9 月 29 日晚，科大讯飞发布《 2025 年度向特定对象发行 A 股股票预案（修订稿）》，拟 变更募集资金用途。 【导读】科大讯飞大幅调整定增方案 中国基金报记者 卢鸰 值得注意的是，在 8 月 22 日发布定增预案之后， 8 月 26 日国务院公开发布了《关于深入 实施 " 人工智能 +" 行动的意见》，明确 " 强化智能算力统筹；支持人工智能芯片攻坚创新 与使能软件生态培育，加快超大规模智算集群技术突破和工程落地 "，" 加强智能算力互联互 通和供需匹配，创新智能算力基础设施运营模式 " 。 科大讯飞称，国家高度重视自主可控大模型算力建设，租用算力模式在国家政策支持下更具 合理性。 基于国产算力进行训练和推理的效率与使用英伟达算力存在较大差距，科大讯飞通过与华 为、中科海光、寒武纪等国产算力厂商建立长期紧密联合攻关协作机制，深度使用国产算力 集群，发现并解决了多个国产算力底层软硬件的隐藏问题。 | 序号 | 募集资金投向 | 投资总额 | 拟使用募集资金 | | --- | --- | --- | --- | | | 星火教育大模型及典型产品 | 112,571.00 | 80,000.00 | | ...

【导读】科大讯飞大幅调整定增方案 中国基金报记者 卢鸰 9月29日晚，科大讯飞发布《2025年度向特定对象发行A股股票预案（修订稿）》，拟变更募集资金用 途。 | 序号 | 募集资金投向 | 投资总额 | [拟使用募集资金 | | --- | --- | --- | --- | | | 星火教育大模型及典型产品 | 112,571.00 | 80,000.00 | | 2 | 算力平台 | 240,000.00 | 240,000.00 | | 3 | 补充流动资金 | 80,000.00 | 80,000.00 | | | 合计 | 432,571.00 | 400,000.00 | 相对于8月22日发布的定增预案，科大讯飞本次修改后的方案中，40亿元的定增募资总额不变，星火教 育大模型及典型产品项目8亿元的募资金额也不变；但用于补充流动资金的金额由32亿元下调为8亿元， 另外新增募集资金24亿元用于算力平台项目，该项目募资金额约占本次融资规模的60%。 | 序号 | 募集资金投向 | 投资总额 | 拟使用募集资金 | | --- | --- | --- | --- | | | 星火教育大模型及典型产品 | ...

相对于8月22日发布的定增预案，科大讯飞本次修改后的方案中，40亿元的定增募资总额不变，星火教育大模型及典型产品项目8亿元的募资金额也不变； 但用于补充流动资金的金额由32亿元下调为8亿元，另外新增募集资金24亿元用于算力平台项目，该项目募资金额约占本次融资规模的60%。 【导读】科大讯飞大幅调整定增方案 中国基金报记者 卢鸰 9月29日晚，科大讯飞发布《2025年度向特定对象发行A股股票预案（修订稿）》，拟变更募集资金用途。 基于国产算力进行训练和推理的效率与使用英伟达算力存在较大差距，科大讯飞通过与华为、中科海光、寒武纪等国产算力厂商建立长期紧密联合攻关协 作机制，深度使用国产算力集群，发现并解决了多个国产算力底层软硬件的隐藏问题。 此外，《关于深入实施"人工智能+"行动的意见》还提出，要推动人工智能普惠共享。把人工智能作为造福人类的国际公共产品，打造平权、互信、多 元、共赢的人工智能能力建设开放生态。 对此，科大讯飞表示，公司算力平台项目的实施，将持续增强星火大模型的多语言支持能力，在为世界提供第二种选择的同时，将形成公司业务收入的新 增长点。 讯飞多语种大模型下一步将锚定海外市场落地需求，针对不同区域 ...

Core Viewpoint - The company, Keda Xunfei, has revised its fundraising plan for the issuance of A-shares, reallocating the use of raised funds to focus on computing power leasing in response to the rapid development of AI and large model technologies [1] Fundraising Plan - The total amount of funds to be raised remains unchanged, with the amount allocated for working capital reduced from 3.2 billion to 800 million [1] - 2.4 billion will be allocated for specialized expenditures related to computing power leasing [1] Computing Power Leasing Project - The company aims to implement a computing power platform project through a leasing model, expanding its computing power scale to support the development of large models and continuous algorithm innovation [1] - This project will enhance the base model upgrades for the Xunfei Spark cognitive large model and provide computing power support for the optimization of various models, including the Spark X1 model and multi-modal models [1] Strategic Implications - The expansion of computing resources is expected to improve the company's capability to provide multi-language large models for global clients, enhancing its technical strength for overseas AI business expansion [1] - The initiative is anticipated to create a new growth point for the company's business revenue [1]

Core Insights - Embodied Intelligence (EI) is a cutting-edge field at the intersection of artificial intelligence and robotics, aiming to integrate physical and cognitive capabilities in machines, enhancing their ability to interact with complex environments [1][7][21] - The market for embodied intelligence in China is projected to grow from 404.7 billion yuan in 2020 to 863.4 billion yuan in 2024, and is expected to reach 973.1 billion yuan by 2025 [1][7][22] Industry Overview - EI seeks to overcome limitations of traditional AI by enabling machines to perceive and interact with their environments like biological entities, thus paving the way for advancements in autonomous decision-making and execution capabilities [1][7] - The Chinese government has recognized the importance of EI, as highlighted in the 2025 government work report, which emphasizes the cultivation of future industries including EI [1][5] Current Industry Status - The EI market in China has seen rapid expansion, with significant advancements attributed to the emergence and iteration of large AI models [7][21] - The application of EI is primarily concentrated in robotics and autonomous vehicles, with the robotics sector accounting for over 55.6% of the market share, while autonomous driving holds 44.4% [9][11] Application Market - The robotics market in China is projected to reach 25 billion USD by the end of 2024, with an annual growth rate of 21% [11] - The autonomous driving market is expected to grow to 399.3 billion yuan in 2024, nearing 450 billion yuan in 2025, indicating a rising demand for EI technologies in this sector [13] Competitive Landscape - Major companies in the EI space include Tesla, Google, NVIDIA, and domestic firms like Yushutech, Xiaomi, and Huawei, all actively innovating and expanding their EI capabilities [15][19] - Notable developments include Huawei's collaboration with robotics firms to explore applications of their large models in humanoid robots, and advancements in AI-driven automation solutions by companies like Estun and Siasun [17][19] Future Trends - The EI industry is expected to experience a surge in technological innovation, particularly in perception, decision-making, and control, with significant improvements in 3D vision and tactile sensing technologies by 2025 [21][22] - The application scenarios for EI are diversifying, extending from industrial manufacturing to healthcare, services, and education, with collaborative robots increasingly replacing hazardous tasks in various sectors [22][23] - The establishment of a shared data ecosystem and open-source initiatives is anticipated to accelerate technological advancements and foster collaboration across the industry [24]

Group 1 - The National Development Bank (NDB) has issued infrastructure loans amounting to 369.1 billion yuan in the first quarter of this year, with the proportion of infrastructure loans steadily increasing within the total loan balance [1] - The NDB is focusing on supporting major infrastructure projects, including transportation, energy, water conservancy, smart cities, and smart logistics, to facilitate the coordinated development of traditional and new infrastructure [1] - The NDB is actively promoting the implementation of the "14th Five-Year Plan" by accelerating loan disbursement for 102 major projects, thereby enhancing effective investment and supporting high-quality development [2] Group 2 - The Dalian Jinzhou Bay International Airport, part of the "14th Five-Year Plan," is under construction using offshore reclamation methods, with the NDB providing early-stage loans for foundational work [2] - The Fujian Kemen Phase III Project has successfully completed its 168-hour test run, with two clean and efficient coal power units now operational, significantly enhancing energy supply and efficiency in Fujian [3] - The South-to-North Water Diversion Project's Xiong'an storage reservoir will supply 300 million cubic meters of water annually to the Xiong'an New Area, improving regional water resource allocation and ecological conditions [4] Group 3 - The NDB is prioritizing the development of new infrastructure, including supercomputing, cloud computing, and artificial intelligence platforms, to enhance the layout and functionality of new infrastructure [5] - The Qingdao branch of the NDB is providing financial support for the Jiaozhou Smart City project, focusing on smart applications that improve citizen life [5] - Automated Guided Vehicles (AGVs) are significantly increasing efficiency at the Ganqimaodu port, with the NDB's Inner Mongolia branch supporting the financing of smart customs and AGV projects to enhance cross-border transport capabilities [6][7] Group 4 - The NDB is supporting iFLYTEK's AI innovation projects through dedicated research loans, facilitating advancements in various sectors such as education, healthcare, and finance [7]

Market Overview - The Chinese multimodal large model market reached CNY 9.09 billion in 2023 and is projected to grow to CNY 66.23 billion by 2028, with a compound annual growth rate (CAGR) of 48.76%[24] - The rapid growth is driven by continuous technological innovation and strong industry demand[24] Industry Insights - Major players in the Chinese multimodal large model sector include Baidu, Alibaba, Tencent, and SenseTime, with significant advancements in model capabilities[31] - The application of multimodal models spans various sectors, with digital humans accounting for 24% of applications, followed by gaming and advertising at 13% each[33] Technological Development - The evolution of multimodal models has transitioned from task-specific to more general architectures, enhancing efficiency and flexibility[22] - Key components of multimodal models include modality encoders, input projectors, large model backbones, output projectors, and modality generators, which work together to process and generate diverse data types[9][12][14][15][16] Training and Evaluation - The training process for multimodal models typically involves two phases: pre-training with multimodal data and instruction fine-tuning to enhance user interaction capabilities[34] - Evaluation of generation capabilities focuses on aspects such as semantic understanding, coherence, and the ability to handle complex scenes[40][41] Future Trends - Future advancements in multimodal models will focus on improving generation consistency, contextual learning, and complex reasoning capabilities[46] - Addressing challenges like multimodal hallucination and enhancing model robustness will be critical for practical applications in fields such as healthcare and autonomous driving[46][50]