观点一：智驾平权 2.0，支撑智驾赛道全年高景气度。 展望 2026 年整车行业面临补贴退坡、原材料涨价等多重压力，市场担心主机厂对智能驾驶功能是否会减配。我们研 究发现 26 年智驾平权趋势非但不会放缓，而是会进入 2.0 阶段，城市 NOA（领航辅助驾驶功能）开始走向千家万户， 背后的原因在于：1）过去 2-3 年智驾景气度的本质来源是新能源车内卷背景下的主机厂增配，这一强大的供给端驱 动力在 2026 年依然存在并且增强。2）小鹏 Mona 已经证明 10-20w 低价格带消费者对优质智驾功能同样具有需求。在 供给和需求双重驱动力下，我们测算 2026 年城市 NOA 硬件配置渗透率有望从 2025 年 16%提升至 25%，全年搭载城市 NOA 功能硬件配置销量有望达到 545 万，同比增速超过 50%。 观点二：L2 进入强监管政策周期，L3/L4 法规体系逐步建立。 1）L2 强监管政策周期下检测类机构充分受益：L2 强标是强监管政策落地最重要的执行依据，目前已进入报批阶段。 此次我国 L2 强覆盖范围广、所涉及检测车型多；同时由于测试严格程度强、检测项目多，预计单次检测价值量较高； 因此我们预计 ...

「如果我想去远远甩开当前的对手，这一代智驾我们应该怎么做？」 时间回到两年前，在美国硅谷小鹏的办公室，何小鹏见到刘先明时，几乎只问了这一个问题。 这个问题非常关键。 刘先明的回答是拆掉语言的 VLA。在一个小时的交流里，刘先明觉得这不像是一场面试，也不需要说服老板接受新的技术方案，而是两个人已经开始商 量去做这件事的具体步骤。 刘先明从何小鹏的办公室出来之后，只有一个想法：「这是一个我必须来的地方。」 而刘先明，已经是小鹏自研智驾十年来的第四任大将。 吴新宙完成了小鹏智驾的「代际领先」；李力耘完成了小鹏从规则时代到端到端的转型。但正在这个阶段，很多玩家靠着端到端火速完成超车。 显然，小鹏没有预想到其他人跟进速度如此之快。 就在小鹏下线第 100 万辆车的时候。外界有一种声音：小鹏销量从 ICU 到了 KTV，但智驾却被「理想、华为们」围追堵截，甚至陷入「吃老本」的质 疑。 小鹏智驾的起伏很像中国新势力智驾突围史的缩影，均围绕着体系、量产、算法三种能力比拼。 但同时它又具备特殊性。 8 年间小鹏三次换帅，牵引出了另一层深意： 真正的智驾战争，不是眼前的技术代差，而是对抗组织惯性。 时代变了、架构在变、主导的人， ...

Core Insights - 2026 marks the beginning of a golden decade for autonomous driving, driven by technological advancements and the adoption of "end-to-end" architecture [1] - The introduction of VLA (Vision-Language-Action) models and world models will enable autonomous systems to learn from vast amounts of real and generated data, achieving safety levels over 10 times that of human drivers [1] - The market is expected to see widespread adoption of urban NOA (Navigation On Autopilot) features in vehicles priced around 100,000 yuan by 2026, a significant improvement compared to the capabilities of many L4 autonomous vehicles just two years prior [1] Market Trends - The penetration rate of passenger cars equipped with L2-level driving assistance features reached 64% in the first three quarters of 2025, with a year-on-year growth of 21.2% [4] - By November 2025, the cumulative sales of passenger cars with urban NOA features reached 3.129 million, accounting for 15.1% of insured passenger vehicles [4] - The majority of urban NOA-equipped vehicles are expected to be priced below 150,000 yuan, with over 68.9% of mainstream models under 300,000 yuan featuring this technology [4] Competitive Landscape - The autonomous driving sector is becoming increasingly competitive, with companies transitioning from traditional technologies like "LiDAR + high-definition maps" to AI-driven solutions [7] - The industry is witnessing a divergence in development directions, particularly between VLA and world models, with a consensus forming around the evolution of autonomous driving technology [7] - The successful differentiation in the market will depend on the ability to translate technology into tangible user experiences rather than just technical concepts [7] Strategic Insights - The million-unit deployment of autonomous driving systems is seen as a critical milestone, with few suppliers able to achieve this level [6] - The compatibility of QCraft's solutions with both new energy vehicles and traditional fuel vehicles provides a strategic advantage for global expansion [5] - The entry of Tesla's FSD into the Chinese market is viewed as a positive development that will expand the overall market rather than just intensifying competition [3][8]

Core Insights - Tesla has identified three core challenges in autonomous driving during its presentation at ICCV2025, which have been widely discussed in both academia and industry [3][6][7] - The event features discussions on solutions to these challenges, including insights from researchers at the University of Hong Kong [3][11] Group 1: Core Challenges - The three main challenges in Tesla's end-to-end architecture for autonomous driving are dimensionality disaster, interpretability and safety guarantees, and closed-loop evaluation [6][7] - Solutions proposed include UniLION, DrivePI, and GenieDrive, which aim to address these challenges [6][13] Group 2: Technical Insights - The presentation includes a detailed explanation of Tesla's end-to-end technology evolution and FSD v14 [6][13] - The discussion will also explore the concept of a general artificial intelligence that can understand and interact with the physical world [6][13] Group 3: Additional Content - The event will provide deeper insights into the technical details, Q&A, and previously unpublished content related to autonomous driving [14] - There will be discussions on the divergence between academic research and mass production, as well as ongoing technical debates in the industry [14]

Core Viewpoint - The article discusses the transformative impact of high-fidelity virtual datasets, specifically SimData, on the development of autonomous driving algorithms, emphasizing the need for high-quality data to overcome the limitations of traditional real-world testing [2][4][29]. Group 1: SimData Dataset Overview - SimData addresses the high demand for quality data in autonomous driving, highlighting the challenges of traditional real-world testing, including high operational costs, subjective bias in manual labeling, and legal constraints [4][5]. - The dataset includes 880 instances, 215,472 keyframe data, and 64,190 annotations, showcasing its extensive scale and diversity [6][7]. - SimData covers critical operational design domains (ODD) such as highways, urban canyons, and parking lots, with a focus on hard-to-capture scenarios like construction zones and extreme lighting conditions [7]. Group 2: Automation Toolchain: aiSim2nuScenes - The aiSim2nuScenes toolchain facilitates the efficient conversion of virtual simulation data into high-value data assets for algorithms, creating a standardized bridge between virtual environments and algorithm applications [11][12]. - It automates the generation of multi-modal sensor data and ensures strict temporal alignment of sensor data, achieving microsecond-level synchronization [13][15]. - The toolchain supports the nuScenes standard format, enhancing compatibility and reducing the engineering team's migration costs [13]. Group 3: Algorithm Empirical Evidence - Training experiments on the pure virtual dataset demonstrated rapid convergence, achieving a mean Average Precision (mAP) of 0.446 and a nuScenes Detection Score (NDS) of 0.428 within 30 epochs [19]. - The consistency between models trained on SimData and those trained on real-world data was validated through AP correlation analysis and attention heatmap analysis, indicating high fidelity in feature extraction [20][22]. - Domain adaptation experiments showed that combining real-world data with virtual data significantly improved model performance across various categories, proving that virtual data complements rather than replaces real data [23][26]. Group 4: Conclusion and Future Outlook - The article concludes that high-fidelity virtual data is essential for training algorithms capable of generalizing to real-world scenarios, emphasizing the importance of accurate modeling of physical processes [29]. - As the demand for high-quality synthetic data grows, the integration of virtual data into the training process is positioned as a key strategy for enhancing the robustness and performance of autonomous driving systems [29].

Core Insights - Tesla has resumed sharing updates on its autonomous driving algorithms after a two-year hiatus, presenting at the ICCV conference instead of its previous AI Day events [1] - The company is facing challenges with its end-to-end architecture for autonomous driving, particularly regarding the "black box" nature of the model and the quality of training data [3][7] Group 1: Technical Developments - Tesla's end-to-end system must address the mapping from high-dimensional to low-dimensional outputs, which is complex due to the nature of the data [5][7] - The company has implemented optimizations in its architecture, including the introduction of OCC occupancy networks and 3D Gaussian features to enhance decision-making [3][8] - Tesla has developed a "neural world simulator" that serves as both a training and validation environment for its algorithms, allowing for extensive testing and refinement [12][15] Group 2: Competitive Landscape - Other companies in the industry, such as Xpeng and Li Auto, have also adopted similar models, indicating a shift in the competitive dynamics of the autonomous driving sector [4][11] - Tesla's previous position as a leader in autonomous driving technology is being challenged, with other players no longer closely following its developments [18] Group 3: Market Reception and Challenges - The subscription rate for Tesla's Full Self-Driving (FSD) feature is low, with only about 12% of users opting for it, raising concerns about the technology's acceptance [4][24] - Despite price adjustments for FSD, consumer interest has waned, with many potential buyers citing concerns over the technology's maturity and reliability [24][25] - Recent investigations into Tesla's FSD have highlighted safety issues, further complicating the company's efforts to promote its autonomous driving capabilities [24][25]

Core Insights - The automotive industry is shifting its focus from electrification to intelligence, with chips, radars, and systems becoming critical for success [3] - Horizon Robotics aims to empower smart vehicles and robots, emphasizing safety through a comprehensive safety development system certified by international standards [3][5] - The company follows a progressive technical path similar to Tesla, aiming to achieve L4 and L5 capabilities while focusing on an "end-to-end" architecture for human-like driving experiences [5][9] Group 1: Strategic Positioning - Horizon Robotics is one of the few domestic companies achieving large-scale production in the intelligent driving sector, positioning itself as an industry pioneer [3] - The company has developed a full-domain safety development system that integrates hardware and software, making it one of the most complete safety systems in the industry [3] - Horizon emphasizes the importance of product strength over marketing gimmicks, aiming to make intelligent driving a standard feature in vehicles [7] Group 2: Technical Path and Market Outlook - The company predicts that true L3 capabilities will be based on L4 capabilities, with expectations of achieving near "100,000 kilometers without takeover" by 2028, contingent on extensive real-world data and insurance models [5] - Horizon has empowered over 25 vehicle models for international markets, collaborating with various Tier-1 suppliers and foreign automakers [7] - The intelligent driving market is seen as a certainty, with Horizon's shipment of millions of chips reflecting genuine market demand [9] Group 3: Future Vision and Industry Dynamics - Horizon Robotics focuses on creating real value for users, rejecting short-term trends in favor of long-term strategies [12] - The company believes the intelligent driving industry is nearing a consolidation phase, with only two to three leading companies expected to emerge in the next three to five years [9] - The ultimate goal is to shift competition from internal struggles to collaborative value creation for users and industry expansion [12]

Core Viewpoint - The field of embodied intelligence is experiencing unprecedented attention, yet key issues remain unresolved, including data scarcity and differing technical approaches [1][2][3] Group 1: Data and Technical Approaches - The industry is divided into two factions: the "real machine" faction, which relies on real-world data collection, and the "synthetic" faction, which believes in the feasibility of synthetic data for model training [5][12] - Galaxy General, representing the synthetic faction, argues that achieving generalization in embodied intelligence models requires trillions of data points, which is unsustainable through real-world data alone [8][9] - The "real machine" faction challenges the notion that real-world data is prohibitively expensive, suggesting that with sufficient investment, data collection can be scaled effectively [12][14] Group 2: Model Architecture - Discussions around the architecture of embodied intelligence models highlight a divide between end-to-end and layered approaches, with some experts advocating for a unified model while others support a hierarchical structure [15][19] - The layered architecture is seen as more aligned with biological evolution, while the end-to-end approach is criticized for potential error amplification [19][20] - The debate extends to the relevance of VLA (Vision-Language Alignment) versus world models, with some experts arguing that VLA is currently more promising due to its data efficiency [21][22] Group 3: Industry Trends and Infrastructure - The scaling law in embodied intelligence is beginning to emerge, indicating that expanding model and data scales could be effective [24] - The industry is witnessing an acceleration in the deployment of embodied intelligence technologies, with various companies sharing their experiences in human-robot interaction and industrial applications [24][29] - Cloud service providers, particularly Alibaba Cloud, are emphasized as crucial players in supporting the infrastructure needs of embodied intelligence companies, especially as they transition to mass production [29][31] Group 4: Alibaba Cloud's Role - Alibaba Cloud has been preparing for the exponential growth in data and computational needs associated with embodied intelligence, having developed capabilities to handle large-scale data processing and model training [33][35] - The company offers a comprehensive suite of cloud-based solutions to support both real and synthetic data production, enhancing efficiency and reducing costs [35][36] - Alibaba Cloud's unique position as a model provider and its engineering capabilities are seen as significant advantages in the rapidly evolving embodied intelligence landscape [37][41]

Core Insights - The rise of AI is leading to a new era characterized by embodied intelligence and intelligent assisted driving, marking the beginning of a competitive landscape in the Agentic AI domain [1] - Companies are increasingly focusing on the transition from modular to end-to-end architectures in intelligent driving technology, which is seen as a paradigm shift [2][3] - The demand for data and computational power is growing exponentially, posing significant challenges for the industry [3][4] Group 1: Industry Trends - The shift to "end-to-end" architecture in autonomous driving has been a game changer, allowing for rapid iteration and adaptation to complex scenarios [2] - The traditional modular approach has been limited by its reliance on manually defined rules and case-by-case adjustments, while the new VLA architecture integrates visual, linguistic, and action capabilities [2] - Companies are investing heavily in cloud infrastructure to support the increasing demands of data processing and model training in both intelligent driving and embodied intelligence sectors [4][5] Group 2: Technological Developments - The integration of large-scale data management and advanced AI infrastructure is crucial for the success of intelligent driving and embodied intelligence applications [4][5] - Alibaba Cloud has upgraded its intelligent assisted driving solutions, achieving significant improvements in data management and model training efficiency [5] - The collaboration between Alibaba Cloud and NVIDIA aims to enhance the capabilities of Physical AI, providing a comprehensive platform for data processing, simulation, and model training [6][8] Group 3: Future Outlook - The future of AI is expected to involve widespread deployment of agents and robots in various sectors, necessitating substantial computational resources [8] - The competition for building "super AI clouds" is intensifying, with a focus on creating robust infrastructure to support advanced AI applications [8] - The industry is moving towards a model where only a few major cloud platforms will dominate, driven by the need for extensive resources and capabilities [8]

Investment Rating - The report maintains a "Recommended" rating for the autonomous driving industry [1] Core Insights - The autonomous driving industry is at a pivotal point of capability evolution, with advancements in AI and high-performance computing driving the development of autonomous driving solutions [5][8] - The report identifies that the differentiation in autonomous driving capabilities among automakers is diminishing as the industry matures, leading to a focus on safety features and user experience [5][8] Summary by Sections 1. Industry Overview - The report outlines the current state of the autonomous driving industry, highlighting the convergence of technology paths and the need for enhanced safety features as the industry transitions to higher levels of automation [5][6] 2. Corporate Strategy and Organization - Companies are adjusting their organizational structures and research focuses to improve R&D efficiency and commercialization pace, with a notable shift towards AI applications [6][52] - The report emphasizes the importance of maintaining product strength and long-term operational capabilities in a price-sensitive competitive landscape [6][52] 3. Technical Capabilities - **Sensors**: The report discusses the parallel development of multiple sensing solutions, including LiDAR, cameras, and radar, to meet safety and reliability requirements [7] - **Computing Power**: It highlights the establishment of cloud-based computing centers for model training and algorithm iteration, with Tesla leading at over 75 Eflops and some Chinese automakers achieving around 10 Eflops [7] - **Vehicle-Cloud Models**: The report notes a shift from rule-based to data-driven models, enhancing decision-making capabilities through the integration of multimodal data [7] 4. Consumer Perception - The report indicates that autonomous driving products are becoming increasingly recognized by consumers, with features such as parking assistance and safety enhancements being continuously optimized [7][49] 5. Investment Recommendations - The report suggests focusing on automakers making significant advancements in R&D and functional deployment, including Tesla, Xpeng, Li Auto, NIO, and Xiaomi, as well as leading third-party solution providers like Momenta and Horizon Robotics [8][50]