Li Auto Accelerates AI Transformation, but Internal Vision Lags Market Expectations

By the end of 2024, Li Xiang, founder of Li Auto, declared in an extensive interview that Li Auto is not a traditional automaker but an artificial intelligence (AI) company. To fulfill this positioning, Li Auto launched multiple initiatives in 2025: restructuring its organization to reduce reliance on external technologies, open-sourcing its Xinghuan OS operating system, initiating an in-house chip development project, and unveiling its first AI glasses. In early 2025, Li Xiang convened an impromptu all-hands meeting, focusing on AI development trends and key milestones. However, many employees expressed confusion or questioned the meeting’s practical value on internal platforms afterward, revealing a lack of consensus within the company regarding its AI strategy. Analysts noted that Li Auto’s team has long prioritized short-term goals, making Li Xiang’s recent emphasis on a long-term technological vision a stark contrast to his previous image as a “pragmatic product manager.” On June 15, 2025, Li Auto held its Livis Day event in Beijing, showcasing software and embodied intelligence advancements. Li Xiang outlined the company’s decade-long goal: “to imbue cars and homes with life.” At the event, Li Auto unveiled its self-developed Mach M100 chip, the large AI model Mach Mind, the autonomous driving system Mach VLA, and a full-year OTA upgrade roadmap—promising a new version rollout in Q3 and capabilities matching Tesla’s FSD V14 by Q4. The Mach M100 chip is built on a 5nm automotive-grade process, delivering 1,280 TOPS of computing power per chip with over 82% real-world operational efficiency. The dual-chip-equipped Li L9 Livis achieves a total system compute capacity of 2,560 TOPS. Li Auto claims it has achieved full-stack in-house development—from chips and compilers to operating systems, AI algorithms, and domain controllers. Li Xiang defined an “embodied intelligent vehicle” as a “quadruple integration”: an electric vehicle, a professional driver (autonomous driving), an AI computer (computing foundation), and a lifestyle assistant (cabin AI)—aiming to redefine the competitive dimensions of smart vehicles. However, critics argue the concept lacks concrete implementation scenarios and differs only marginally from prevailing industry technical pathways. Although some owners responded positively—describing Li Auto as “leaps ahead” in intelligent driving—the capital markets remained unimpressed: Li Auto’s stock price fell 1.5% on June 16 (the day after the event) and dropped another 3.75% the following day. Technologically, Li Auto faces a disadvantage in data scale. Tesla’s FSD leverages real-time data from millions of vehicles, while Huawei’s ADS benefits from broad partnerships across automakers. In contrast, Li Auto’s ADAS mileage still lags by orders of magnitude. To compensate for insufficient real-world data, Li Auto plans to use high-compute chips to support synthetic data generated via simulation for training world models. However, this strategy carries significant risk. XPeng’s in-house Turing chip entered mass production in Q3 2025, with cumulative shipments exceeding 200,000 units and a target of nearly one million units for 2026. Meanwhile, Li Auto’s chip has just completed its initial vehicle integration and has yet to undergo large-scale production validation or establish a functioning data feedback loop. Currently, Li Auto is undergoing a strategic shift from product-driven to technology-driven growth—but a clear gap persists among internal employee understanding, user perception, and capital market expectations.