行业研究丨深度报告丨汽车与汽车零部件 [Table_Title] 智驾平权系列六：AI 智能涌现新阶段，智驾 VLA 与世界模型之争 %% %% %% %% research.95579.com 1 丨证券研究报告丨 报告要点 [Table_Summary] 通用人工智能大模型迎来跨越式发展，模型规模、训练范式与推理能力持续突破，为各类 AI 应 用构建了坚实的底层技术底座。智能驾驶本质是"物理 AI"的应用，因此注定了往大模型方面 演进。通用大模型能力涌现，赋能智驾模型基座，智驾模型架构持续进阶，逐步进入 VLA 和世 界模型的物理 AI 时代，迈向智能涌现新阶段。 分析师及联系人 [Table_Author] 高伊楠 张扬 SAC：S0490517060001 SAC：S0490524030004 SFC：BUW101 请阅读最后评级说明和重要声明 2 / 30 %% %% %% %% research.95579.com 2 汽车与汽车零部件 cjzqdt11111 [Table_Title 智驾平权系列六： 2] AI 智能涌现新阶段，智驾 VLA 与世界模型之争 [Table_Summary2] 引 ...

Core Viewpoint - The company aims to establish itself as a foundational platform in the humanoid robotics industry, likening its role to that of "selling shovels" in a gold rush, focusing on service rather than immediate commercialization [5][42]. Company Overview - Beijing Humanoid Robot Innovation Center was established at the end of 2023 with an initial investment of 460 million yuan from various stakeholders [2][39]. - The CEO, Xiong Youjun, emphasizes a work culture that allows for mistakes but does not tolerate stagnation, aiming for continuous exploration in technology [4][41]. Technological Development - The company is navigating a chaotic phase in the industry, debating whether to focus on VLA models or world models, and whether to rely more on simulation data or real data [4][41]. - A significant goal is to release the first generation of the "Embodied Tiangong" robot in 2024, followed by the "Wisdom Open Object" platform in 2025, and to enhance capabilities with the "Embodied Tiangong 3.0" in 2026 [4][41][42]. Market Positioning - The company positions itself within a projected trillion-dollar market for embodied intelligence, asserting that current technological breakthroughs are merely stepping stones and do not yet create a competitive moat [5][42]. - The company aims to create an Android-like ecosystem in the robotics field, focusing on open-source technology to foster industry growth [5][23]. Funding and Investment - The company recently completed its first market-oriented financing round, raising 700 million yuan, which was initiated in late 2025 due to favorable market conditions [7][45]. - The CEO notes that there are no performance guarantees set by investors, reflecting confidence in the company's technological advancements [9][48]. Industry Context - Comparatively, domestic humanoid robot companies are valued at over 10 billion yuan, while leading companies like Ubiquity are valued below 100 billion yuan, contrasting with U.S. companies valued in the hundreds of billions [12][47]. - The CEO argues that the perception of a bubble in the industry stems from a lack of understanding of the future market potential [12][48]. Open Source Strategy - The company has opened its platforms, "Embodied Tiangong" and "Wisdom Open Object," significantly lowering barriers for other companies to enter the humanoid robotics space [19][57]. - The open-source approach is seen as a way to accelerate industry maturity and does not conflict with the company's commercial goals [23][60]. Challenges and Solutions - The company faces numerous challenges, including technological, product, talent, and funding issues, and is focused on building a robust platform to address these gaps [15][69]. - A mid-test verification platform has been established, achieving a production capacity of 5,000 units annually, which aims to bridge the gap between laboratory prototypes and industrial applications [33][70].

Core Viewpoint - The humanoid robot industry is shifting focus from entertainment to practical applications, with a significant increase in production expected in the coming years, driven by the need for robots to demonstrate real-world value rather than just perform on stage [1][3][4]. Industry Trends - The humanoid robot market is projected to see a shipment increase of over 650% in 2025, reaching approximately 18,000 units, and is expected to rise to 62,500 units in 2026 [3]. - The industry is moving past a phase of "wild growth," where mere performance was sufficient for sales, to a more mature phase where practical applications and real-world scenarios are essential for success [4][14]. Technology Development - Three main technical routes are emerging in the humanoid robot sector: VLA (Visual Language Action) model, world model, and layered decision-making with hardware-software collaboration, each with distinct advantages and challenges [7][9][10]. - The VLA model aims for general intelligence, relying on vast data for training, while the world model focuses on simulating physical environments to predict actions [9][10]. - The layered decision-making approach breaks down complex tasks into manageable components, enhancing reliability and efficiency in real-world applications [9][10]. Market Demand - There is a growing demand for robots that can operate in specific scenarios, such as factories and logistics, where they can perform tasks like assembly and packaging, thus providing tangible economic value [13][14]. - Users are increasingly looking for robots that can reduce production costs and alleviate humans from repetitive or hazardous tasks, indicating a shift towards practical applications [13][14]. Investment Focus - Investors are prioritizing companies that can demonstrate viable application scenarios and tangible products, moving away from those that lack a clear path to market [4][5]. - The consensus among investors is that companies must integrate their technology with real-world applications to avoid being sidelined in a competitive landscape [4][5]. Future Outlook - The next 3 to 5 years are critical for the practical deployment of humanoid robots, with expectations that they will increasingly complement human labor rather than replace it [13][19]. - The industry is expected to see rapid technological advancements, with a focus on improving the stability and reliability of robots in various operational environments [18][19].

Core Viewpoint - The CEO of Li Auto, Li Xiang, presented an ambitious vision focused on AI during a company-wide meeting, which did not resonate well with employees who are more concerned with immediate sales and performance metrics [1][2][4]. Group 1: AI Strategy and Investment - Li Auto is making a significant bet on AI, with a timeline indicating that 2026 is the last chance for companies to establish themselves as leaders in AI [2][3]. - The company plans to invest over 12 billion RMB in R&D in 2024, with a substantial portion allocated to AI technologies, including foundational models and inference chips [2][3]. - The VLA (Vision-Language-Action) model aims to unify various intelligences to create a vehicle that can understand, think, and act like a robot [3][4]. Group 2: Employee Concerns and Company Performance - Li Auto's third-quarter revenue was 27.4 billion RMB, a 36.2% year-over-year decline, with vehicle deliveries down 39.0% [6]. - The company reported a net loss of 624 million RMB, marking a significant shift from profitability in previous quarters [6]. - Employees expressed dissatisfaction regarding year-end bonuses and high work pressure, indicating a disconnect between the CEO's vision and their immediate concerns [6][7]. Group 3: Organizational Changes and Challenges - The restructuring of the R&D organization into three teams (base model, software, and hardware) reflects a shift in focus from traditional automotive production to AI technology [11][12]. - Employees are experiencing anxiety and confusion due to the new organizational structure, which may divert attention from current automotive projects to long-term AI goals [12][13]. - Sales and marketing teams are particularly concerned about how the AI strategy will translate into actionable products and market strategies to meet their KPIs [13][14]. Group 4: Leadership and Communication - Li Xiang's inability to effectively communicate the AI vision in a way that resonates with employees highlights a leadership challenge [14][15]. - Employees are looking for concrete plans and strategies that connect the ambitious AI goals with their day-to-day responsibilities and performance metrics [15].

Core Insights - AI technology is experiencing three major breakthroughs: the evolution from chatbots to intelligent agents, the lowering of entry barriers through open-source models, and the understanding of the physical world through physical AI [1] - Physical AI is recognized as the next wave of AI development, showcasing its potential in understanding complex scientific principles [1] Group 1: VLA and World Models - The VLA (Vision-Language-Action) model and world models are emerging as a dual-model paradigm to address the data scarcity and safety issues in physical AI [2][3] - World models can generate infinite simulation data at a low cost, allowing VLA to learn from various scenarios without the risks associated with real-world data collection [3] - The integration of VLA and world models is seen as the optimal solution for enhancing embodied intelligence in physical AI [3] Group 2: Development Stages - The development of VLA and world models can be structured into four stages: cold start, interface alignment, training in simulated environments, and real-world transfer and calibration [4][5] - The cold start phase involves training a basic VLA model using existing robot datasets while the world model is pre-trained on vast amounts of video data [4] - The interface alignment phase focuses on mapping VLA's action outputs to the world model's input conditions to simulate the resulting scenarios [4] - In the training phase, VLA operates within the simulated environments generated by the world model, allowing for extensive reinforcement learning without physical wear on robotic components [4] Group 3: Addressing Challenges - Generative models often produce inconsistent outputs, leading to incorrect physical assumptions; introducing 3D geometry and material constraints can mitigate this issue [6] - A reward model can be implemented to evaluate the success of tasks in generated scenarios, providing feedback to the VLA [6] - The speed of world model predictions is crucial for training efficiency; techniques like latent consistency models can enhance prediction speed by focusing on feature changes rather than pixel-level details [6] Group 4: Data Sharing and Best Practices - The architecture of world models is evolving, but the necessity for real and synthetic data remains constant [7] - Sharing visual encoders between VLA and world models can optimize memory usage and ensure synchronized understanding of the environment [7] - Generating counterfactual data allows VLA to learn from hypothetical failure scenarios, improving robustness and reducing real-world testing costs [7] Group 5: Towards General Artificial Intelligence - The future of world models involves generating interactive 4D environments, enabling VLA to train in dynamic settings rather than static ones [8] - The integration of fast and slow systems within AI, where VLA handles real-time responses and world models manage long-term planning, is a key goal for advancements in autonomous systems [8] - Ultimately, VLA and world models may converge into a unified model capable of predicting both actions and future states, aligning with the vision of AI understanding physical laws [9][10]

Core Insights - The article discusses the significant reshuffling in the Chinese advanced driving assistance market, particularly focusing on urban NOA (Navigation on Autopilot) as it approaches a critical penetration rate of over 10% by 2025, with expectations to reach 22% by 2026, leading to a market scale targeting millions of units [6][14]. Group 1: Market Dynamics - A number of players, including Maimo Zhixing and Dazhuo Intelligent, have exited the market, while Huawei, Yuanrong Qixing, and Momenta have emerged as the dominant trio, collectively holding 99% of the urban NOA market share from January to October 2025 [2][8]. - In October 2025, Yuanrong Qixing achieved the highest urban NOA installation rate, with a significant growth rate of 2.7 times compared to the average monthly installation from January to October 2025 [8][10]. Group 2: Competitive Strategies - The three leading companies have adopted different market strategies: Huawei focuses on high-end models, Momenta has a broad brand coverage, and Yuanrong Qixing targets mass-market models for data accumulation [9][10]. - Yuanrong Qixing's strategy emphasizes deep collaboration with a few key models to create "explosive" sales, resulting in significant sales increases for models like the Blue Mountain and Galaxy M9 [9][10]. Group 3: Future Projections - The competition is expected to intensify as the industry moves towards "million-unit production," with a projected annual output of 5 million high-level autonomous vehicles by 2026 [14][15]. - The article highlights the importance of data efficiency and closed-loop systems as critical competitive advantages in the evolving landscape of intelligent driving [14][15]. Group 4: Investment Trends - The investment landscape for autonomous driving has seen a significant rebound in 2025, with nearly 60 billion yuan raised, indicating a strong focus on leading and commercially viable projects [14][15].

Core Viewpoint - The article discusses recent organizational changes at Li Auto, focusing on the shift towards embodied intelligence and the integration of the VLA model for autonomous driving development [2][6]. Group 1: Organizational Changes - Li Auto is reallocating resources towards embodied intelligence as competition in automotive intelligence enters a "modeling" phase [2]. - Key personnel changes include Zhan Kun taking over the VLA integration and development work, reporting directly to the CTO, while Chen Wei, responsible for the LLM direction, is leaving the company [2][5]. - The internal restructuring reflects a preference for promoting from within, indicating strong confidence in the existing technical team [6]. Group 2: Technological Developments - Significant upgrades to the VLA model have been made in recent months, with high internal confidence in version 8.2 [6]. - The integration of robotics and autonomous driving is being coordinated under a larger embodied paradigm, with Shua Yifan now responsible for the autonomous driving product [4][5]. - The development of a new generation closed-loop system is being emphasized, combining base models, cloud, and vehicle-end technologies [8]. Group 3: Industry Trends - The trend towards integrated hardware and software solutions is expected to be a major industry focus by 2026 [10]. - The success of Horizon Robotics' HSD is noted as a contributing factor to the recent organizational adjustments at Li Auto [8].

Core Viewpoint - Xiaopeng Motors is increasingly positioning itself as an AI company, focusing on software and AI capabilities rather than just hardware in its new vehicle releases [6][8]. Group 1: New Vehicle Launches - On January 8, 2026, Xiaopeng launched four new models: P7+, G7 extended range version, 2026 G6, and 2026 G9, all featuring self-developed Turing AI chips [6][7]. - The new models will utilize Xiaopeng's second-generation VLA model, enabling basic Level 4 (L4) assisted driving capabilities [6][7]. - The Turing chip in the MAX version has an effective computing power of 750 TOPS, while the Ultra SE version uses two Turing chips for intelligent driving, and the Ultra version uses three chips [7]. Group 2: Strategic Focus on AI - Xiaopeng's CEO emphasizes that the value of AI will surpass traditional performance upgrades, with software's contribution to vehicle value expected to rise from 10% to 50% over the next decade [8][14]. - The company is committed to developing both the models and the software for chips, believing that the best AI companies will choose to customize their chips [17]. Group 3: Market Position and Sales Strategy - Xiaopeng has sold over 400,000 vehicles in the past year, showing significant growth, and plans to expand its product lineup with multiple new models in 2026 [10][24]. - The company aims to enhance its global supply chain and channel management, with plans to expand into more countries and regions [9][10]. Group 4: Future Outlook and Challenges - The CEO predicts that 2026 will be a pivotal year for the automotive industry, with a significant shift towards AI integration in vehicles [14][20]. - Xiaopeng is aware of the competitive landscape in the Chinese automotive market and acknowledges the uncertainties associated with its AI strategy [7][8].

Core Insights - The autonomous driving industry is approaching a state of transparency, with multiple automakers and tech companies targeting the sub-100,000 yuan market for advanced driving features [1][4][9] - The competition has intensified, with companies like BYD, Chery, and Geely launching models equipped with high-level autonomous driving capabilities at lower price points [5][6][7] - The industry is witnessing a convergence of algorithmic approaches, with discussions around VLA and world model architectures, as companies aim to enhance the usability of autonomous driving features [3][20][21] Group 1 - The introduction of advanced driving features in vehicles priced below 100,000 yuan has become a key focus for both automakers and autonomous driving technology providers [1][2][4] - Companies like Horizon Robotics and Huawei are planning to expand their advanced driving capabilities into lower price segments, aiming to break the perceived barriers of high costs associated with such technologies [8][11] - The competition is not only about vehicle pricing but also about the technological advancements in algorithms and hardware, with a significant push towards reducing costs in chips and sensors [10][11][13] Group 2 - The industry is transitioning from a phase of diverse approaches to a more consolidated focus on effective algorithmic solutions, with VLA and world model architectures being at the forefront of discussions [3][20][21] - The recent approval of L3-level conditional autonomous driving vehicles by the Ministry of Industry and Information Technology marks a significant milestone, indicating a potential "breakthrough year" for L3 technology [4][28] - Companies are also exploring the integration of autonomous driving technologies into broader applications, such as logistics and robotics, indicating a shift towards a more comprehensive ecosystem [30][33] Group 3 - The competitive landscape is evolving into a "red ocean" as more players enter the sub-100,000 yuan market, necessitating a focus on efficiency and differentiation [9][13] - The advancements in chip technology and the reduction in costs for essential components like lidar are enabling more companies to offer competitive autonomous driving solutions [10][11] - The push towards L3 and L4 capabilities is becoming a race among automakers and tech firms, with significant investments and developments expected in the coming years [29][30]

Core Insights - The automotive industry's core competitiveness is shifting from traditional mechanical performance to smart technology, safety, and integration with energy networks [3] - Eight key component areas have emerged as focal points for change in the automotive supply chain by 2025 [3] Group 1: AI and Smart Technology - AI large models, including VLA and VLM, are reshaping the perception, decision-making, and interaction systems in smart vehicles [4] - Companies like Li Auto and XPeng are actively developing and deploying VLA-based autonomous driving systems, with plans for mass production by 2026 [4] - The competition in AI models is intensifying, with a focus on the underlying support systems like computing power and data [4] Group 2: Vehicle-to-Grid (V2G) Interaction - V2G is becoming a hot topic as electric vehicles can act as distributed energy storage units within new energy systems [5] - Government policies are driving the adoption of V2G, with pilot projects and plans to expand the scope of V2G applications by 2027 [5][6] - Companies like GAC Group are implementing V2G functionalities in their models and developing charging infrastructure to support this transition [6] Group 3: Battery Safety Standards - The new national standard for electric vehicle batteries, effective July 2026, emphasizes safety by requiring batteries to be "non-flammable and non-explosive" [7] - The updated standards will compel battery manufacturers to innovate in materials, design, and production processes to meet stricter safety requirements [7] - Leading battery companies like BYD are already adapting to these new standards, which will enhance safety and consumer trust in electric vehicles [7] Group 4: Door Handle Innovations - Electric hidden door handles are becoming a focal point due to safety concerns arising from their failure in collision scenarios [8][9] - New regulations are being proposed to ensure that all door handles, including electronic ones, have a mechanical release function for emergency situations [9] Group 5: Solid-State Batteries - Solid-state batteries are gaining traction due to their advantages in energy density and safety, with several companies planning to launch new products or production lines [10] - The development of solid-state batteries is seen as a key competitive factor for companies in the next generation of electric vehicles [10][11] Group 6: Human-Car-Home Ecosystem - The "Human-Car-Home" ecosystem is emerging, integrating automotive, home, and personal devices into a cohesive smart system [12] - Companies like Haier and Midea are collaborating with automotive brands to create interconnected systems that enhance user experience [12][13] Group 7: Humanoid Robots - The automotive industry is increasingly intersecting with humanoid robotics, with companies exploring the integration of robotic technology into manufacturing processes [14][15] - The demand for precision and adaptability in manufacturing is driving the development of humanoid robots tailored for automotive applications [14] Group 8: Zero-Gravity Seats - Zero-gravity seats are becoming a key feature in mid to high-end vehicles, enhancing passenger comfort and experience [16] - The lack of standardized regulations for these seats poses challenges, particularly regarding safety during vehicle operation and collisions [16]