前言 在AI的世界里，"疯狂"往往不是一个贬义词，而是突破边界的标志。特斯拉FSD 14.1.2版本的推出，正是这样一个"疯狂的信号"。 这次更新中最受关注的，就是那个代号"Mad Max"的驾驶模式——疯狂模式。很多人以为这只是个彩蛋、一个噱头，实际上，它背后反映的是特斯拉自动 驾驶团队前所未有的自信：他们终于敢让AI，像一个真正的人那样，在复杂的路况中主动出击。 顺便说下，会员双十一抢跑已经开始， 前200 名加送两本小册子：《全球配置指南》《牛熊&回撤分析》。（目前还剩最后50+本啦～）详细会员介绍在 文末 从保守到主动 AI驾驶的分水岭 过去的所有自动驾驶系统：无论是国内还是海外的车企。都有一个共同点：保守。能不加塞就不加塞，能不超车就不超车，能慢就绝不快。因为他们的策 略核心是"安全第一"，在极端复杂的真实环境中，一点点保守能换来巨大的容错空间。 但特斯拉这次反过来了。FSD V14.1.2让车辆可以在确保安全的前提下，执行更具侵略性的驾驶动作：精准的超车、果断的变道、在多车汇流区主动"抢 位"。 这不是简单的"更快"，而是更像人类驾驶——一种拥有判断力和胆量的AI决策。 这也意味着，特斯拉已经不再 ...

Group 1 - The humanoid robot market is on the verge of explosive growth, with a projected market size of approximately 9 billion in 2025, expected to soar to 150 billion by 2029, reflecting a compound annual growth rate (CAGR) exceeding 75% [2] - The core drivers of this market growth will be industrial handling and medical applications, highlighting the importance of advanced processing capabilities in humanoid robots [2][5] - The performance of processors is critical as it directly influences the intelligence level and application potential of humanoid robots, making them the foundational element of the robotics industry [1][5] Group 2 - The current processor supply for humanoid robots is dominated by NVIDIA and Intel, while domestic chip manufacturers are still in the catch-up phase [6] - Tesla is noted for its capability to develop its own chips, such as the Dojo chip for AI model training and the FSD chip for real-time operations in robots, while other manufacturers primarily rely on Intel and NVIDIA chips [6][8] - The Jetson Orin series from NVIDIA is widely used, providing up to 275 TOPS of computing power, significantly enhancing the capabilities of humanoid robots [9][10] Group 3 - Domestic manufacturers are accelerating the development of their own humanoid robot chips to compete with foreign dominance, focusing on integrating general intelligence with practical application needs [10][11] - The RK3588 and RK3588S chips from Rockchip have been adopted by several humanoid robot manufacturers, showcasing their potential in the robotics field [11] - The RDK S100 development kit from Horizon Robotics integrates both "brain" and "cerebellum" functions into a single SoC, simplifying hardware architecture and reducing development costs [12][14] Group 4 - The trend towards "brain-cerebellum fusion" architecture aims to enhance the synchronization and efficiency of humanoid robots by integrating cognitive decision-making and motion control into a unified system [15][17] - Current challenges in the humanoid robot market include insufficient data accumulation, hardware architecture optimization, high costs, and safety concerns, which need to be addressed for large-scale commercialization [18][19][20]

在科技与资本的世界里，有些传言，起初只是爱好者论坛里的窃窃私语，而后发酵为财经媒体的捕风捉影，最终，它们会以一种不容忽视的方式，摆上最 严肃的议事日程。 特斯拉（Tesla）与马斯克旗下另一家AI初创公司xAI的合并，正是这样一则正在逐渐为现实的预言。 上周日，天桥资本（SkyBridge Capital）创始人Anthony Scaramucci的一条推文，为这本已滚烫的预期再添了一把烈火。他直言：随着马斯克加速将AI整合 到其商业版图中，特斯拉与xAI的合并"感觉不可避免"。 这并非孤立的喊话。当一份敦促特斯拉投资xAI的股东提案被正式提交，当马斯克那份价值可能高达万亿美元的新薪酬方案中，被摩根士丹利扒出隐藏 的"合并条款"时，所有人都意识到：一个横跨数字世界与物理世界，估值可能冲击8.5万亿美元的AI巨头，正在浮出水面。 01 将一个千亿美元估值的当红炸子鸡（xAI），并入一个已是万亿市值的巨头（特斯拉），绝非心血 interracial。我们从四个关键的信号中，可以清晰地看到 这条路径是如何被一步步铺就的。 1. 一纸提案：从民间呼声到正式议程 一切始于一位名叫Stephen Hawk的特斯拉散户股东 ...

Core Viewpoint - Tesla is focusing on the development of its AI5 and AI6 chips, which are expected to significantly enhance the performance and cost-effectiveness of its future products, particularly in AI and autonomous driving applications [1][4]. Group 1: AI Chip Development - Tesla's CEO Elon Musk announced that the AI5 chip is expected to be the best inference chip for models with fewer than 250 billion parameters, highlighting its low silicon cost and high performance-to-power ratio [1][2]. - The AI6 chip is anticipated to be even more advanced and will serve as the "unified heart" of Tesla's future AI ecosystem, with production expected to begin in 2025 at Samsung's Texas factory [3][4]. - The AI5 chip is designed for vehicle inference tasks and is projected to start mass production by the end of 2026, while the AI6 chip will first be used in Tesla's Cybercab and Optimus robot [3]. Group 2: Strategic Shift - Tesla has decided to discontinue its Dojo chip design project to concentrate resources on a single chip architecture, which Musk believes is a clear and correct decision for the company [2][3]. - This strategic shift aims to consolidate all chip talent towards the development of the AI5 and AI6 chips, enhancing the company's capabilities in creating critical AI technology [3]. Group 3: Integration and Future Plans - The self-developed chips are a key step in Tesla's "Master Plan Part 4," which aims to reduce reliance on external suppliers and provide a solid computational foundation for rapid iterations of its autonomous driving and robotics technologies [4].

Core Viewpoint - Arm has decided to develop its own chips, marking a significant shift from its traditional IP licensing model to a more direct involvement in chip manufacturing [1][3]. Group 1: Arm's Business Model and Market Position - Arm is known for its successful processor architecture, particularly in low-power, high-performance applications, widely used in billions of devices across mobile, embedded systems, and IoT [3][5]. - The company's IP licensing model allows various chip manufacturers to utilize its technology without fear of being "choked," fostering widespread adoption among companies like Xiaomi, MediaTek, and Apple [5][6]. - Arm's neutrality in the semiconductor field has been a key factor in its success, as it has acted as a technology provider without competing directly with its clients [10]. Group 2: Recent Financial Performance and Challenges - Arm's recent financial reports have shown troubling signs, with a 9% lower-than-expected revenue guidance following a record quarter of $1.24 billion in revenue and a 55% net profit growth [9]. - The company's net profit for the first quarter of fiscal year 2026 was $130 million, a 42% year-over-year decline, attributed to slowdowns in its core business areas: data centers, smart vehicles, and consumer electronics [9][11]. - Major clients like Tesla and Qualcomm are moving towards self-developed technologies, which poses a significant threat to Arm's traditional revenue model based on IP licensing [11]. Group 3: Strategic Shift and Market Implications - Arm's decision to enter chip manufacturing is seen as a response to declining revenues and the need to counteract clients who are attempting to bypass its licensing system [11][13]. - This move could lead to a major shake-up in the mobile chip market, potentially disrupting the current dominance of Qualcomm and MediaTek [13].

特斯拉的 Dojo 是目前全球最大的两款处理器之一。这些巨大的晶圆级芯片使用整块 300 毫米晶圆， 这意味着一次性构建更大的计算能力根本不可能。每个 Dojo 晶圆级处理器最多可容纳 8,850 个核 心，但其中一些核心在部署后可能会引发静默数据损坏 (SDC)，从而破坏大规模训练运行的结果。 大型处理器 鉴于 Dojo Training Tile（大型晶圆尺寸芯片）的极端复杂性，即使在制造过程中也不容易检测到有 缺陷的芯片，但是当涉及到静默数据损坏（SDC）时，事情会变得更加复杂。 请记住，所有类型的硬件都不可避免地会出现 SDC，但 Dojo 处理器的电流消耗高达 18,000 安培， 功耗高达 15,000 瓦，这会产生影响。不过，所有核心都应该按预期运行，否则特斯拉的 AI 训练将 变得更加复杂，因为数据损坏导致的一个错误就可能使数周的 AI 训练付诸东流。 公众号记得加星标⭐️，第一时间看推送不会错过。 来源：内容 编译自 tesla 。 在大型处理器上检测故障核心并将其禁用是一项挑战，但特斯拉开发了 Stress 工具，该工具不仅可 以在 Dojo 处理器上检测容易出现静默数据损坏的核心，还可以 ...

Investment Rating - The report maintains a positive outlook on the humanoid robot industry, emphasizing the importance of software development for commercialization [3][4]. Core Insights - The report identifies that the hardware maturity of humanoid robots is currently higher than that of software, with software being the key to commercialization. It highlights the need for advancements in algorithms, data, and control systems to drive the industry forward [3][5][6]. Summary by Sections 1. Algorithms: The Core of Embodied Intelligence - The algorithm framework is divided into two levels: the upper "brain" focuses on task-level planning and decision-making, while the lower "cerebellum" handles real-time motion planning and joint control [3][11][18]. - The report discusses the evolution of control algorithms, noting a shift from traditional methods to modern approaches like reinforcement learning (RL) and imitation learning (IL) [3][19][29]. - The VLA (Vision-Language-Action) model is highlighted as a significant advancement in upper-level control, enabling robots to understand and execute tasks through natural language processing [3][36][40]. 2. Data: The Foundation of Algorithm Learning - Data quality and diversity are crucial for algorithm performance, with sources categorized into real data, synthetic data, and web data. Real data is the most accurate but least abundant [3][74][76]. - The report emphasizes the importance of remote operation and motion capture technologies for collecting high-quality real data [3][79]. 3. Control Systems: The Foundation of Embodied Intelligence - The control system is described as the "brain" of humanoid robots, consisting of hardware (SoC chips, CPUs, GPUs, NPUs) and software components [3][3][3]. - The report notes that the industry lacks a unified consensus on the structure of the "brain" and "cerebellum" in humanoid robots, which are essential for executing complex algorithms and tasks [3][3][3]. 4. Investment Opportunities - The report identifies several key companies in the humanoid robot industry worth monitoring, including: - Controller segment: Tianzhun Technology, Zhiwei Intelligent, Desay SV [4][4]. - Motion control technology: Huichuan Technology, Xinjie Electric, Leisai Intelligent, Gokong Technology, Tosida [4][4]. - Chip manufacturers: Rockchip, Horizon Robotics [4][4]. - Data collection equipment: Lingyun Optical, Aofei Entertainment [4][4].