图片来源：地平线 继与哈啰出行达成合作，地平线在Robotaxi领域又有新动作。 9月26日，地平线发布融资公告称，计划筹集约58亿港元。 地平线通过 "先旧后新" 配股模式募资，计划发行 6.39 亿股新股（占当前总股本 4.6%），配售价 9.99 港元 / 股，较前一日收盘价10.60 港元折让 5.75%。最终计划募资总金额约为63.4亿港元，扣除费用后， 地平线净得约58亿港元，是其近年规模最大的单次融资之一。 所谓 "先旧后新"，即现有大股东先向市场投资者转让部分持股，再用转让资金同步认购公司新发行的 同等数量股票 —— 既避免新增股份对现有股东股权的过度稀释，又快速引入外部资本，同时通过小幅 折让吸引投资者，兼顾 "融资效率" 与 "股价稳定"。 同时，公司承诺，未来90天内，不会再额外发新股或做类似融资动作，以免对股价造成太大冲击。 在介绍本次募资资金用途时，地平线明确提到投资新领域，比如与Robotaxi相关的计划。 除了投资新领域（Robotaxi）之外，这笔资金还要用来拓展海外业务（之前的融资主要用于国内扩 张）；投入研发，提升技术，尤其是中高阶辅助驾驶解决方案；以及战略投资上下游的合作 ...

Core Insights - Horizon Robotics and Hello signed a strategic cooperation agreement to develop an end-to-end L4 autonomous driving core system for Robotaxi operations, aiming to accelerate urban deployment and commercialization [1][3] Group 1: Strategic Partnership - The partnership leverages strong financial backing from investors like Ant Group and CATL, providing robust funding and a large user base to support scalable commercial applications [3] - Hello's first mass-produced Robotaxi model, "HR1," was globally unveiled, with plans for mass production by 2026, targeting over 10 cities and a fleet of 10,000 vehicles, and aiming to deploy over 50,000 vehicles by 2027 [3] Group 2: Technology and System Capabilities - Horizon Robotics will provide a complete core capability system for L4 autonomous driving, including key components like the Journey 6P chip and HSD urban auxiliary driving system, designed for high computational redundancy and integration efficiency [3][4] - The company is transitioning from a hardware and software provider to a technology infrastructure platform builder, enhancing its role in the industry and creating new growth opportunities and valuation models [5] Group 3: Development Approach - Horizon Robotics adopts a "gradual approach" starting from L2 driving assistance, utilizing real-world data to iteratively improve safety and regulatory strategies, which contrasts with the high-cost and unstable experience of a leapfrogging approach directly to L4 [4] - The gradual approach allows for the accumulation of real road data and alignment with evolving regulatory environments, facilitating market entry and expansion [4]

Core Insights - Horizon and Hello signed a strategic cooperation agreement to develop low-cost, high-safety, and reliable intelligent driving technology for Robotaxi operations [1] - The collaboration marks a significant step in Horizon's strategic layout, enabling L2 to L4 scenarios in intelligent driving [1] - The Robotaxi service market in China is projected to reach a potential scale of $183 billion by the late 2030s, with approximately 4 million Robotaxis expected to be operational nationwide [1] Group 1: Technology and Collaboration - Horizon has built a technology foundation suitable for Robotaxi deployment, leveraging its full-stack technology and mass production experience [2] - The Zhencheng 6P provides 560 TOPS of computing power, with a modular system capable of achieving over 1,000 TOPS, meeting the demands for multi-sensor fusion and large model applications [2] - The collaboration aims to optimize the core software and hardware systems of Robotaxi, reducing development cycles and integration costs [2] Group 2: Business Model and Market Strategy - Hello's Robotaxi business is supported by investments from Hello, Ant Group, and CATL, facilitating large-scale commercial applications [3] - The first mass-produced Robotaxi model, HR1, was unveiled, with plans for mass production in 2026 across over 10 cities, targeting a fleet of 10,000 vehicles [3] - Two main technical routes for L4 autonomous driving are identified: a leapfrogging route focusing on regulatory compliance and a progressive route driven by user experience [3] Group 3: Industry Perspectives - The progressive route, represented by companies like Tesla and Horizon, emphasizes speed and iterative development from L2 to L4 capabilities [4] - Horizon's president stated that a solid L2 foundation is essential for achieving L4, advocating for a data-driven approach to expand operational design domain (ODD) boundaries [4] - The approach aims to significantly reduce per-vehicle investment and enhance return on investment (ROI) [4]

这种被称为"远程操作"的技术方案，在自动驾驶行业已形成两种典型实践模式。以谷歌(GOOGL.US)旗 下Waymo为代表的渐进式路线，采用"车辆自主决策为主，人工远程介入为辅"的混合模式。当自动驾驶 系统遭遇道路施工、紧急车辆通行等复杂场景时，车辆会通过蜂窝网络向远程安全员发起咨询，由人类 判断是否放行。这种模式在Waymo的运营中已形成标准化流程，其前CEO约翰·克拉夫西克曾强调"软件 始终掌握最终决策权"，人类仅在必要时提供辅助判断。 与之形成对比的是其他企业采用的"虚拟驾驶员"方案，由远程操作员直接通过5G网络操控车辆方向 盘、油门和制动系统。这种模式对网络稳定性要求极高，卡内基梅隆大学自动驾驶专家菲利普·库普曼 教授指出："在公共道路场景下，蜂窝网络延迟和连接中断风险始终存在，当前技术条件下远程驾驶本 质上仍属不可靠方案。"他特别警示，当测试车辆规模从10辆级扩大至百万辆级时，通信中断概率将呈 指数级上升。 特斯拉的自动驾驶出租车方案呈现独特的技术路径特征。马斯克今年1月曾承诺将推出基于"完全无人监 督"系统的付费服务，但5月采访时改口称，初期运营将限定在奥斯汀特定安全区域，避开复杂路口，并 配置远程 ...