Li Auto has released its 2025 financial report, marking its first earnings statement framed around an AI company identity. The document reveals a company in the midst of a profound self-reassessment, characterized by a striking contrast: three consecutive years of profitability, cash reserves exceeding 100 billion yuan, yet research and development expenditure hitting a record 11.3 billion yuan, with over half allocated to AI. As the most profitable among China's new automakers decides to spend like a tech firm, key questions arise about its view on profits, the allocation of funds, and organizational transformation. This analysis delves into the strategic logic behind this unconventional automotive earnings report through five core questions.
On March 12, Li Auto announced its 2025 financial results: revenue reached 112.3 billion yuan, achieving profitability for the third consecutive year, with a net profit of 1.1 billion yuan. To some investors, if measured by traditional automotive manufacturing standards, this profit appears modest relative to revenue. However, when viewed through the lens of a technology company, its balance sheet holds a staggering 101.2 billion yuan in cash reserves.
In the landscape of Chinese tech firms, Li Auto has always been an outlier. Historically, it stood alone as the only new automaker capable of sustained profitability, often described as the most fiscally prudent car company. Now, it is directing over 50% of its earnings into a bold AI venture—developing its own chips, heavily investing in large models, and incubating robotics projects.
The 2025 report functions as a dual-sided mirror: one side reflects the stability of manufacturing, while the other projects the ambitions of a tech company. What broader strategy is Li Auto pursuing? The answers lie within five critical areas of inquiry.
Why emphasize "three consecutive years of profitability"? In 2025, Li Auto reported a net profit of 1.1 billion yuan. Viewed in isolation, this figure might suggest a mismatch between profit and revenue scale. However, a broader perspective reveals several telling details: this marks the third straight year of profitability, making Li Auto the only domestic new automaker to achieve both over 100 billion yuan in annual revenue and sustained profitability for three years. More importantly, by the end of 2025, Li Auto's cash reserves amounted to 101.2 billion yuan, placing it in the top tier among all Chinese automakers.
Maintaining profitability during a period of strategic investment is key to understanding this report's contrasts. The term "strategic investment" is quantified by the annual R&D expenditure of 11.3 billion yuan, a historic high, averaging nearly 1 billion yuan per month, with AI-related investments accounting for 50% of this total.
Why is AI-related investment so high? Li Auto has redefined its identity, no longer viewing itself solely as an automotive manufacturer but increasingly as a technology entity evolving toward "embodied intelligence." Within this framework, the 1.1 billion yuan net profit signifies the ability to maintain financial health while undertaking a large-scale transformational investment.
The 101.2 billion yuan cash reserve underpins this reassessment. Amid ongoing price wars in the new energy vehicle sector and intensifying technological competition, cash provides strategic flexibility. It supports the mass production and integration of self-developed chips, funds technological breakthroughs in steer-by-wire chassis systems, and offers room for experimentation in new ventures like AI glasses and humanoid robots, which are incubated with a startup-like approach.
During the earnings call, Li Xiang, Chairman and CEO of Li Auto, clearly articulated this positioning: "In the vertical integration of embodied intelligence technologies, we are 100% committed. We believe embodied intelligence shares common systemic technologies, including edge-side inference chips, various models and operating systems, and the underlying data and training frameworks. This is where we will focus our efforts entirely; we consider it the most critical foundation for all embodied intelligence."
This explains the report's apparent contradictions: leading revenue scale without "super profits," massive R&D spending alongside substantial cash reserves. This profile does not resemble a traditional manufacturing firm focused on short-term profit maximization but rather a technology company undergoing a generational technological leap, exhibiting typical characteristics of a transition period.
Where is the money going? Half is invested in intangible AI. Discussing Li Auto's 2025 R&D expenditure, Li Tie, Chief Financial Officer, stated during the earnings call: "We see cars and AI as interdependent. The core of our R&D investment is building AI capabilities and integrating them into our existing business model. In our current model, all R&D investments are unified; AI R&D is not a separate business unit."
The most direct manifestation of this integration is the upcoming mass production of Li Auto's self-developed chip, the Mahe 100. Scheduled for integration into vehicles in the second quarter of this year, the Mahe 100 delivers three times the effective computing power of Nvidia's Thor-U when running the VLA large model, significantly enhancing the safety and fluency of intelligent driving assistance.
Xie Yan, Chief Technology Officer of Li Auto, explained the design philosophy behind the Mahe 100 during the call: "For the same silicon area, the Mahe 100 provides higher effective computing power, thereby offering greater design space for VLA algorithms. For instance, we can deploy VLA models with parameter sizes six times larger and computational requirements ten times greater than the previous generation, while still achieving higher frame rates and faster inference speeds."
Crucially, this represents a full-stack in-house development. The self-reliance across chips, compilers, operating systems, and base models equips Li Auto with the capability to customize AI computing power from the ground up, building a technological barrier difficult for competitors to replicate.
Xie Yan used an analogy to illustrate the strategic intent behind this layout: "When we initiated chip development in 2022, we predicted that starting from 2025, the industry would enter an era where competitive advantage would be gained through vertically integrated design of models, chips, and operating systems. We believe the gap in chips, operating systems, and the software-hardware ecosystem will ultimately resemble the difference between Apple and Android."
The deep integration of the Mahe 100 with the Xinghuan OS and steer-by-wire systems is already yielding tangible product improvements. According to Xie Yan, this synergy enables more efficient communication between autonomous driving computation, pre- and post-processing, and execution control, significantly reducing the overall latency from sensor photon input to vehicle actuator output.
Beyond performance gains, Xie Yan also revealed that the Mahe 100 offers substantial cost advantages, with the Bill of Materials cost per chip being significantly lower than purchasing external solutions.
The value of this technological foundation is already extending beyond vehicles. The Li Auto AI glasses, Livis, launched in December 2025, serve as the first validation of this AI technology migrating to consumer terminals. It demonstrates that the technological base built for embodied intelligent vehicles possesses cross-product applicability.
"We expect R&D expenses to remain at around 12 billion yuan this year, with AI-related R&D investment accounting for approximately half," stated Li Tie.
Automaker vs. AI Company: Why is Li Auto charting a third path? In China's new energy vehicle arena, players are being rapidly categorized. On one side are traditional automakers and new entrants, most struggling near the breakeven point, with few achieving sustained profitability. On the other side are AI tech companies, burning through raised capital in pursuit of AGI, with commercial viability still distant.
Li Auto's 2025 financial report presents a hybrid "middle state" that is difficult to classify, simultaneously embodying the stability of manufacturing and the growth potential of a tech company.
Compared to traditional automakers, where price wars erode profits and R&D costs keep rising, many firms are mired in losses. Even multinational giants like Volkswagen and Toyota face challenges like profit declines, cost-cutting, and layoffs.
Li Auto managed to maintain its profitability baseline during its 2025 strategic investment phase—essentially, its ability to generate cash flow remained intact. More importantly, this profitability does not come at the expense of future growth. This model of "high investment alongside profitability" is rare in the automotive industry.
The contrast with AI companies is stark. Large model startups often raise hundreds of millions of dollars but face unclear commercialization paths, with the vast majority continuously burning cash. Even leading AI companies grapple with the dilemma of "massive investments, distant returns."
Li Auto's distinction lies in its clear monetization path—selling cars. According to the company, production bottlenecks for the Li i6 have been overcome, with monthly capacity reaching 20,000 units in March, and the proportion of pure electric models is steadily increasing. A dual-drive strategy combining extended-range and pure electric vehicles is now firmly established.
Beyond automobiles, Li Auto's expansion into other areas avoids a "reckless spending" approach. For new domains like AI glasses and humanoid robots, Li Xiang's strategy is clear: "In terms of actual commercial and product deployment, we will proceed cautiously, continuously exploring. When venturing into new fields, we will adopt a startup approach, including for AI glasses and robotics projects. We aim to avoid the lavish spending typical of large corporations, instead opting for a genuine startup team approach to incubate products."
Essentially, Li Auto's path combines the cash-generating capability of its automotive business with long-term investment in AI technology, creating a self-reinforcing cycle: car sales generate cash flow, cash flow funds R&D, R&D builds technological barriers, which in turn enhance product competitiveness, driving further sales growth.
In capital markets, if viewed as an automotive manufacturer, Li Auto is evaluated on sales volume, profit margins, and per-vehicle profit. If viewed as an AI tech company, the market is willing to pay a premium for its technological moat, data assets, and future potential. Li Auto possesses both the manufacturing "stability" of three consecutive profitable years and the AI "growth option" backed by hundred-billion-yuan cash reserves.
How is AI helping Li Auto restructure "people" and "organization"? Li Xiang stated: "For Li Auto, AI primarily involves two aspects: creating AI and using AI. Creating AI is about giving life to products; using AI is about enhancing work efficiency."
Reviewing Li Auto's organizational changes over the past year reveals a clear evolutionary path: shifting from traditional hierarchical management to a platform-based, enabling organization.
In January, Li Auto completed an internal reorganization that some insiders found "somewhat abstract." R&D teams were no longer divided by traditional software/hardware functions but were regrouped based on a new logic—what Li Xiang termed "organizing based on the way we create digital humans and silicon-based lifeforms."
In this restructuring, the R&D system was reconfigured into four major segments: visceral systems, brain systems, software ontology, and hardware ontology. The underlying logic of this organizational change, in Li Xiang's words, is: "Building products is like building a person."
On the R&D front, by restructuring the architecture to encompass chips (analogous to the heart), datasets (lungs), and operating systems (nerves), Li Auto broke down the departmental silos between software and hardware, enabling true collaboration among experts from different fields.
The effects of this adjustment became apparent quickly. "Recently, everyone has noticed that work efficiency has increased. The iteration cycle for intelligent driving model training has shifted from once every two weeks to once per day," said Li Xiang.
Behind this new organizational logic lies Li Auto's rethinking of corporate structure in the AI era. Li Xiang summarized it as: "AI is reversing the decline in information flow and decision-making efficiency that typically accompanies company growth. By introducing Agent-assisted collaboration, our iteration and evolution speed is regaining the high efficiency and agility of a startup. This change has begun to manifest in our daily operations this year. It can be said that AI not only brings innovation in production tools but also fosters a more dynamic organizational form."
When corporate scale becomes an enemy of efficiency, AI and organizational transformation together constitute Li Auto's solution. The former provides empowerment at the tool level, the latter represents restructuring at the mechanism level. Combined, they allow this company with tens of thousands of employees to maintain the evolutionary pace of a startup.
How does the L9 Livis define the "flagship of embodied intelligence"? In the second quarter of this year, the all-new Li L9 and the Li L9 Livis will launch. Internally, the Li L9 Livis is defined as "a flagship SUV completely redefined based on the logic of embodied intelligence, bringing a generational leap in user experience."
Behind this somewhat complex positioning lies the first comprehensive realization of Li Auto's hundreds of billions of yuan in R&D investment over the past three years. From perception and decision-making to execution, the all-new Li L9 Livis undergoes a technological reconstruction across three dimensions.
"Whether for autonomous driving or embodied intelligence, the fundamental challenge is training and learning by watching videos, and controlling the vehicle based on video input. Essentially, this technical system doesn't truly understand the physical world; it's learning through imitation," Li Xiang said.
The all-new Li L9 Livis aims to break this limitation. It will move beyond the "2D camera + LiDAR" approach, upgrading to 3D ViT to perceive and understand the physical world truly like a human, not merely through video analysis.
Li Xiang explained the significance of this breakthrough: "This requires joint design and breakthroughs from the video encoder to the entire model development, and the powerful computing provided for the encoder and model—that is, at the chip level. I believe a major global breakthrough will occur this year, whether in autonomous driving or embodied intelligence."
Accompanying this breakthrough, the VLA model will evolve into a true "VLA"—where the "Language" component will no longer be solely for object recognition but will genuinely understand and reason about the physical world like a human.
The computational headroom of the Mahe 100 in the Li L9 Livis also reserves space for future evolution. Xie Yan said: "Higher local computing power also gives us the opportunity to provide richer intelligent capabilities beyond autonomous driving, making the car closer to a robot. These capabilities will debut on the L9 and continue to expand in the future."
If perception and the brain determine how clearly the vehicle "sees" and how quickly it "thinks," then the body determines how accurately it "acts."
The all-new Li L9 Livis features the world's first "complete" full steer-by-wire chassis, including steer-by-wire, four-wheel steering, and the world's first fully electric mechanical brake (EMB). According to the company, it is the first vehicle under 600,000 yuan to be equipped with an 800V fully active suspension, delivering top-tier handling and driving pleasure.
Li Xiang explained the significance of this system: "This makes the vehicle's response speed and safety far exceed those of traditional cars. The vehicle is no longer controlled via transmissions between traditional MCUs; instead, the model can output commands directly to the control system."
In traditional vehicles, the driver's intent is transmitted mechanically or electronically to actuators; the car is a passive tool. In the all-new L9 Livis, the large model can directly control all actuators via the MCP; the vehicle is no longer just a tool but an intelligent agent capable of active response.
Li Xiang's positioning for the Li L9 Livis is clear: "The all-new Li L9 Livis has completed a technological reconstruction across the three dimensions of perception, decision-making, and execution." This is not merely a facelifted model but Li Auto's first complete creation since transitioning from an automaker to an embodied intelligence enterprise.
Comments