9月5日，在2025世界智能产业博览会上，中科曙光协同AI芯片、AI整机、大模型等20多家产业链上下 游企业，共同发布国内首个AI计算开放架构，并宣布全面开放"DeepAI深算智能引擎"，涵盖GPU异构 计算工具套件、人工智能基础软件栈OpenDAS，以及人工智能应用管理平台DAP。 中科曙光总裁助理、智能计算产品事业部总经理杜夏威对观察者网表示，"表面上看，这是一场关于工 具链的开放；但在更深层的逻辑中，它折射出中国AI算力生态的走向：从追赶、模仿，到开始尝试制 定自己的规则"。 CUDA之外的突围 在AI产业的技术版图里，英伟达的CUDA体系长期像一堵高墙。它以软硬件深度绑定的方式，锁定了全 球开发生态。 换言之，英伟达GPU之所以受欢迎，并不仅是因为硬件的性能高，同时也因为CUDA软件系统生态的完 善，让开发者可以更高效的发挥GPU本身的性能，降低使用成本。 对大多数国产厂商而言，要么花费巨大成本适配CUDA，要么在某些应用场景中"掉链子"。开发者更是 常常抱怨——想跑通一个模型，需要自己写一堆复杂脚本，还可能面临版本兼容问题。 "DeepAI深算智能引擎"的出现，意在改变这种局面。GPU异构计算工具套 ...

Core Viewpoint - Zhongke Shuguang's performance in the first half of 2025 exceeded expectations, with operating revenue of 5.85 billion yuan, a year-on-year increase of 2.41%, and a net profit attributable to shareholders of 729 million yuan, a significant year-on-year growth of 29.39% [1][4] Financial Performance - Operating revenue reached 5.85 billion yuan, reflecting a year-on-year growth of 2.41% [4] - Net profit attributable to shareholders was 729 million yuan, marking a year-on-year increase of 29.39% [4] - Deducting non-recurring gains and losses, net profit grew by 55.20% to 569 million yuan [1][4] - The net cash flow from operating activities was -1.381 billion yuan, worsening by 447 million yuan compared to the same period last year, primarily due to increased procurement payments and the purchase of large bank time deposits [1][4] Business Development - The company's profitability significantly improved, with the gross margin of high-end computer products reaching 15.33% [1][2] - IT equipment business revenue accounted for over 90%, with high-end computer products generating 4.559 billion yuan, representing 78.07% of total revenue [2] - The company has completed a full industry chain layout covering general computers, liquid cooling hardware, distributed storage, and network security, creating strong synergy effects [1][2] Strategic Initiatives - A major asset restructuring has been initiated with Haiguang Information, aiming to achieve complementary advantages in chip design and data center infrastructure [3] - The strategic integration is expected to accelerate the construction of the computing power industry ecosystem and enhance competitiveness against international leaders [3] - The company is increasing R&D investment in intelligent computing, advanced storage, and green computing, promoting technological innovation and product upgrades [3]

Core Insights - The article highlights that Zhongke Shuguang's DeepAI intelligent engine has successfully passed a joint evaluation by the China Academy of Information and Communications Technology, Tair Laboratory, and the Foundry Plan, becoming the first end-to-end AI acceleration suite to receive this certification in the industry [1] - The DeepAI engine's DAP (AI Application Platform) is recognized for its comprehensive functionality, business security, technological advancement, and operational stability, positioning it as a leader in the industry [1] Group 1: Product Features - The DAP platform adheres to the "out-of-the-box, secure and controllable" philosophy, integrating deeply with domestic GPU acceleration cards to provide a full-stack solution for large model application development [1] - Key capabilities of the DAP platform include dialogue services, RAG knowledge base, and Agent intelligence, along with enterprise-level features such as SSO, fine-grained permissions, and comprehensive auditing [1] Group 2: Market Applications - The DAP platform has been successfully deployed in various clients, including Bank of China, Beijing Aerospace General Hospital, Nanjing University of Aeronautics and Astronautics, and Shenzhen Discipline Inspection Commission, demonstrating the superior support capabilities of domestic AI computing hardware and software for complex large model tasks [3] - The platform addresses challenges in IT operations management by providing a cost-effective and efficient solution for knowledge management, particularly in industries like finance and healthcare where the quality of product requirement documents (PRD) is critical [2][3] Group 3: Future Developments - Following the certification from the China Academy of Information and Communications Technology, the company plans to continuously iterate on the DeepAI intelligent engine, focusing on enhancing product functionality in line with actual scene requirements and integrating deeply with client datasets and business processes [3] - The company aims to collaborate with more ecosystem partners to further enhance the competitive edge of domestic AI computing hardware and software [3]