second half, we must thoroughly resolve past issues like supply and product launches, establishing our pure electric products firmly in the mid-to-high-end market.
The two supporting strategies are: First, our sustained investments in intelligent technology over the past few years—including chips, models, and technical research—will be transformed into differentiated product experiences this year. These experiences will be proactive, high-frequency, and integrated into users' daily driving scenarios. Second, we expect significant progress in overseas market expansion. This year marks our first official year of overseas布局, representing an important long-term growth opportunity. This "3+2" strategy is our core guarantee for achieving over 20% sales growth in a fiercely competitive market.
Ma Donghui: I will answer the second question. Indeed, the recent price increases for core components like batteries and memory chips have put significant pressure on per-vehicle costs. Facing this challenge, our response focuses on three main areas:
First, strengthen collaboration with supply chain partners to stabilize prices and ensure supply. On the cost side, we have signed long-term agreements (LTAs) with key suppliers to lock in prices for critical raw materials and components in advance, hedging against short-term market fluctuations. On the supply side, addressing tight supply for key intelligent components like memory chips, we have secured supply quotas with core suppliers in advance to meet the demands of new vehicle launches and production. For contracts with price adjustment mechanisms, we strictly adhere to them. For those without, we insist on sharing costs and weathering cycles with suppliers for mutual benefit, and we thank our supply chain partners for their long-standing support.
Second, promote end-to-end cost reduction efforts, exploring savings opportunities across the entire chain from product R&D, manufacturing, and logistics to quality control. Simultaneously, we aim to improve parts reuse rates through platform-based R&D, leveraging economies of scale to absorb external price pressures internally as much as possible. For instance, our self-developed and self-manufactured range extenders, electric drive power modules, self-developed but outsourced domain controllers, silicon carbide power chips, the MACH 100 autonomous driving chip, and customized in-car Pads all effectively help us control costs.
Third, new model pricing will be more reasonable and稳健. For this year's new models, we will comprehensively consider raw material price fluctuations, R&D investment, and user value to ensure the company's healthy and sustainable profitability, allowing the gross margin for new products to return to normal levels.
Overall, we are confident that through supply chain collaboration, long-term price locking, platform-based and technological cost reduction, plus reasonable product pricing, we can keep the impact of rising raw material costs within a controllable range, ensuring the stability of the company's gross margin and operational quality.
Jing Chen, CICC Analyst: My first question is: Recent media reports suggest the company is considering a share repurchase. Are there any relevant plans? The second question concerns R&D investment planning: In 2025, 50% of the company's R&D investment was directed towards AI-related fields. Could you provide guidance on the total R&D expenditure for 2026 and the expected proportion allocated to AI/Embodied Intelligence?
Li Tie: As a company with dual primary listings in the US and Hong Kong, we view share repurchases as one tool to enhance shareholder value. However, there are currently no disclosable arrangements regarding this matter.
Regarding the second question, in 2026, the company's R&D expenses are expected to remain around 12 billion yuan. AI-related investments will still account for approximately 50% of this. This portion includes R&D for AI infrastructure like inference chips and computing power, as well as AI product development such as autonomous driving systems. It's important to note that we do not view the automotive business and AI business as separate segments. Instead, we integrate the development of AI capabilities into the company's overall business model. All R&D investments will serve the upgrade of our existing business system.
Analyst: My first question is about the i-series pure electric models: Could you provide details on the orders and sales of the Li i6 and i8, as well as the production ramp-up progress for the i6? Also, how does management evaluate the safety of Sunwoda batteries and their contribution to future cost reduction? The second question is about the self-developed MACH 100 chip: Could you detail the mass production timeline for this chip and its specific role in reducing costs and increasing efficiency? Furthermore, how should we understand the company's "integrated hardware and software" strategy? When will the gaps between automakers resulting from this strategy become apparent?
Ma Donghui: Since its launch last July, the Li i8 has seen continuously improving word-of-mouth regarding driving experience, charging convenience, and intelligent features as user mileage accumulates. Its Net Promoter Score (NPS) has increased by over 20% compared to the initial launch period. During the Spring Festival, our 5C supercharging service and OTA 8.3 autonomous driving system upgrade were highly praised by users, and the i8's NPS reached a record high. In J.D. Power's second-half 2025 survey, the Li i8's NPS ranked first among mid-to-large SUVs. Leveraging this positive reputation, orders for the i8 have steadily recovered, with March orders up 180% compared to January, indicating sustained strong market demand.
Regarding the Li i6, we have passed the most challenging phase of production ramp-up and entered a stable delivery stage. The i6's product strength has been fully validated; its exterior design, interior space, energy consumption, and handling performance precisely meet the needs of young families, resulting in strong order performance post-launch. Currently, supply chain bottlenecks for the i6 have been completely resolved. The short-term battery supply fluctuation issue was addressed through collaboration with our core supplier, which increased capacity. We also promptly introduced purchase tax subsidies and care policies for delivery delays, and we especially thank Li i6 users for their understanding and patience. Going forward, the monthly delivery capacity for the i6 will reach 20,000 units, and existing orders will be efficiently delivered within the next 1-2 months. More importantly, the success of the i6 demonstrates that Li Auto's brand momentum has successfully extended from extended-range electric vehicles (EREVs) to the pure electric market.
Regarding battery strategy, we adhere to the principle of open cooperation, partnering with leading industry players, but we maintain主动权: we lead the battery solution design based on overall vehicle performance requirements and enforce strict quality control at every level. Regardless of the battery supplier, their products' performance, quality, and safety must meet Li Auto's unified standards to ensure a consistent user experience. Additionally, starting in 2026, all Li Auto models will be equipped with batteries co-branded by Li Auto and CATL, signifying a deeper level of integration with our core partner. Please trust the Li Auto brand, because the quality of our products is never determined by a single supplier. It is defined by our full-stack self-developed technology, stringent quality control system, and long-held values. Choosing Li Auto means choosing the most reliable assurance.
Xie Yan: I will answer the second question. The MACH 100 chip will be delivered to the market simultaneously with the new-generation Li L9 series, and mass production of the chip has already been achieved. The core advantage of the MACH 100 chip is its ability to deliver higher effective computational power per unit of silicon area, providing greater design flexibility for our visual large model architecture. For example, the chip can run a visual large model with a parameter scale 6 times larger and computational power 10 times greater than the previous generation, while still ensuring high frame rates and fast inference.
More importantly, with the coordinated iteration of our self-developed models, compiler architecture, and operating system, the true potential of our full-stack self-developed autonomous driving system is being unleashed. While the current performance improvements are significant, the core benefit is that this system-level integration will substantially accelerate the iteration speed of our autonomous driving capabilities. Once the system is officially deployed, the pace of our ADAS upgrades will significantly increase.
The MACH 100 chip is also deeply integrated with Li Auto's vehicle operating system and by-wire vehicle control system, enabling tighter coordination between pre-processing and post-processing of autonomous driving computational tasks and vehicle control actions. This reduces the end-to-end latency from sensor photon input to vehicle actuator response to 200-300 milliseconds, directly enhancing the user's driving experience. Higher local computational power also enables the vehicle to perform more intelligent functions beyond autonomous driving, making the car increasingly resemble an intelligent robot. These features will debut on the new-generation L9 and continue to expand in the future.
In terms of cost reduction and efficiency improvement, the MACH 100 chip delivers triple value: First, the material cost per chip is significantly lower than externally sourced solutions. Second, by replacing the MCU controllers of the previous platform with the MACH 100 chip, combined with the virtualization technology of Li Auto's OS, we can save over 1,000 yuan per vehicle. Third, leveraging the data flow architecture and the joint design of the chip and models, the chip's operational efficiency is greatly enhanced, with ample room reserved for future performance upgrades.
We predicted back in 2022, when we started self-developing chips, that the industry would enter a new phase of co-design for models, chips, and operating systems by 2025. This vertical integration will create genuine differentiation for companies in terms of performance, efficiency, and user experience. The gap between automakers in the future will resemble that between Apple and the Android阵营 in the smartphone industry. Once full-stack hardware-software integration is achieved, it creates a structural advantage that continues to widen.
Tina Hou, Goldman Sachs Analyst: My first question is about embodied intelligence: What is the company's strategic plan for the embodied intelligence field over the next 1-2 years? What will be the form and timeline for product implementation? Among smart cars, Robotaxis, and humanoid robots, what is the company's strategic priority? The second question is about R&D organizational adjustments: Has the restructuring of the intelligent R&D organization been completed? Under the new organizational structure, what is the progress of the intelligent driving business?
Li Xiang: Regarding our embodied intelligence strategy, we can look at it from two levels: First, in terms of full-stack vertical technology integration for embodied intelligence, we will dedicate 100% of our resources. We believe embodied intelligence shares commonalities at the system technology level, including edge inference chips, various large models, operating systems, and the underlying end-to-end data and training systems. This is the core foundation for all embodied intelligence products, and we will go all-in on R&D.
Second, in terms of commercialization and product implementation, we will remain cautious and continue exploring. When expanding beyond automobiles, we will incubate projects like a startup, including AI glasses, robotics projects, etc. We will abandon the "big spending" R&D model of large corporations, letting entrepreneurial teams lead product exploration to avoid resource waste. This is our core strategy for embodied intelligence.
Li Xiang: Regarding the second question, in January this year, we completed a major organizational adjustment. The core was a restructuring of the R&D team management model, shifting from a division based on hardware and functions to a model centered around "creating digital humans, silicon-based humans." We reorganized the R&D teams according to the logic of "human organs":
First, the "Basic Organs" team integrates teams responsible for chips (heart), datasets (lungs), operating systems (nervous system), etc., under one management system. This forms the foundational support for embodied intelligence.
Second, the "Brain System" team integrates teams for perception, model pre-training, post-training, reinforcement learning, and underlying computing infrastructure. This team is explicitly required to focus solely on foundational R&D, avoiding specific applications, preventing the "brain" from being distracted by tasks meant for "limbs."
Third, the "Software Embodiment" team, similar to application teams at OpenAI or Anthropic, focuses on building system-level intelligent agents (Agents). This includes software tool development, MCP protocol R&D, agent construction, memory systems, and skill system development, all requiring deep integration with the vehicle OS. The core goal of this team is to enable agents to perform actual tasks, not just engage in dialogue or output text/images. This team is also required to focus on application落地 and avoid底层 deep learning R&D, preventing the "body" from growing an inefficient "brain," thereby improving R&D efficiency.
Fourth, the "Hardware Embodiment" team is dedicated to embodied intelligence R&D, integrating teams for energy systems, drive systems, control systems, etc., no longer managed by traditional single controller divisions. We have already achieved direct command output from large models to all actuators via the MCP protocol, realizing integrated hardware management.
This adjustment has significantly improved collaboration efficiency within the R&D teams. For example, the iteration speed for autonomous driving models has increased from once every two weeks to once per day. It has also broken down departmental silos; previously dispersed teams now work together in co-located offices, with the unified goal of creating embodied intelligence products and building a human-like intelligent system.
Let me add one more point: From the end of last year to the beginning of this year, several core R&D managers who had grown with the company chose to leave to start their own businesses. These individuals accompanied Li Auto through the startup phase from 0 to 1, gained rich experience, and earned recognition in the capital markets. First, we congratulate these key talents and hope they become mainstays in the market competition. Secondly, this change has also created opportunities for young technical and business managers within the company. Now, in core areas like foundational models, humanoid robots, and product lines, post-90s and even post-95s generations have begun taking leadership roles (一号位). Even among our campus recruits, many post-00s have become core contributors to technical solutions and high-difficulty research papers.
This is a win-win situation for Li Auto: externally, it incubates excellent entrepreneurial forces; internally, it builds a solid梯队 of young talent. This is also a core reason for our confidence in the next decade. Everyone can rest assured that the new generation of young management teams can effectively take the baton and, building upon the foundation laid by their predecessors,迎接 even greater challenges.
Comments