元戎启行CEO周光表示："芯片是辅助驾驶的核心算力底座，黑芝麻智能在车规级高性能芯片领域的技术积累与产品实力，与元戎启行的算法研发需求高度 契合。此次合作将实现'芯片+算法'的软硬一体深度融合，不仅能加速高阶辅助驾驶方案的量产落地，更能为车企提供更高效、可靠的集成化选择。期待与 黑芝麻智能携手，共同推动中国智能汽车产业链的高质量发展。" 黑芝麻智能深耕车规级芯片领域，自主研发的华山系列辅助驾驶芯片与武当系列跨域计算芯片，凭借自研IP核、全栈式算法及配套软件驱动，构建了完整的 技术支撑体系。其中，华山A2000芯片是黑芝麻智能2025年推出的面向下一代AI模型更高性能、更高效率的芯片平台，可充分满足高阶辅助驾驶在复杂场景 处理与多任务并行中的算力需求。A2000全面支持全场景通识辅助驾驶，通过赋予AI模型类人的常识与认知能力，使其能够在动态场景中做出更智能的决 策，显著提升驾驶体验的流畅度与安全性。A2000作为目前市场上少数几个坚持开放平台赋能生态的辅助驾驶计算平台计划于2026实现量产。 元戎启行致力于打造"物理世界的通用人工智能"，以创新技术打造AI司机，实现RoadAGI，引领人工智能行业变革。凭借富有 ...

12月12日，黑芝麻（000716）智能科技有限公司(以下简称"黑芝麻智能")与深圳元戎启行科技有限公司(以下简称"元戎启行")正式签署合作协议。双方在此 次合作中将建立资本与业务的双重纽带，叠加各自在芯片技术与辅助驾驶模型与算法领域的核心优势，基于黑芝麻智能(02533)下一代车规级高性能计算芯 片平台，联合打造高性能、高性价比的高阶辅助驾驶解决方案，共同拓展行业应用场景，推进技术规模化落地。 元戎启行致力于打造"物理世界的通用人工智能"，以创新技术打造AI司机，实现RoadAGI，引领人工智能行业变革。凭借富有竞争力的产品和服务，元戎启 行已与多家车企达成量产合作，共同推进十余款组合辅助驾驶汽车落地。截至目前，元戎启行已交付20万辆搭载城市NOA(领航辅助驾驶)的量产车型；2025 年10月，元戎启行在辅助驾驶城市NOA第三方供应商市场的单月市占率接近40%。 根据合作规划，黑芝麻智能将提供芯片产品及全套应用开发工具链，为解决方案的高效研发奠定硬件基础。元戎启行将依托自身在算法领域的深厚积累，在 黑芝麻智能下一代高性能芯片平台上完成高阶辅助驾驶算法的深度适配与优化，与黑芝麻智能联合打造"芯片+算法"的软硬 ...

Core Viewpoint - The article discusses the evolution and strategic positioning of Hezhima Intelligent in the AI chip industry, particularly focusing on its A1000 and A2000 series chips for autonomous driving applications, emphasizing the importance of collaboration with automotive manufacturers and the need for a comprehensive solution to drive adoption [2][3][4]. Group 1: Product Development and Market Positioning - Hezhima Intelligent was established in 2016, and its A1000 chip, launched in 2020, was the first domestic 16nm AI chip with over 50 TOPS of computing power, marking a significant milestone in the industry [2][3]. - The A1000 began mass production in 2022, with Hezhima adopting a "joint development" approach with automotive companies to ensure successful integration and application of its technology [2][3]. - The A2000 series, set to launch by the end of 2025, will feature a self-developed NPU architecture and will cater to various applications, including Robotaxi and general reasoning computing [3][4]. Group 2: Strategic Collaborations - Early collaborations with companies like Geely and FAW were crucial for Hezhima's development, allowing for rapid iteration and deployment of autonomous driving solutions [3][10]. - The partnership with Wuhan University aims to enhance the capabilities of humanoid robots using Hezhima's chips, showcasing the company's commitment to expanding its technological applications [4]. Group 3: Industry Challenges and Trends - The autonomous driving sector has faced challenges, including the need for significant investment and the slow adoption of new technologies by automotive manufacturers [9][10]. - The article highlights the importance of balancing chip performance and cost, as the industry moves towards more cost-effective solutions while maintaining high computational capabilities [6][15]. - The demand for higher computing power in autonomous driving is expected to evolve, with projections indicating that the necessary computing power for city NOA could converge to below 300 TOPS in the next three years [36]. Group 4: Future Directions and Innovations - Hezhima Intelligent is focusing on the integration of driving and cabin technologies, with its products designed to meet the needs of entry-level vehicles while maintaining cost efficiency [39][43]. - The company is also exploring the potential of embodied intelligence and robotics, indicating a strategic move towards diversifying its product offerings beyond automotive applications [48][54].

Core Insights - The 2025 IAA Mobility exhibition in Munich showcased the advancements in smart automotive technology, with Hezhima Intelligent presenting its innovative solutions for AI-driven smart vehicles [2] - Hezhima Intelligent introduced its safety intelligent foundation solution, which utilizes the Wudang C1200 family of cross-domain integration chips to enhance vehicle safety and reduce development costs [2] - The company emphasized the importance of integrating electronic and electrical architectures to facilitate seamless upgrades from entry-level to flagship models [2] Group 1 - Hezhima Intelligent's Wudang C1200 family of chips is central to its safety intelligent foundation solution, providing hardware security isolation and compatibility throughout the product lifecycle [2] - The Wudang C1200 solution aims to significantly lower development costs and shorten development cycles for automotive manufacturers [2] - The company is collaborating with leading global automotive firms to explore the future of smart vehicles powered by AI [2] Group 2 - The Huashan A2000 chip family was highlighted for its high performance and efficiency, integrating multiple functional units for advanced AI applications [3] - The Huashan A2000 platform supports a robust driving assistance technology foundation, ensuring optimal performance and flexibility [3] - Hezhima Intelligent's Wudang C1236 and C1296 chips are recognized for their capabilities in navigation assistance and multi-domain integration, respectively [3] Group 3 - The C1296 cross-domain integration solution has gained market recognition, with plans for mass production in multiple new models from Dongfeng Motor by the end of 2025 [4] - The Huashan A1000 family also showcased a mature software ecosystem, with successful applications in various vehicle models [4] - Hezhima Intelligent's dual-driven model of chips and solutions aims to provide cost-effective and reliable driving assistance technologies for the automotive industry [4]

Group 1 - The 2025 IAA Mobility exhibition in Munich showcased Hezhima Intelligent's advanced chip products and solutions for integrated cockpit and high-level assisted driving [1] - Hezhima Intelligent launched the "Safe Intelligent Base" solution, centered around the Wudang C1200 family of cross-domain integration chips, aimed at providing seamless upgrades for automotive companies from entry-level to flagship models [1] - The solution is designed to significantly reduce development costs, shorten development cycles, and enhance product competitiveness for automotive manufacturers [1] Group 2 - Hezhima Intelligent has established partnerships with leading companies, including Aptiv, to commercialize the Wudang C1296 chip-based integrated cockpit solution, which is expected to reach mass production by the end of 2025 [2] - The Huashan A1000 family of chips has already been integrated into multiple models from brands such as Geely, Lynk & Co, and Dongfeng [2]

Group 1 - The 2025 IAA Mobility exhibition in Munich showcased Black Sesame Intelligence's various chips and solutions, focusing on AI-driven smart vehicles [1] - Black Sesame introduced the "Safe Intelligent Foundation" solution, which utilizes the Wudang C1200 family of cross-domain integration chips to address safety and cost challenges in automotive cross-domain integration [1][2] - The Wudang C1200 family includes the C1236, the first domestic SoC chip supporting navigation-assisted driving, and the C1296, the industry's first multi-domain integration computing chip [2] Group 2 - The Huashan A2000 chip sample was a highlight, integrating CPU and NPU functionalities, and is designed to support advanced driver assistance systems and robotics [1][2] - The production results of the Huashan A1000 family have been successfully implemented in multiple vehicle models, including Geely Galaxy E8 and Dongfeng Yipai eπ007 [2] - Black Sesame aims to drive innovation and global collaboration in the automotive industry by providing cost-effective and reliable advanced driving assistance technology [2]

Core Insights - The 2025 IAA Mobility in Munich showcased the advancements in smart automotive technology, with Hezhima Intelligent presenting its capabilities in AI-driven smart vehicles [1] Group 1: Product Innovations - Hezhima Intelligent introduced its "Safe Intelligent Base" solution, addressing safety and cost challenges in cross-domain integration for automotive companies [3] - The Wudang C1200 family of chips serves as the core of the "Safe Intelligent Base," enabling seamless upgrades from entry-level to flagship models [3] - The Huashan A2000 chip family was highlighted for its high performance and efficiency, integrating multiple processing units for advanced driving assistance [4] Group 2: Market Collaborations - Hezhima Intelligent is collaborating with various partners, including the team led by Liu Sheng from Wuhan University, to develop humanoid robots and dexterous hands [4] - The Wudang C1200 family, particularly the C1236 and C1296 chips, has gained market recognition, with multiple automotive companies developing cross-domain solutions based on these chips [5] - The C1296 chip is set to be the first mass-produced chip platform for integrated cockpit and driving assistance solutions, with plans for production by the end of 2025 [5] Group 3: Commercialization and Ecosystem - The Huashan A1000 family has demonstrated successful commercialization, with chips already integrated into several vehicle models, including those from Geely and Dongfeng [6] - Hezhima Intelligent's dual approach of "chips + solutions" aims to provide cost-effective and reliable driving assistance technologies [6] - The company emphasizes innovation and open collaboration to enhance the automotive industry's auxiliary driving technology [6]

Core Viewpoint - Nvidia's new Jetson Thor chip significantly enhances the computational power and data processing capabilities for robotics, enabling more complex tasks to be performed directly on the device rather than relying on cloud processing [3][4]. Group 1: Nvidia's Jetson Thor Chip - The Jetson Thor chip, based on the Blackwell architecture, offers a peak performance of 2070 TFLOPS at FP4 precision, representing a 7.5 times improvement over the previous Orin chip and a 3.5 times increase in energy efficiency [4]. - The chip allows robots to handle high-resolution, high-frequency sensor inputs directly, potentially shifting many tasks from cloud processing back to local execution [5]. Group 2: Robotics Deployment Models - Current robotics typically utilize a hybrid deployment model combining cloud and edge processing, with edge systems focusing on real-time tasks and cloud systems handling more complex reasoning tasks [4]. - The reliance on cloud processing introduces latency issues that can affect the safety and feasibility of high-frequency tasks, such as rapid decision-making and continuous grasping [4]. Group 3: Competitive Landscape - Domestic companies are also developing their own solutions, such as Diguo Robotics' RDK S100 development kit and Hezhima Intelligent's chips for humanoid robots, which focus on real-time control and multi-modal data processing [6][7]. - The advantages of domestic chips include higher cost-effectiveness and tailored services that cater to local market needs, providing differentiation in scene optimization [7].

Group 1 - Nvidia has launched the Jetson Thor robot computing platform, which is now available for developers at a price of $3,499 [2] - The Thor chip, based on the Blackwell architecture, offers a peak computing power of 2070 TFLOPS at FP4 precision, representing a 7.5 times improvement over the previous Orin chip and a 3.5 times increase in energy efficiency [2] - The current deployment model for robots typically combines cloud and edge computing, with a need for significant computational power for more generalized robot models [2][3] Group 2 - The Thor chip enhances data processing capabilities and interface bandwidth, allowing robots to handle high-resolution, high-frequency sensor inputs locally, which could reduce reliance on cloud processing [3] - Nvidia aims to establish foundational infrastructure in the robotics sector, similar to its strategy before the AIGC boom, to influence industry standards ahead of a potential market explosion [3] - Domestic companies in China are also developing their own solutions, such as the RDK S100 development kit and various chips for different robotic functions, highlighting the competitive landscape [4] Group 3 - The advantages of domestic chips include higher cost-effectiveness and tailored services, which may provide differentiation in market optimization [4] - Despite Nvidia's advancements, the fragmented nature of robot applications presents opportunities for competitors to carve out niches in low-power chips and specialized scenarios [4]

Core Insights - Nvidia has launched the Jetson Thor robot computing platform, which significantly enhances computational power and data processing capabilities for robots, allowing them to run large-scale models directly on-device [1] - The new Thor chip offers a peak performance of 2070 TFLOPS, a 7.5 times improvement over the previous Orin chip, with a 3.5 times increase in energy efficiency [1] - The robotics industry is characterized by fragmented applications, with cost pressures and the need for long-term scenario validation, presenting opportunities for competitors in low-power chips and niche markets [5] Group 1: Nvidia's Technological Advancements - The Jetson Thor chip is now available for developers at a price of $3,499, marking a significant step in robot intelligence [1] - Thor's architecture allows for real-time processing of high-resolution, high-frequency sensor inputs, potentially reducing reliance on cloud processing for complex tasks [2] - The chip's advancements may accelerate the deployment of robots in high-frequency, complex interaction scenarios [2] Group 2: Industry Dynamics and Competitor Landscape - The robotics sector employs a hybrid deployment model of cloud and edge computing, with challenges such as latency affecting real-time applications [4] - Competitors are exploring various architectures, such as heterogeneous designs, to balance inference and real-time control in robotics [6] - Domestic chip manufacturers are leveraging cost advantages and tailored services to compete with Nvidia, suggesting that the latter's technological lead may not guarantee market dominance [8]