【环球网科技报道 记者 林梦雪】海洋作为覆盖地球 71% 表面积的蓝色疆域，是人类社会发展的重要支 撑。随着高性能计算与人工智能技术的深度渗透，海洋科学研究正经历颠覆性转型，从风暴潮灾害风险 预警到海洋卫星高质量数据集构建，从四维海洋监测技术升级到高时空分辨率海洋预报大模型落地，技 术创新正为海洋安全保障、资源开发与环境治理提供前所未有的赋能。 李津宇称，这种创新模式正与海洋数据预报领域深度契合，"目前海洋领域已有几十个专属大模型，未 来这些模型都将在 MaaS 场景中实现具体落地"。并行科技的 MaaS 平台已整合 DeepSeek、千问、智 谱、豆包等通用领先模型，以 AI Coding 技术为基础，为海洋预报研发工作提供辅助支持。随着行业预 报大模型的成熟稳定，平台将进一步进入更深层次的业务支撑体系。 在算力支撑方面，并行科技已构建起覆盖 45 家智算中心、15 家超算中心的强大算力网络，总调度能力 超 200 万 CPU 核心、5 万卡以上 GPU。不同于传统算力服务的 "一刀切" 模式，并行科技基于多年行业 经验，针对海洋环境模拟、资源勘探、数据预报等不同场景的特点，提供定制化算力方案。"海洋预报 ...

陈健自身便是算力的用户。1993年，陈健进入清华大学就读流体力学专业，流体力学风洞等实验恰恰需 要大量的高性能计算来通过精细建模逼近真实物理过程，陈健与算力领域的缘分也就此展开。 "计算机专业人士更关心的可能是我做得好不好，但用户更关心的其实是我用得爽不爽。并行科技发展 18年以来，整个业务逻辑就是用户用得爽不爽、是不是让计算更简单。"陈健认为，算力行业本质上是 服务业。 不是建好算力设施等待需求，而是为客户需求匹配合适的算力。"我们搭建算力服务平台，用户仅需五 分钟即可完成开户，提交计算任务。我们负责把程序有关的所有问题，包括选什么算力最合适、什么算 力性价比最高、应用在算力上跑的性能如何提高，所有这些事情全包了。用户只管用，我们叫'可用、 好用、降本'，这些都做到之后，用户为什么要离开我们？你是找不到理由的。"陈健表示。 算力只有用起来才有价值，并行科技做到了算力供给与需求之间的"翻译官"和"优化师"，破解行业普遍 存在的供需错配难题。数据显示，并行科技的算力服务已覆盖超1300家机构，包括600余家企业、400余 所高校、300余个科研机构，以高质量、高性能、高性价比的算力服务，推动各领域实现数据驱动、 ...

Core Insights - The demand for computing power has surged dramatically, with daily Token consumption increasing over 300 times in just over a year, driven by the needs of AI large models [2][6][7] - The company, Parallel Technology, has positioned itself as a leading provider of computing power services in China, focusing on simplifying computing for users rather than just selling hardware [2][3] Company Strategy - The core logic of Parallel Technology is to prioritize user experience, emphasizing that the essence of computing power services is value delivery rather than hardware accumulation [3][4] - The company has developed a service platform that allows users to complete tasks in just five minutes, handling all related issues to ensure a seamless experience [3][4] Market Position - Parallel Technology's services cover over 1,300 institutions, including more than 600 enterprises, 400 universities, and 300 research institutions, providing high-quality and cost-effective computing services [4] - The company has a market share exceeding 20% in the general supercomputing sector and has established partnerships with numerous computing centers [8] Financial Performance - In 2024, the company's revenue reached 655 million yuan, a year-on-year increase of 32.07%, with a net profit of 12.06 million yuan, marking a return to profitability [7] - For the first three quarters of 2025, revenue grew by 69.25% to 734 million yuan, with net profit increasing by 178.80% to 8.41 million yuan [7] Future Outlook - The computing power demand is expected to grow exponentially, with projections indicating a potential increase of 100,000 times by 2035, necessitating improvements in software and hardware collaboration [9][11] - The company aims to evolve from simplifying computing to simplifying models, indicating a strategic shift to meet future user needs [10][11]

Core Insights - The explosive growth in daily Token consumption, which has increased over 300 times in just over a year, highlights unprecedented demand in the IT industry driven by computing power [1][3] - The company, Parallel Technology, has positioned itself as a leading provider of computing power services in the A-share market, focusing on simplifying computing for users rather than merely selling hardware [1][4] Company Strategy - The core logic of Parallel Technology revolves around delivering value to users by ensuring that computing services are user-friendly and efficient, rather than just providing hardware [2][4] - The company has built a computing service platform that allows users to complete tasks in just five minutes, handling all related issues to enhance user experience [2][3] Market Demand - The demand for computing power has surged, with over 30 trillion daily Tokens consumed in China as of June, marking a growth of over 300 times in 18 months [3][4] - The rapid increase in demand is reflected in the performance of major AI model companies, with significant growth in Token usage reported by firms like ByteDance [3][4] Financial Performance - Parallel Technology reported a revenue of 655 million yuan in 2024, a year-on-year increase of 32.07%, and a net profit of 12.06 million yuan, marking a return to profitability [4][5] - In the first three quarters of 2025, the company achieved a revenue of 734 million yuan, a 69.25% increase year-on-year, with net profit rising by 178.80% [4][5] Customer Base and Growth - The company serves over 1,300 institutions, including more than 600 enterprises and 400 universities, establishing a solid foundation for growth [3][4] - A unique customer conversion chain has been developed, where users from educational and research backgrounds transition into industry clients, contributing to sustained growth [4][5] Industry Trends - The "AI+" initiative is expected to further activate computing power demand across various sectors, including traditional manufacturing and agriculture [5][6] - The company has achieved over 20% market share in the general supercomputing field and is expanding its computing resource pool through various collaborative efforts [5][6] Future Outlook - The computing power demand is projected to increase dramatically, with expectations of a tenfold rise by 2035, necessitating improvements in software, hardware, and algorithm efficiency [6][7] - The company aims to evolve its mission from simplifying computing to simplifying models, indicating a strategic shift to meet future demands [7]

◎记者 窦世平 "人工智能并非一项新技术，而是超级计算机上的新应用。"并行科技董事长陈健近日接受上海证券报记 者专访时，分享了他对人工智能的深刻理解。并行科技作为国内领先的算力运营服务商，深耕高性能计 算领域十八年，率先洞察到从传统超算向AI智算转型的趋势，通过打造"算力买手"模式，在激烈的市场 竞争中开辟出一条独特的发展路径。 面对这一变化，并行科技展现出深厚的技术积淀和快速响应能力。陈健介绍，公司在DeepSeek模型优 化上取得了显著进展，从2月到现在不断优化性能，将在线服务的Token成本从最初的130元降至4元以 下，离线模式更是可以降到1元以下。这种技术突破使得公司能够在保证服务质量的同时，大幅降低用 户使用成本，进一步刺激了市场需求。 公司的客户黏性也令人印象深刻。陈健分享了两组关键数据：一是客户复购率大于80%，客户留存极 好；二是客户的复购额从2018年到现在常年保持在100%。"这种高度的客户忠诚度，源于公司对用户需 求的深刻理解和服务能力的持续提升。" "算力服务的核心在于真正理解并满足客户需求。"陈健认为，一些算力供应商仅从供给端出发提供算力 资源，这种思维方式难以获得市场认可。正确的 ...

Core Insights - Zhengzhou Zhizhen Parallel Computing Technology Co., Ltd. has been established with a registered capital of 10 million yuan [1] - The company's business scope includes data processing and storage support services, computer system services, IoT technology services, and communication equipment manufacturing [1] - The company is jointly held by Parallel Technology and Zhizhen Technology (stock code: 003007) [1]

Group 1 - The report highlights the release of significant technologies and products at the 2025 Yunqi Conference, emphasizing the focus on high-quality and scarce targets in the AI industry chain on the Beijing Stock Exchange [2][11] - As of September 26, 2025, there are 25 listed companies in the AI-related industry chain on the Beijing Stock Exchange, with a total market capitalization of 97.089 billion yuan, covering areas such as chips, computing center infrastructure, and computing services [17][18] - Parallel Technology (839493.BJ) is identified as a leading provider of supercomputing cloud services and computing operation services, achieving a revenue of 458 million yuan in the first half of 2025, a year-on-year increase of 69.27% [19][24] Group 2 - The report indicates that five major industries on the Beijing Stock Exchange experienced an average decline in stock prices, with the high-end equipment industry down by 6.35% and the information technology industry down by 6.17% [3][25] - The average price-to-earnings (P/E) ratio for the high-end equipment industry fell to 40.7X, while the information technology industry saw its P/E ratio drop to 93.5X [25][28] - The technology new industry, comprising 156 companies, saw a median P/E ratio decrease from 51.7X to 48.9X, with a total market capitalization decline from 524.131 billion yuan to 494.938 billion yuan [39][40] Group 3 - The report notes that HaiDar (836699.BJ), a precision rail manufacturer, achieved a revenue of 182 million yuan in the first half of 2025, with a net profit of 33.0587 million yuan, reflecting a year-on-year growth of 3.79% [21][22] - The report emphasizes the importance of the AI computing server market, with Parallel Technology planning to procure AI computing servers and related equipment, indicating ongoing investment in infrastructure [59]

Core Viewpoint - The company, Bianhang Technology, announced plans to purchase AI computing servers and related equipment from Mianchun Technology and Zhejiang Tianji, with a total estimated amount not exceeding 113 million yuan [1] Group 1 - The total procurement amount for the AI computing servers and equipment is expected to be up to 113 million yuan [1]

北京并行科技股份有限公司 购买资产的公告 证券代码：839493 证券简称：并行科技 公告编号：2025-112 北京并行科技股份有限公司（以下简称"公司"）拟向绵存（浙江）科技有 限公司、浙江天极集成电路技术有限公司采购 AI 算力服务器及配套服务器主 机设备，并签署相关采购合同，采购合同总金额预计不超过人民币 11,338.98 万 元。 （二）是否构成重大资产重组 本次交易不构成重大资产重组。 公司购置算力设备形成自有算力资源池是向用户提供服务的经营采购行 为，是公司各项主营业务的必要组成部分，属于与公司日常经营相关的交易行 为。根据《上市公司重大资产重组管理办法》第二条、《北京证券交易所上市公 司重大资产重组审核规则》第八条规定，公司据经营所需购置相关算力设备资 产的交易活动为日常经营活动中购买资产的交易活动，不构成重大资产重组。 （三）是否构成关联交易 本次交易不构成关联交易。 （四）决策与审议程序 1、本次购买资产的决策与审议程序 本公司及董事会全体成员保证公告内容的真实、准确和完整，没有虚假记 载、误导性陈述或者重大遗漏，并对其内容的真实性、准确性和完整性承担个 别及连带法律责任。 一、交易概况 ...

证券代码：839493 证券简称：并行科技 公告编号：2025-113 北京并行科技股份有限公司 出售资产的公告 本公司及董事会全体成员保证公告内容的真实、准确和完整，没有虚假记载、 误导性陈述或者重大遗漏，并对其内容的真实性、准确性和完整性承担个别及连 带法律责任。 一、交易概况 （一）基本情况 北京并行科技股份有限公司（以下简称"公司"）拟向北京乐讯科技有限公司 出售部分早期型号的算力服务器专用设备，预计合同总金额不超过人民币 12,392.72 万元。 （二）是否构成重大资产重组 本次交易不构成重大资产重组。 2、需履行信息披露义务的出售资产事项 本次董事会召开日前，公司分别向北京清云算力科技有限公司、北京企智科 技有限公司等出售 AI 算力服务器、网卡等固定资产（固定资产销售明细见下表）， 合同金额为人民币 3,404.43 万元。根据《北京证券交易所股票上市规则》第 7.1.2 条规定，该等出售资产事项披露如下： | 序 | 合同相对方 | | | 签署时间 | | | | | 标的物类别 | 合同金额 | | --- | --- | --- | --- | --- | --- | --- | -- ...