其中，江原D20 AI加速卡是国内首款实现设计、制造、封装、境内流片量产的云端AI推理卡，供应链 自主可控，搭载256GB显存，INT8算力峰值达320TOPS，支持FP32/FP16/BF16/INT8多精度混合计算， 为大模型运行提供坚实硬件支撑。 单机性能层面实现关键突破：单卡机型可稳定流畅运行32B参数级大模型，四卡机型凭借多芯片协同架 构与品高自研4D并行调度策略，可单机承载Qwen 235B、DeepSeek 685B等千亿级大模型运行，打破桌 面级工作站超大模型部署的算力限制。 配合一体式液冷散热架构，AI Station系列实现"高性能+低噪音+长稳定"的三重优势：热传导效率较传 统风冷提升40%，可避免高负载下的算力降频，保障7×24小时连续运行；运行噪音低至28dB（接近图 书馆环境音），适配办公、实验室等安静场景。 通过液冷系统的小型化集成，四卡机型仍可保持紧凑机身设计，兼顾算力密度与空间适配性。场景适配 方面，全系列机型深度支持RAG（检索增强生成）、工作流自动化、智能体部署等前沿AI应用场景， 依托本地算力部署模式构建数据私域安全屏障，有效规避核心数据外泄风险，契合政企客户数据合规需 ...

证券日报网讯12月25日，宝信软件（600845）在互动平台回答投资者提问时表示，公司天行PLC已突破 大型PLC产品多项关键核心技术，首发天行系列全栈自主可控T3、T4PLC产品，实现从操作系统、到运 行时软件、再到编程环境的全栈自主可控，已在多个行业形成国产化替代示范应用；宝之云华北基地的 建设正常推进中。 ...

Core Viewpoint - China Railway Signal and Communication Group Co., Ltd. (通号集团) plans to divest its entire stake in Kaifa Electric (凯发电气), amounting to 36.6869 million shares, which represents 11.63% of the total share capital after excluding shares in the company's repurchase account [1] Group 1: Shareholding Changes - The divestment will result in 通号集团 no longer holding any shares in Kaifa Electric, where it was the second-largest shareholder as of September 30 [1] - 通号集团 initially acquired shares in Kaifa Electric on December 26, 2018, through a share transfer agreement, purchasing 14.2263 million shares at a price of 6.45 yuan per share, totaling approximately 91.76 million yuan [2] - After subsequent investments in convertible bonds, 通号集团's total equity stake in Kaifa Electric was calculated to be 11.18% if all bonds were converted at the initial conversion price of 8.15 yuan per share [2] Group 2: Company Development Plans - Kaifa Electric recently announced a plan to raise 300 million yuan through a private placement to fund technology upgrades and the development of a power supply automation product platform [3] - The company is transitioning from traditional power supply automation systems to higher-value products, including intelligent power supply dispatch and smart monitoring systems, necessitating advanced technology [3] - The funds raised will support the establishment of an integrated production line for power supply automation products, enhancing product quality and reducing costs while aligning with national strategic goals for self-controlled technology [3]

Core Insights - Sichuan Bank aims to achieve a fully autonomous and controllable technology stack through the implementation of TencentOS, which is designed to address core challenges in the bank's digital transformation [1][5][8] Company Overview - Sichuan Bank was established on November 7, 2020, as the first provincial-level urban commercial bank in Sichuan Province, with an asset scale approaching 700 billion [1] - The bank has achieved a significant growth trajectory, targeting a scale of 1 trillion by 2024 [1] Project Timeline - The new generation information system project commenced in May 2022, with a comprehensive technology stack being built by September 2023, including 13 project groups [2] - The project aims for full autonomy and control over 93 business systems by May 2025 [2] Technical Requirements - The core requirement for the new system is full-stack autonomy from chip to system, focusing on security, ecosystem compatibility, and performance [4] - Key considerations include stable operation, compatibility with mainstream CPUs, and optimization for financial scenarios [4] Operating System Selection - TencentOS was chosen for its robust product capabilities, having been deployed over 10 million times and proven in critical business scenarios [5] - The operating system meets Sichuan Bank's requirements for security, stability, and performance optimization [5][6] Business Value - The cloud platform for Sichuan Bank is built on over 1,300 physical servers using TencentOS, supporting 93 application systems and over 60,000 containers [6] - This setup ensures a fully controllable supply chain from hardware to application systems [6] Future Outlook - Sichuan Bank plans to deepen its digital transformation by focusing on AI applications in risk control, customer service, and operational efficiency [8] - Future work with TencentOS will include performance optimization for high-frequency trading, enhancing ecosystem compatibility, and exploring AI integration [8]

Core Viewpoint - Kaifa Electric (300407.SZ) plans to issue shares to specific investors through a simplified procedure, aiming to raise a total of 300 million yuan for technology upgrades and project development [1] Group 1: Fundraising Details - The company intends to issue shares to no more than 35 specific investors [1] - The total amount to be raised is 300 million yuan, with the net proceeds after issuance costs allocated entirely to technology upgrades and industrialization projects for self-controlled power supply automation products [1] Group 2: Project Focus - The raised funds will also support the development of a large-scale model for power supply in rail transit [1]

Core Viewpoint - Kefa Electric (300407.SZ) plans to issue shares to specific investors through a simplified procedure, aiming to raise a total of 300 million yuan for technology upgrades and project development [1] Group 1: Fundraising Details - The company intends to issue shares to no more than 35 specific investors [1] - The total amount to be raised is 300 million yuan, with the net proceeds after issuance costs allocated entirely to technology upgrades and industrialization projects [1] Group 2: Project Focus - The raised funds will be used for the upgrade of self-controlled power supply automation product technology [1] - Additionally, the funds will support the development of a large model research platform for rail transit power supply [1]

Group 1: Financial Performance - In 2024, the company achieved a revenue of 23.343 billion, marking an 18.79% year-on-year growth, and a net profit attributable to shareholders of 560 million [1] - The Q1 2025 report showed a revenue of 4.658 billion, a 27.74% increase year-on-year, with net profit and net profit excluding non-recurring items growing by 35.68% and 48.29% respectively [1] - The operating cash flow reached a historical high of 2.495 billion, growing over six times compared to the previous year, attributed to improved sales collection [4] Group 2: Business Strategy and Structure - The company has restructured its business model focusing on optimizing C-end, strengthening B-end, and selectively partnering with G-end clients, leading to improved revenue collection rates [5] - High-margin businesses such as smart education and medical AI have seen an increase in revenue share, maintaining an overall gross margin above 40% [5] - The company’s core business segments, including consumer, education, automotive, and medical, reported significant revenue growth, with smart education revenue reaching 7.229 billion, up 29.94% [5] Group 3: Technological Advancements - The company has established a fully autonomous and controllable AI technology system, integrating domestic computing power and self-developed algorithms, which is crucial for maintaining a competitive edge [6][7] - The Spark-X1 model, launched in April 2024, has shown significant improvements in various tasks, achieving performance levels comparable to leading international models despite having fewer parameters [9][10] - The company’s focus on self-developed foundational models is seen as essential for the application of large models across various industries, enhancing reliability and trust among clients [12][13] Group 4: Market Position and Future Outlook - The company is positioned as a leader in the domestic AI market, with the largest number of bids and bid amounts in 2024 across multiple sectors including communication, finance, and healthcare [10] - The integration of AI technology into smart hardware products has opened new growth avenues, with AI learning machines seeing sales growth exceeding 100% [14] - The company emphasizes the importance of productization to ensure the scalability of AI solutions, aiming to create replicable products that address specific market needs [17][18]