Core Viewpoint - The rapid commercialization of intelligent driving is surpassing expectations, with Tesla's production of directionless vehicles and the global operation of unmanned fleets marking a significant turning point in the industry [2][3]. Group 1: Intelligent Driving Explosion - The first directionless vehicle designed for Robotaxi by Tesla is set to begin mass production in April 2026, eliminating the steering wheel and pedals, fully controlled by AI [3][4]. - Tesla's Full Self-Driving (FSD) system is expected to receive full approval in China by 2026, which will significantly transform the domestic intelligent driving market [3][4]. - The scale of autonomous driving fleets in China is expanding, with companies like Baidu and Xiaopeng actively deploying unmanned vehicles in various cities and even internationally [4]. Group 2: Scientific Principles of Intelligent Driving - Autonomous driving hardware is surpassing human limits, with AI capable of perceiving distances over 200 meters and making decisions with a reaction time of 10-30 milliseconds, far exceeding human capabilities [5][7]. - The learning ability of autonomous driving systems allows for rapid sharing of experiences across vehicles, enabling exponential evolution in driving capabilities [7][8]. Group 3: Industry Trends - The industry is witnessing a rapid transition to L2+ level intelligent driving becoming a standard feature, with high-level driving assistance (NOA) becoming widely available [10][11]. - Many companies are bypassing L3 technology due to its ambiguous responsibility issues and are directly targeting L4 autonomous driving [13][14]. - The competition is intensifying as companies leverage large-scale data and advanced algorithms to establish competitive advantages in the intelligent driving space [15][18]. Group 4: Future Trends - The technological foundation of intelligent driving is being restructured with end-to-end models and world models, significantly reducing the complexity of programming [17]. - High-quality data loops are becoming critical, with companies that can process data faster gaining a competitive edge [18]. - The cost of hardware for intelligent driving is decreasing, leading to widespread adoption and making advanced driving features more accessible [19]. Group 5: Future Scenarios of Intelligent Driving - The business model will shift from selling cars to providing mobility as a service (MaaS), fundamentally changing consumer behavior and market dynamics [24][25]. - Urban planning will be transformed, reducing traffic congestion and parking needs as vehicles are algorithmically managed [26][27]. - Autonomous vehicles will become integral to energy management, acting as mobile energy sources that can charge during off-peak hours and supply energy back to the grid [28]. - The personal experience of commuting will be revolutionized, as vehicles become multifunctional spaces for entertainment and work [29][30].

Core Viewpoint - BYD is set to hold a groundbreaking technology launch event on March 5, 2026, with potential announcements regarding key technological advancements in battery and intelligent driving systems [1] Group 1: Battery Technology - The upcoming event may feature the second generation of blade batteries, which are expected to enhance safety and energy density [1] - An upgraded "megawatt flash charging" technology is anticipated, with peak power potentially reaching megawatt levels [1] - A high-voltage platform supporting ultra-fast charging is likely to become more widely adopted [1] Group 2: Intelligent Driving - The "Tianshen Eye" advanced driving assistance system may receive a version update, possibly incorporating new capabilities based on end-to-end large models [1] - A new generation of hybrid systems is also considered a potential announcement at the event [1] - Electric vehicles priced below 200,000 yuan are expected to undergo collective upgrades, with several models achieving a range of 700 km, and lidar-based driving solutions will be significantly introduced in this price segment [1]

Investment Rating - The report maintains an "Outperform" rating for the industry [4] Core Insights - The commercialization of L3 autonomous driving has reached a turning point, transitioning from "optional" to "preferred" [1][14] - The approval of L3 licenses marks a significant shift from technical validation to operational readiness, establishing a clear regulatory framework [14] - The competitive landscape has evolved from hardware specifications to a focus on end-to-end models driven by data and algorithms, enhancing industry concentration [1][20] Summary by Sections 1. Commercialization of L3 Autonomous Driving - The dual approval of operational licenses and road rights has established a commercial closure for L3 autonomous driving, with clear responsibilities defined [14] - User perception of intelligent driving has matured, with advanced features becoming a key factor in purchasing decisions, as 60% of consumers view autonomous driving as the most anticipated technological breakthrough [16][17] - The paradigm of intelligent driving is shifting towards an end-to-end model, where the core competitive logic is now based on data quality, computational power, and model iteration efficiency [20][21] 2. Huawei's QianKun Intelligent Driving - Huawei's QianKun Intelligent Driving has evolved from a single supplier to a public technology platform, enhancing its market position [30][45] - The company has developed a multi-tiered cooperation model, including component supply, HI mode, and Harmony Intelligent Driving, to cover various market segments [31][32] - The QianKun Intelligent Driving system has undergone significant iterations, establishing a technological moat centered around end-to-end models [36] 3. Independent Entity "Yinwang" - The establishment of Yinwang as an independent entity has alleviated concerns among automakers regarding technology control, allowing for broader collaboration [41][42] - The strategic partnership with automakers has led to a valuation of 115 billion RMB, positioning Yinwang as a unicorn in the industry [45] - Yinwang aims to become a neutral public platform for the smart electric vehicle industry, similar to Bosch's role in the traditional automotive sector [48][49] 4. Hardware Cost Breakdown - The cost of intelligent driving hardware per vehicle exceeds 10,000 RMB, with significant portions attributed to chips and PCB components [2][3] - The potential market space for various components in the intelligent driving supply chain is substantial, with estimates reaching billions in growth opportunities [2] 5. Investment Opportunities - The report identifies potential investment opportunities in advanced process wafer fabs, packaging and testing companies, and PCB manufacturers, highlighting the growth potential in the intelligent driving sector [3]

Core Viewpoint - The U.S. is shifting its autonomous driving policy significantly with the introduction of the "2026 Autonomous Vehicle Act," aimed at overcoming competition from China and facilitating the commercialization of autonomous vehicles [1][20]. Group 1: Legislative Changes - The new legislation increases the annual exemption limit for manufacturers from 2,500 to 90,000 vehicles, enabling large-scale production for companies like Tesla and Waymo [1][5]. - The act establishes federal preemption, preventing states from creating their own autonomous vehicle performance standards, thus ensuring a unified regulatory framework [2][10]. - The regulatory approach evolves from requiring detailed original code submissions to a "Safety Case" report, allowing companies to demonstrate safety through evidence rather than code inspection [3][15]. Group 2: Industry Implications - The legislation is tailored to support U.S. companies like Tesla and Waymo by addressing production capacity and expansion barriers, which are critical for their growth [4][9]. - The introduction of the "Safety Case" aligns with the technological evolution of autonomous driving, allowing companies to protect their proprietary algorithms while proving safety outcomes [19][24]. - The act signals a strategic move to enhance the competitive position of U.S. firms against Chinese counterparts, with implications for data security and market access [20][25]. Group 3: Challenges for Chinese Companies - Chinese Robotaxi companies face legal and regulatory hurdles that hinder their ability to scale, particularly in the absence of a standardized production model [21][23]. - The U.S. legislation's provisions may create a competitive disadvantage for Chinese firms, as they struggle with local regulations and lack of a cohesive national framework [22][24]. - The ongoing regulatory fragmentation in China complicates the operational landscape for Robotaxi companies, making it difficult to achieve the necessary scale for profitability [23][24].

Core Viewpoint - Waymo has secured a record-breaking $16 billion funding round, marking the largest financing in the history of global autonomous driving, significantly boosting its valuation to $126 billion [1][3][7]. Funding and Valuation - The $16 billion funding is equivalent to approximately 111.1 billion RMB, which is over four times Tesla's projected net profit for 2025 [4]. - Cumulatively, Waymo has raised a total of $27.1 billion (about 188.1 billion RMB) across four funding rounds [6]. - Waymo's valuation has reached a peak of $126 billion (approximately 875.3 billion RMB), surpassing that of BYD [7]. Industry Context - The autonomous driving sector has seen significant investments, with major players like Cruise, Argo AI, and Zoox receiving billions in funding, although many have failed to achieve sustainable business models [14][16]. - In China, companies like Baidu's Apollo have invested heavily, with Baidu alone spending around 150 billion RMB over ten years [18]. Investment Dynamics - The allure of the autonomous driving market is driven by its potential for high returns, with estimates suggesting that the Robotaxi segment alone could reach a trillion-dollar market size [24]. - Despite the high risks and many companies failing, capital continues to flow into the sector, indicating a strong belief in the future profitability of autonomous driving [22][23]. Market Trends and Future Outlook - The investment landscape is changing, with a notable reduction in the number of companies receiving funding, leading to a "dragon effect" where leading firms like Waymo attract the majority of capital [51]. - Waymo is expanding its commercial operations, with plans to operate in over 20 cities globally by 2026, indicating a growing acceptance and integration of autonomous vehicles into everyday life [46][48]. Consumer Demand - Waymo's Robotaxi service is adapting to meet real-world needs, such as providing safe transportation for teenagers, which reflects a growing acceptance of autonomous vehicles in various aspects of daily life [54][59]. - The emergence of new use cases for Robotaxi services demonstrates the evolving landscape of consumer demand and the potential for autonomous vehicles to integrate into everyday scenarios [61][66].

Core Viewpoint - Tesla's Full Self-Driving (FSD) is fundamentally a large model that utilizes a significant neural network architecture to achieve end-to-end driving capabilities, contrary to claims that it relies on numerous smaller models for various tasks [7][17]. Summary by Sections FSD Model Architecture - Tesla FSD is characterized as a large model, confirmed by Ashok at ICCV, which utilizes a massive neural network for computations from Photon In to Control Out [7][14]. - The architecture includes numerous model parameter files, which are primarily small task heads rather than independent models, indicating a more complex integration than previously assumed [6][10]. Parameter File Insights - The discovery of hundreds of neural network parameter files has led to skepticism about FSD being a large model; however, these files are largely associated with smaller tasks rather than the core end-to-end model [8][10]. - The parameter sizes for HW3 and HW4 show significant growth, with HW4's B core reaching 7.5GB, indicating a substantial increase in model complexity and capability [8][12]. Memory and Bandwidth Considerations - HW3's limited memory bandwidth of 68GB/s restricts the model size to approximately 1.8 billion parameters, while HW4's bandwidth of 384GB/s allows for a theoretical capacity of around 10 billion parameters [12][13]. - The use of a mixture of experts (MOE) architecture enables Tesla to optimize memory usage and enhance model performance without exceeding bandwidth limitations [13][16]. Technological Advancements - The assertion that Tesla's technology is outdated is challenged by the argument that significant engineering innovations contribute to advancements, similar to the development of reusable rockets [17]. - The integration of advanced engineering practices and innovative architectures positions Tesla as a leader in the autonomous driving sector, countering claims of technological inferiority [17].

Core Insights - The implementation of a 128 TOPS chip for city NOA (Navigate On Autopilot) has been successfully launched, challenging the previous consensus that a minimum of 200 TOPS was required for such technology, indicating a shift towards mainstream adoption in the market [1] - A report by the China Automotive Industry Economic and Technological Research Institute forecasts that by November 2025, the cumulative sales of passenger cars equipped with city NOA will reach 3.129 million units, with a penetration rate of 15.1%, an increase of 5.6 percentage points from 2024 [1] - The trend shows that city NOA is moving from high-end vehicles to mainstream passenger cars, with over 68.9% of city NOA-equipped vehicles priced below 300,000 yuan [1] Market Penetration - In 2024, the penetration rate of NOA in the domestic automotive market is projected to be 7.3%, with city NOA at 1.52%, indicating a significant increase in adoption within a year [3] - By November 2025, 15 out of every 100 passenger vehicles are expected to be equipped with city NOA, marking a rapid scale-up in its market presence [3] Competitive Landscape - The focus of the industry has shifted from highway NOA to city NOA, with the latter being more complex to implement [4] - Over 78.3% of city NOA-equipped vehicles sold by November 2025 are expected to be self-developed by automakers, indicating a strong market position for companies that invest in in-house technology [4] Key Players - Notable brands in the self-developed city NOA segment include Tesla, NIO, Xpeng, Li Auto, and Xiaomi, each leveraging their unique technological capabilities to enhance their offerings [5] - Approximately 21.7% of city NOA-equipped vehicles are developed in collaboration with third-party suppliers, with traditional automotive brands making up 64.4% of these partnerships [5] Supplier Dynamics - The market for third-party city NOA suppliers is dominated by Momenta and Huawei, which together account for about 80% of the market share [6] - By November 2025, Momenta is expected to have a leading position with 414,400 units, while Huawei's HI model will account for approximately 19.76% of the third-party supplier market [6] Future Outlook - The upcoming mandatory national standard for intelligent connected vehicles is expected to set a clear safety baseline and further promote the marketization of related technologies [7] - The integration of end-to-end large models is seen as a key driver for the acceleration of city NOA, enhancing safety and user experience [8] - By 2030, city NOA is projected to become a mainstream feature in advanced driver assistance and autonomous driving systems, with significant market penetration expected in the 150,000 to 200,000 yuan price range [6][9]

Core Viewpoint - Tesla's Full Self-Driving (FSD) is characterized as an end-to-end large model, challenging the notion that it is merely a combination of nearly 200 small scene models [1][11]. Group 1: Model Architecture and Parameters - The B-core neural network parameters significantly exceed those of the A-core, with only 61 shared parameter files, indicating that the redundancy design between A and B cores has become impractical with the rapid expansion of the neural network scale in Tesla V12 [5]. - The discovery of many model parameters being parts of a large model, indicated by naming conventions like FSD E2E FACTORY PART X, suggests a distributed deployment strategy for model parameters across different chips, which is common in the era of large models [6]. - Tesla's HW3 has limited memory bandwidth of 68GB/s, theoretically allowing for a maximum of 1.8GB of model parameters to support a 36Hz output, while HW4, with a bandwidth of 384GB/s, could theoretically support around 10 billion parameters [7][8]. Group 2: Mixture of Experts (MoE) Architecture - The use of a Mixture of Experts (MoE) architecture allows Tesla to run large-scale end-to-end models at high frequencies on relatively older chips by activating only a subset of expert networks, thus optimizing memory bandwidth usage [8][10]. - Elon Musk and Ashok Elluswamy have indicated that the FSD employs MoE architecture, which supports the idea of localized parameters for different regions while maintaining a generalized approach [9][10]. Group 3: Technological Advancement - The assertion that FSD is a backward technology is dismissed, emphasizing that technological advancement is not solely defined by scientific discoveries but also by engineering innovations, as exemplified by Tesla's achievements in rocket technology and engineering [11].

Core Insights - The automotive industry is increasingly focusing on intelligent assisted driving, with companies like BYD investing over 100 billion yuan and forming specialized teams to enhance their capabilities [2] - The shift in the industry is moving from scale and accessibility to user experience, emphasizing the importance of value creation in the competitive landscape [3][12] Investment and Development - BYD has established a team of over 5,000 for assisted driving and aims to deploy its "Tian Shen Zhi Yan" system in over 2.5 million vehicles by December 2025, generating over 150 million kilometers of effective driving data daily [3] - The establishment of new companies, such as the joint venture between BAIC and Horizon Robotics, indicates a growing trend towards collaboration in the intelligent driving sector [2][4] Technological Pathways - Companies are adopting full-stack self-research as a key strategy to deepen their understanding of technology and user needs, with BYD exemplifying this approach through significant investment and team size [3][5] - Open collaboration with third-party solution providers is also becoming a vital strategy, allowing companies to enhance their systems' performance and functionality through shared resources [4][5] User Experience and Market Trends - The concept of "intelligent driving equity" is emerging, indicating a shift towards making advanced driving features accessible in the mid-range vehicle market, particularly in the 100,000 to 150,000 yuan price range [12][14] - User feedback highlights the importance of stability in high-frequency scenarios, ease of use in interaction design, and the need for cost-effective solutions [13][14] Industry Dynamics - Suppliers are transitioning from providing single hardware products to offering integrated software and hardware solutions, becoming deep partners in the development of intelligent driving systems [6][8] - The competitive focus is shifting from merely having intelligent driving capabilities to ensuring their reliability, safety, and user-friendliness in complex driving environments [7][11] Future Outlook - The industry is expected to evolve towards a model where technology is not only usable but also provides a good user experience, ultimately leading to a sustainable business model [14] - As regulations and infrastructure improve, the Chinese intelligent driving industry is poised for high-quality development on a global scale, driven by technology and user experience [14]

Core Insights - The report indicates that the cumulative sales of passenger cars equipped with urban NOA functions in China reached 3.129 million units from January to November 2025, marking a significant expansion in the industry [1][2] - The urban NOA market is characterized by a "dual strong" pattern dominated by Huawei and Momenta, which together hold approximately 80% of the third-party supplier market share [1][3] Group 1: Market Overview - The penetration rate of urban NOA-equipped passenger cars in China was 15.1% of the insured passenger car volume from January to November 2025, an increase of 5.6 percentage points compared to 2024 [2] - The report highlights that the urban NOA market has entered a phase of large-scale popularization, with profound changes in market structure and competitive factors [1][3] Group 2: Competitive Landscape - The market is driven by both self-research by car manufacturers and partnerships with third-party suppliers, with domestic brands accounting for 81.1% of the sales of urban NOA-equipped passenger cars [3] - Momenta leads the third-party supplier market with a total of 414,400 units equipped with urban NOA, representing about 61.06% of the third-party supplier share, while Huawei's HI model has approximately 134,100 units, accounting for about 19.76% [3][4] Group 3: Technological Advancements - The report emphasizes that algorithm, data loop capability, and mass production experience are critical factors determining the market position and development speed of auxiliary driving suppliers [6][8] - Momenta's R6 reinforcement learning model is noted as the first in China to achieve mass production based on an end-to-end architecture, enhancing safety and efficiency [8] Group 4: Future Trends - The urban NOA is expected to accelerate systemic changes in intelligent driving technology routes, core architectures, and industrial ecosystems, extending applications from highway NOA to urban NOA [9] - The competitive landscape is evolving with a focus on end-to-end models, pushing for integrated system architecture and the maturation of vehicle-road-cloud collaboration and regulatory standards [9]