本周硬科技领域投融资重要消息包括：央行：将科技创新和技术改造贷款额度增加至1.2万亿元；上 海：到2027年高级别自动驾驶应用场景实现规模化落地；核聚变公司超磁新能获数亿元天使轮融资。 》》政策 央行：将科技创新和技术改造贷款额度从8000亿元增加至1.2万亿元 央行副行长邹澜表示，增加科技创新和技术改造再贷款额度，并扩大支持范围，将科技创新和技术改造 贷款额度从8,000亿元增加4,000亿元至1.2万亿元，并将研发投资水平较高的民营中小企业纳入支持领 域。 工信部印发《推动工业互联网平台高质量发展行动方案（2026—2028年）》 工业和信息化部印发《推动工业互联网平台高质量发展行动方案（2026—2028年）》。方案指出，到 2028年，工业互联网平台高质量发展取得积极成效，"专业型+行业型+协作型"多层次平台体系持续壮 大，具有一定影响力的平台超450家；平台的要素资源连接能力大幅增强，重点平台的数据增值、模型 沉淀和人工智能开发应用能力显著提升，工业设备连接数突破1.2亿台（套）；平台普及率达到55%以 上，基本建成泛在互联、数智融合、深度协同、开源开放的新一代工业互联网平台生态。 上海：到2027 ...

Group 1 - The People's Bank of China and the National Financial Regulatory Administration have adjusted the minimum down payment ratio for commercial housing loans to no less than 30% [1] - Local branches of the People's Bank of China and the National Financial Regulatory Administration can set lower limits for the down payment ratio based on local government policies [1] Group 2 - The Eastern Theater Command monitored and guarded the passage of U.S. naval vessels through the Taiwan Strait, maintaining a high level of readiness to safeguard national sovereignty and regional stability [2] - The Southern Theater Command conducted normal drone flight training near the Dongsha Islands, asserting the legality of the operation [2] Group 3 - The Ministry of Industry and Information Technology has revised the management measures for the cultivation of high-quality small and medium-sized enterprises, now including technology-based SMEs in the cultivation scope [3] - The new measures aim to enhance the quality of cultivated enterprises and improve the standards for recognizing specialized and innovative SMEs [3] Group 4 - The Boao Forum for Asia 2026 Annual Conference is scheduled to be held from March 24 to 27, with the theme "Shaping a Shared Future: New Situations, New Opportunities, New Cooperation" [4][6] Group 5 - The China Passenger Car Association predicts a significant optimization in the structure of automotive exports by 2025, with a focus on enhancing core technology and local production [7] - The association emphasizes the need for companies to strengthen collaboration and address external risks to promote high-quality development in automotive exports [7] Group 6 - Star River Dynamics announced the failure of the first flight test of the Vesta II rocket, with ongoing investigations to determine the cause [12] - Novo Nordisk's new CEO emphasized that all business expansions must focus on the core needs of diabetes or obesity patients, marking a strategic shift away from previous diversification attempts [13] Group 7 - Avita Technology clarified that the recent traffic accident involving 15 vehicles in Nanning was due to driver error, with the vehicle's assisted driving feature not activated at the time of the incident [14] - The company highlighted the importance of enhancing safety in automated driving technology and user education [15] Group 8 - Aiko Micro has initiated the A-share listing guidance process, with no single shareholder holding more than 30% of the company [17] - Zhuhai Wanda Commercial Management has appointed a new CEO, reflecting adjustments in management in response to market pressures [18] Group 9 - LoBo Quick Run has launched a fully autonomous driving service in Abu Dhabi, marking its first overseas public operation [18]

Core Viewpoint - Tesla's CEO Elon Musk announced that the design work for the AI5 chip used in the fully autonomous driving system is nearly complete, and the company is in the early stages of developing the next-generation AI6 chip [1] Group 1: Chip Development - The design cycle for future chip iterations is aimed to be nine months [1] - Following Musk's previous update on chip progress on November 24, the company's stock price surged nearly 7% [1] Group 2: Application of Chips - These chips support Tesla's autonomous driving features and the nascent robotics product line [1]

Core Insights - The total electricity consumption in China is projected to reach 10.3682 trillion kilowatt-hours by 2025, marking a significant milestone as it is the first time the country's electricity consumption surpasses 10 trillion kilowatt-hours. This figure is more than double the annual electricity consumption of the United States and exceeds the combined annual consumption of the EU, Russia, India, and Japan [1][9]. Group 1: Energy Structure - The shift towards electric vehicles is evident, with non-fossil energy sources expected to account for over 60% of installed capacity by 2025, indicating a significant transition towards cleaner energy. One-third of the total electricity consumed will be green energy [3][11]. - In 2025, China's automobile production and sales are projected to reach historical highs of 34.531 million and 34.40 million units, respectively, with a year-on-year growth of 10.4% and 9.4%. New energy vehicle sales are expected to exceed 16 million units, making up over 50% of new car sales [3][11]. Group 2: Technological Transformation - The industrial sector, which is the largest consumer of electricity, is undergoing a technological revolution characterized by advancements in AI, quantum technology, and automation. In the past year, the electricity consumption of the secondary industry reached 66,366 billion kilowatt-hours, accounting for approximately 64% of total consumption [5][13]. - The computer, communication, and other electronic equipment manufacturing sectors are expected to consume over 50 billion kilowatt-hours of electricity in 2025, surpassing traditional industries like steel and chemicals [5][13]. Group 3: Quality of Life - The electricity consumption of the tertiary industry is projected to reach 19,942 billion kilowatt-hours by 2025, with a year-on-year growth of 8.2%. Urban and rural residential electricity consumption is expected to reach 15,880 billion kilowatt-hours, growing by 6.3% [6][14]. - The increase in electricity consumption in the service sector indicates a vibrant social and market environment, with the electric vehicle sector driving a nearly 50% increase in charging service electricity consumption [6][15].

Core Viewpoint - Elon Musk has decided to phase out the one-time purchase option for Tesla's Full Self-Driving (FSD) by February 14, 2026, favoring a Software as a Service (SaaS) model instead [1]. Pricing and Market Strategy - In the U.S., the one-time purchase price for FSD is $8,000, while the monthly subscription price will decrease to $99, making the subscription equivalent to a purchase over 81 months. In China, the buyout price is approximately 64,000 RMB, with the subscription model expected to lower the barrier for adoption and increase subscription rates [2]. - Reports indicate that FSD has received "partial approval" in China, with full approval anticipated around February or March 2026. The monthly subscription fee in China is projected to be between 499 and 699 RMB [2]. Technological Developments - Tesla's FSD continues to utilize an end-to-end Variational Autoencoder (VA) architecture, with ongoing optimizations. The focus is on user acceptance and engineering improvements, indicating a challenging period ahead for autonomous driving [3]. - Recent advancements in FSD include the development of 3D Gaussian closed-loop simulation capabilities, which are expected to enhance action optimization [2]. Future Outlook - The company is also making strides with its Optimus V3 project, which is anticipated to be a transformative technology, potentially overshadowing Tesla's automotive legacy [3].

IDEA团队 投稿 量子位 | 公众号 QbitAI 3D模型的实例分割一直受限于稀缺的训练数据与高昂的标注成本，训练效果有待提升。 近年来使用成熟海量的2D实例分割数据来辅助实现3D实例分割成为了一个极具潜力的研究方向，但实现思路不尽相同。 近日IDEA计算机视觉与机器人研究中心 （CVR，Computer Vision and Robotics） 的张磊团队提出了一种名为OVSeg3R的开集3D实例分 割学习新范式。 与传统训练范式相比， OVSeg3R无需对输入点云进行人工后处理 ，也不需要高成本的物体3D掩码人工标注，大幅降低了3D实例分割的训练 成本，有望让3D实例分割从"只能识别已知类别"的闭集模式，到"可识别未知类别"的开集模式实现商业落地，广泛应用于 自动驾驶、智能家 居、机器人导航等需要精准3D场景理解的领域 。 △ 视频1：从视频到开集空间感知工作流 △ 视频2：OVSeg3R与闭集SoTA模型 （SegDINO3D） 感知结果对比可视化 一、3D感知的"卡脖子"难题 3D实例分割，就是让计算机像人眼一样，看懂三维空间里的每一个物体，以及精准勾勒出每个物体的边界范围。 这项技术是自动驾驶、 ...

炒股就看金麒麟分析师研报，权威，专业，及时，全面，助您挖掘潜力主题机会！ 来源：中信证券研究 PART 1 聚焦资源和传统制造定价权提升 建议资源和传统制造定价权提升逻辑打底，增配非银金融，兼顾反共识品种降低组合波动。 图片来源：摄图网 继续聚焦资源和传统制造 定价权的提升 裘翔 张铭楷 陈峰 高玉森 陈泽平 刘春彤 策略聚焦 从当前的量价和风偏指标来看，预计主题、小票轮动的震荡上行格局可能还会延续到两会前后，直至内 需预期的上修，市场才会回到基本面驱动状态。站在全年维度，真正具备持续入场体量的配置型资金今 年对权益组合降低波动的诉求是更高的，需要考虑的是这部分可持续的"大钱"会持续流向的方向。配置 上，建议资源和传统制造定价权提升逻辑打底，增配非银金融，兼顾反共识品种降低组合波动。 风险因素：中美科技、贸易、金融领域摩擦加剧；国内政策力度、实施效果或经济复苏不及预期；海内 外宏观流动性超预期收紧；俄乌、中东地区冲突进一步升级；我国房地产库存消化不及预期。 关注情绪上行的主题轮动板块 刘易 田鹏 王涛 王子昂 白弘伟 主题聚焦 从催化因素和时序上看，国产算力、AI应用、合成生物、人形机器人、自动驾驶的主题热情 ...

Core Viewpoint - The competition for the title of "Automobile Capital" in China is intensifying, with Chongqing set to achieve a record automotive production of 2.788 million vehicles in 2025, marking a 9.7% increase, and a significant rise in new energy vehicle (NEV) production to 1.296 million units, up 36% [6][8]. Group 1: Historical Context and Challenges - Chongqing aimed to become "China's Detroit" in 2013, reaching a peak production of 3.156 million vehicles in 2016, but faced a decline starting in 2017, with production dropping to 1.383 million vehicles by 2019 [8][9]. - The decline was attributed to a mismatch between Chongqing's focus on mid-to-low-end vehicles and the market's shift towards mid-to-high-end preferences, alongside production capacity issues [8][9]. - The rise of NEVs began to disrupt the traditional automotive landscape, with national sales surpassing 100,000 units within three years after first exceeding 10,000 units in 2015 [9]. Group 2: Strategic Shifts and Collaborations - Changan Automobile, a leading player in Chongqing, announced a plan in 2017 to phase out traditional fuel vehicles by 2025, but initially struggled to keep pace with national NEV growth [9][10]. - A turning point occurred in 2021 when Changan partnered with Huawei and CATL to launch new high-end NEV brands, resulting in a significant increase in sales, with total vehicle sales reaching 2.913 million units and NEV sales up 51.1% to 1.11 million units [10][14]. - The collaboration with Huawei has been pivotal for both Changan and Seres, with Seres becoming the first company to benefit from Huawei's "Smart Selection" model, leading to increased competitiveness in the NEV market [13][14]. Group 3: Future Prospects and Competitive Landscape - The automotive industry is entering a new competitive phase, with Chongqing positioned to lead in NEVs and smart driving technologies, especially after receiving approval for L3-level autonomous driving vehicles [17][19]. - Chongqing's strategic focus on becoming a "Smart Connected New Energy Vehicle Capital" by 2024 aims to leverage its unique geographical features for testing smart vehicles [19][20]. - However, challenges remain, including a lack of competitiveness in AI technology and talent shortages, which could hinder Chongqing's ability to maintain its leading position in the next industrial competition [21][22].

Core Insights - The automotive cloud industry is rapidly evolving, driven by technological advancements, user demands, and market competition, with a significant focus on "smart driving equality" [1][4] - The integration of cloud computing with smart mobility is expected to create substantial growth opportunities, with the market projected to reach 12.21 billion yuan in 2024 and 14.04 billion yuan in 2025, reflecting a 15% year-on-year growth [1][8] Industry Overview - Automotive cloud serves as a digital foundation for the entire smart vehicle industry chain, leveraging cloud computing to enhance the digital transformation across R&D, production, sales, and usage [2][3] - The industry is characterized by a four-dimensional collaborative ecosystem involving vehicles, cloud, roads, and people, facilitating a comprehensive upgrade of the automotive sector [2] Development Background - The acceleration of automotive intelligence is evident, particularly in the new energy vehicle sector, where the installation rate of L2-level and above assisted driving functions is expected to reach 87% by October 2025 [4][5] - The Chinese government is actively promoting the "vehicle-road-cloud integration" strategy, which aims to enhance the safety, energy efficiency, and comfort of intelligent connected vehicle systems [5][6] Market Status - The Chinese automotive cloud market is experiencing robust growth, with a projected market size of 12.21 billion yuan in 2024 and 14.04 billion yuan in 2025, driven by AI models, smart vehicle intelligence, and supportive policies [8][10] - Public cloud services are expected to account for 65.3% of the market in 2024, while private cloud services will represent 34.7%, indicating a diverse deployment strategy among automotive companies [8][9] Customer Distribution - The customer base for automotive cloud services is extensive, covering key segments of the smart vehicle industry, including OEMs, parts suppliers, autonomous driving technology providers, and mobility service suppliers [9] - In 2024, OEMs and parts suppliers are expected to represent 43.2% and 27.6% of the customer distribution, respectively, highlighting their reliance on cloud services for core applications [9] Market Landscape - The market is dominated by a multi-supplier strategy, with leading players maintaining significant market shares due to their resource availability and technological advantages [10] - In the public cloud infrastructure sector, the top five service providers are expected to hold a combined market share of 86.9%, while the top five in the private cloud sector will account for 60% [10] Future Trends - The automotive cloud industry is anticipated to continue its growth trajectory, with expectations of surpassing 20 billion yuan by 2028, driven by the increasing demand for smart and connected vehicles [10] - Global expansion is expected as Chinese automotive companies increase their international presence, leading to a rise in demand for localized cloud services [10] - The industry will likely shift towards a hybrid architecture, optimizing computing resources and ensuring compliance with local data regulations [10]

除了调查特斯拉在启用FSD功能时违反交通安全法规的行为外，NHTSA还在审查该公司车辆检测并妥 善应对阳光眩光、雾霾和其他低能见度环境的能力。在发生多起事故(其中一起造成人员死亡)后，该监 管机构于2024年10月启动了这项调查。 NHTSA上个月向特斯拉发出信息请求，要求该公司提供与所指控缺陷相关的消费者投诉、现场报告、 财产损失索赔、仲裁程序或诉讼的数量。特斯拉于1月12日告知监管机构，需要更多时间来确定可能与 调查相关的交通违章报告。 据特斯拉向NHTSA提交的文件显示，截至今日，仍有8313条记录需要人工审核。该公司估计每天能够 处理约300条记录，并需要在2月23日前做出回应。 对于特斯拉而言，FSD的熟练程度日益重要。该公司首席执行官埃隆·马斯克寄希望于该系统能够提振 市场需求，此前特斯拉的年度汽车交付量连续下滑。尽管马斯克经常吹捧特斯拉自动驾驶功能的有效 性，但加利福尼亚州指责该公司夸大了其汽车的性能，并威胁要暂停销售30天，该禁令可能于今年年初 生效。 智通财经APP获悉，美国监管机构给予特斯拉(TSLA.US)五周的延期，以开始调查该公司在启用全自动 驾驶系统(FSD)时，其车辆违反交通法规 ...