在云端，要对海量的真实与合成数据进行预训练和完成世界模型的构建，可能需要数十万张AI加速卡 和数十个EFLOPS（百亿亿次浮点运算）级别的算力支撑，这构成了极高的资金与技术壁垒。在车端， 为了实现低成本、低延迟与高效能的实时响应，车载智能芯片的算力需求正从目前的最高500-600 TOPS，朝着2500 TOPS以上迈进。这场竞赛不仅考验着企业的资源投入，更考验着其在芯片设计、架 构创新与系统整合上的综合实力。 面对业界最关心的数据挑战，尤其是类似FSD入华可能面临的"水土不服"问题，邓志东表示，一个出租 车司机比新手安全，主要源于他积累了更长的驾驶里程，而非智商更高或书本知识更丰富。要让自动驾 驶的安全性超越人类，若采用世界模型智能体方法，AI所需要的学习里程必须是人类司机的上千倍。 依靠实车路测采集真实数据成本极高、周期漫长，因此利用"数字孪生"技术生成海量的"合成数据"成为 了破局点。这意味着，能提供高质量仿真平台与数据服务的公司，将在未来产业链中更有价值。 邓志东介绍，目前一场激烈的"算力军备竞赛"已经拉开帷幕。这是一场在云端与车端同时进行的双线战 争。 新京报贝壳财经讯（记者林子）"智能驾驶正迎来它 ...

Core Insights - Tesla's Dojo supercomputer project has been terminated, while Waymo is advancing its autonomous driving services in Denver and Seattle, highlighting a divergence in the autonomous driving industry [1][20]. Group 1: Dojo Project - Tesla announced the dissolution of the Dojo team and the termination of the supercomputer project in August 2025, marking the end of a six-year effort [1][6]. - The Dojo was designed to train Tesla's Full Self-Driving (FSD) neural networks and was expected to achieve significant computational power by 2024 [4][5]. - Despite substantial investment and development, the project was deemed a "dead end" by Elon Musk, leading to the departure of key personnel and the formation of a new company, DensityAI [6][8]. Group 2: Tesla's New Strategy - Following the termination of Dojo, Tesla has shifted its focus to the Cortex training cluster, which consists of 50,000 H100 GPUs, enhancing FSD performance [9][11]. - Tesla has signed a $16.5 billion order with Samsung for AI6 chips, indicating a strategic pivot from in-house chip development to partnerships [11] . Group 3: Waymo's Expansion - Waymo is set to launch autonomous taxi services in Denver and Seattle, with testing beginning under human supervision [12][14]. - The company plans to expand its services to ten new cities by 2025 and has already established operations in several major cities [14][15]. Group 4: Global Competition - The competition in the autonomous driving sector has intensified, particularly between companies in the U.S. and China, with various firms pursuing different technological paths [15][20]. - Notable developments include Baidu's Apollo Go service and the introduction of Robotaxi GXR by Chinese company WeRide in Singapore [15][20]. Group 5: Future of Autonomous Driving - The termination of the Dojo project and Waymo's service expansion signify a pivotal moment in the autonomous driving industry, with three distinct paths emerging: Tesla's vertical integration strategy, Waymo's gradual expansion, and the AI network approach represented by companies like MogoMind [20][21].

Summary of Key Points from the Conference Call Industry Overview - The autonomous driving industry is at a pivotal point transitioning from L2 to L3 commercialization, with full-stack self-research manufacturers and third-party suppliers gaining a competitive edge [1][4] - Major players in the autonomous driving sector include Tesla, Xpeng, Li Auto, NIO, and third-party suppliers like Momenta and Yunrong Qixing [1][5] Core Insights and Arguments - The development of cloud-based intelligent computing centers and mass production of high-performance chips are crucial drivers for the industry [1] - Companies are investing heavily in R&D, with Tesla's HW5.0 featuring 4D millimeter-wave radar and Li Auto's L series equipped with laser radar [6][10] - Regulatory policies significantly impact the industry, with L2 standardization and multiple regions opening L4 commercialization pilot projects [8] Technological Developments - Xpeng is shifting to a pure vision solution to enhance visual perception and reduce hardware costs, while Huawei's ADS 4.0 supports high-speed L3 commercialization [3][12] - The VLA model integrates visual, language, and behavioral modules to optimize vehicle decision-making [3] - The industry is witnessing a shift towards data-driven development, with companies showcasing their cloud-based world models and parameter scales [29] Competitive Landscape - Leading companies in autonomous driving include Tesla, Xpeng, Li Auto, NIO, and Xiaomi, with significant contributions from domestic suppliers like SUTENG, Hesai Technology, and others [5][26] - Traditional manufacturers are increasingly opting for third-party solutions to shorten product cycles and reduce time costs [17] R&D and Investment Trends - Companies like NIO have invested over 10 billion yuan in R&D for three consecutive years, but face challenges in achieving commercial breakthroughs [14] - Xiaomi's growth in the autonomous driving sector is driven by its potential rather than current capabilities, with expectations for its models to feature laser radar [16] Consumer Perception and Market Trends - The development of intelligent driving technology includes advancements in features like high-speed NOA and parking functionalities [32] - Safety features are evolving, with the introduction of proactive avoidance systems to enhance driving experience [33] Investment Opportunities - Investors should focus on leading autonomous driving solution providers and full-stack self-research manufacturers, especially as regulatory frameworks evolve [36]

Core Insights - The report focuses on the LiDAR industry, highlighting its dual-driven development in advanced intelligent driving and robotics sectors [1] Downstream Market Applications - Sensor fusion trend: LiDAR collaborates with cameras and millimeter-wave radars to compensate for the shortcomings of pure vision solutions, achieving a target tracking accuracy of 75% compared to 56% for pure vision in 2023 [1][42] - Market size: The global automotive LiDAR market is projected to reach $5.26 billion in 2023 and $3.632 billion by 2029 [2] - Robotics sector: 2025 is anticipated to be the commercial year for humanoid robots, with companies like Tesla planning to produce 5,000 units of Optimus [2] - Market potential: In 2023, robotics accounted for 68.2% of LiDAR applications, with the Chinese robotics LiDAR market expected to reach 28 billion yuan by 2030, reflecting a compound annual growth rate (CAGR) of 67.9% [2] Industry Development Drivers - Cost reduction: Leading manufacturers are lowering costs through self-developed SoC chips and optical integration, with prices for mainstream automotive LiDAR models expected to drop from 350,000-400,000 yuan in 2023 to 300,000-350,000 yuan in 2024 [5] - Increased vehicle integration: L3 level requires one front-facing and 2-3 blind-spot radars, while L4 may require up to 10 units, driving demand growth [5] - Policy and technology support: National and local policies are promoting intelligent driving development, with L3 and above levels creating urgent demand for LiDAR [5] - Market share: By 2024, Chinese manufacturers are expected to lead the global market, with Hesai Technology (33%), RoboSense (24%), Huawei (19%), and TuSimple collectively holding 88% [5] Competitive Landscape and Manufacturer Dynamics - Hesai Technology: Projected revenue of 530 million yuan in Q1 2025 (+46.3%), with an expected annual delivery of 1.2 to 1.5 million units, including nearly 200,000 units for robotics [3] - RoboSense: Q1 2025 robot product sales reached 11,900 units (+183.3%), launching the MX LiDAR to break the $200 price barrier [3] Future Trends - The dual-driven development of intelligent driving and robotics is expected to push the Chinese LiDAR market to 43.18 billion yuan by 2026, with chip and solid-state technologies further driving cost reductions [12]

Investment Rating - The investment rating for the company is "Buy" and is maintained [6]. Core Insights - The company is developing a 720 billion parameter large-scale autonomous driving model, named "Xiaopeng World Base Model," which aims to significantly enhance the intelligence of AI vehicles and support various applications such as AI robots and flying cars [2][4][8]. - The company has established a robust AI infrastructure, including a computing cluster with a cloud capacity of 10 EFLOPS, which is crucial for building a "cloud model factory" [8][30]. - The company has achieved significant milestones in AI development, including the successful implementation of scaling laws in autonomous driving, demonstrating that larger models yield better performance [36][41]. Summary by Sections Company Overview - Xiaopeng Motors is focusing on AI-driven autonomous driving technology, with a significant investment in developing a large-scale model that is 35 times the parameter size of mainstream models [27][30]. AI Infrastructure - The company has built the first 10,000-car intelligent computing cluster in the domestic automotive industry, achieving a cloud computing capacity of 10 EFLOPS, with a high operational efficiency of over 90% [8][25]. - The data infrastructure has been self-developed to enhance data access efficiency, with video data for training the base model currently at 20 million clips, expected to increase to 200 million clips by the end of the year [30][20]. Technological Advancements - The company is leveraging a multi-modal model that incorporates visual understanding, chain reasoning, and action generation capabilities, which are essential for achieving L3 and L4 level autonomous driving [8][23]. - The company has initiated the training of the 72 billion parameter base model, focusing on reinforcement learning to enhance the model's performance and adaptability [45][50]. Market Position - The company is positioned to benefit from a new vehicle launch cycle in 2025, with multiple new models expected to drive sales growth [8][30]. - The anticipated revenue for 2025 is projected to be 99.1 billion, corresponding to a price-to-sales ratio of 1.4X, indicating a strong market outlook [8].