计算机行业事件点评 ➢ 事件概述：当地时间 2025 年 10 月 10 日，美国总统特朗普在社交媒体宣 布：从 11 月 1 日起，对中国产品征收 100%关税；并对所有"关键软件"实施 出口管制。 ➢ 政策层面：政策大力支持，国产化持续推进。2025 年 9 月 30 日，国务院 办公厅印发《关于在政府采购中实施本国产品标准及相关政策的通知》。政策明 确指出，政府采购活动中既有本国产品又有非本国产品参与竞争的，依法对本国 产品给予价格评审优惠，对本国产品的报价给予 20%的价格扣除，用扣除后的价 格参与评审。 ➢ 产业层面：重要订单落地，信创不断加速。1）工商银行 2025 年度海光芯 片服务器采购项目正式进入公示期，项目规模或达 30 亿元。国产芯片在金融领 域已积累广泛认可，尤其在处理高并发、高敏感性业务场景中展现出可靠性能。 随着信创工作步入收官阶段，金融机构在系统选型时更加注重稳定性与迁移效 率，成熟品牌的优势逐渐凸显。 2）2025 年中国联通通用服务器集中采购中，通用服务器标包二(鲲鹏/海光芯片） 78395 台，国产化占比 90.1%。本次集采划分为两个标包，共计通用服务器 87020 台， ...

vivo 移动互联网业务为全球超过 4 亿用户提供应用商店、短视频、广告等服务，其分布式消息中间件 平台承担了实时数据接入与分发的关键角色，日均处理数据量达十万亿级别。面对业务规模的持续增 长，vivo 在消息系统架构演进中通过引入 Apache Pulsar，有效解决了原有 Kafka 架构在多集群管理、 弹性扩缩容和海量分区场景下面临的诸多瓶颈。 业务痛点与挑战 随着 vivo 业务流量的数十倍增长，原有基于 Kafka 的消息系统逐渐显露出架构局限性。Topic 和分区数 量的持续增加严重影响了集群性能，大量分区导致磁盘随机读写加剧，违背了 Kafka 依赖顺序读写实现 高性能的设计初衷。集群规模扩张后，资源组隔离与集群拆分的运维成本显著上升，且 Kafka 无法实现 动态扩缩容，机器资源利用率低。在面对突发流量时，扩容速度缓慢，流量均衡耗时较长，消费端性能 过度依赖分区数量，造成元数据急剧膨胀。此外，集群基数增大导致硬件故障频发，且故障直接传导至 客户端，缺乏有效的中间容错层。 技术选型 滴滴大数据团队于 2021 年 1 月开始调研 Apache Pulsar，同年 8 月正式上线首个 Pulsar ...

| 股票简称 | 股票代码 | 限售股类型 | 解禁股数 | | --- | --- | --- | --- | | 国新健康 | 000503 | 股权激励限售流通 | 167.98万 | | 康缘药业 | 600557 | 股权激励限售流通 | 31.55万 | | 森特股份 | 603098 | 股权激励限售流通 | 27万 | | 三人行 | 605168 | 股权激励限售流通 | 16.53万 | | 海量数据 | 603138 | 股权激励限售流通 | 566.29万 | | 天益医疗 | 301097 | 延长限售锁定期流通 | 4060万 | | 铭利达 | 301268 | 延长限售锁定期流通 | 1.9亿 | | 恒宇信通 | 300965 | 延长限售锁定期流通 | 3510.45万 | | 普源精电 | 688337 | | 1.18亿 | | 中复神鹰 | 688295 | | 2.98亿 | | 长光华芯 | 688048 | | 7007万 | 今日具体限售解禁股情况如下： 智通财经APP获悉，10月9日共有11家上市公司的限售股解禁，解禁总市值约243.03亿元。 ...

据了解，联合体构建了覆盖"靶点确认-分子设计-分子优化-临床前研究-IND申报"全链条的AI赋能研发 体系，深度融入跨学科协作与数据驱动决策。本次针对恶性肿瘤的靶点研究中，AI系统从海量数据 （603138）中筛选出三类高优先级临床需求，经跨学科专家委员会多轮论证后锁定最终研究方向；而多 组学数据与AI建模的交叉验证，则大幅提高了靶点生物学可信度。团队在立项阶段即同步评估靶点的 成药性、专利布局空间及商业转化可行性，将立项周期缩短至传统模式的1/3。 目前，联合体在快速推动该项目进展的同时，正在持续推进三方面的工作。一是建立"AI+专家"协作的 标准化靶点确立与验证流程，形成可复制的方法论体系；二是通过高质量湿实验数据持续训练优化生物 医药大模型，强化全生命周期赋能能力；三是吸引药企、同行和科研机构加入，共建开放创新生态，加 速商业转化与合作。 中证报中证网讯（王珞）成都先导9月29日消息，由成都先导、水木分子、嘉华药锐及英飞智药四家生 物医药企业组成的"AI赋能的源头创新联合体"（以下简称"联合体"）发布了最新研究进展，通过整合人 工智能（AI）技术、多组学数据与专家经验反馈，联合体成功锁定一个针对恶性肿瘤 ...

Group 1: Event Overview - The fourth Global Digital Trade Expo was held from September 25 to 29 in Hangzhou, featuring an exhibition area of 155,000 square meters and 1,812 exhibitors, including 381 international exhibitors, accounting for 21% of the total [1][2] - The theme of this year's expo was "Seeing the Innovative Future at the Digital Trade Expo," highlighting the opportunities presented by AI and robotics in various sectors [1] Group 2: AI and Robotics Development - Wang Xingxing, founder of Yushu Technology, emphasized that AI and robotics represent significant opportunities for the younger generation, with humanoid robots increasingly penetrating entertainment and practical applications [1] - The AI exhibition area expanded to include over 300 companies showcasing applications of AI across agriculture, business, finance, technology, and emergency management [3] Group 3: Innovations and Products - The "Black Panther," a quadruped robot developed by Zhejiang University, can run over 10 meters per second and has undergone ten years of development with four iterations [4] - Companies like Kangbeijing (Hangzhou) Environmental Technology are innovating with products like the P05 device for high-cleanliness laboratories, while others are developing AI health assistants for personal health management [5] Group 4: Global Trade Cooperation - The expo has become a vital platform for promoting global trade cooperation, with 11,000 international merchants registered, a 64% increase from the previous year [6] - Indonesia, as the guest country, has expressed procurement intentions exceeding 5 billion yuan, with total expected procurement reaching 30 billion yuan [6] - The event aims to leverage AI to enhance e-commerce data utilization, enabling companies to better understand and penetrate overseas markets [7]

近期引起市场关注的明星产品——高德扫街榜，已经实现每天超过15亿公里生活服务导航总里程，榜单 发布一天后，相关用户数就超过了4000万。刘振飞称，要将如此巨量的真实出行数据精准量化，同样需 要空间智能。 面向产业生态，刘振飞透露，智能眼镜、汽车、机器人及低空飞行等领域的企业已开始接入高德的空间 智能能力。刘振飞表示，高德将自身角色定位为"负责搞定世界的复杂性"的底层服务提供者，目标是把 空间智能做成类似云计算的社会化基础设施，降低各行业理解与连接物理世界的门槛，"像云计算一 样，成为360行与物理世界交互的标配。" 附：高德地图董事长刘振飞致辞全文： 大家好，我是高德的刘振飞，欢迎大家来到这个会场，非常荣幸能在又一年的云栖大会上与各位相聚， 共同探讨和学习关于AI的一切。也想借此机会，向大家汇报高德在AI领域的最新进展和思考。 2025年对于高德来说，注定是非常热闹且意义非凡的一年。前不久，我们刚上线了全球首个基于真实到 店行为的生活服务榜单——高德扫街榜；上个月初，我们公布了全球首个地图AI原生智能体——高德 地图2025；还有红绿灯AI领航、全域车道级导航、卫星求救等，一系列新功能。 9月25日，阿里云栖 ...

Group 1 - The stock prices of companies such as 信安世纪 (Xinan Century) increased by over 14%, 星环科技 (Star Ring Technology) by over 13%, and 飞荣达 (Feirongda) by over 11% [1] - Other companies including 意华股份 (Yihua Co.), 烽火通信 (Fenghuo Communication), and 海量数据 (Mass Data) also saw their stocks hit the daily limit [1]

Core Insights - Artificial intelligence is becoming a significant driving force behind a new wave of technological revolution and industrial transformation, fundamentally altering production methods, lifestyles, and social governance [1] - The development of large AI models requires vast amounts of data, which raises concerns about the protection of personal information rights and presents new challenges to the personal rights system [1] Group 1: Protection and Utilization of Publicly Available Personal Information - The protection of publicly available personal information is increasingly important in the training of AI models, as much of the training data comes from such sources [1] - The Personal Information Protection Law in China allows for the processing of publicly available personal information without consent, provided it meets certain conditions, including reasonable scope and significant impact on personal rights [1] - The challenge arises when AI models collect fragmented personal information, potentially leading to the reconstruction of sensitive personal data, which necessitates obtaining consent [1] Group 2: Safeguarding Sensitive Personal Information - The advancement of AI technology enhances data analysis capabilities, posing new threats to personal information security, particularly sensitive data [2] - During the training phase of generative AI, it is crucial to anonymize sensitive personal information to prevent severe consequences from potential leaks [2] - Historical incidents, such as vulnerabilities in ChatGPT, highlight the risks associated with sensitive information exposure and the need for ongoing regulatory measures [2] Group 3: Challenges in Generative AI Operations - Generative AI poses significant challenges to the protection of personal privacy and information, necessitating measures to prevent sensitive data from being included in generated content [3] - The risk of generative AI producing malicious or false content is a concern, as inaccuracies in training data can lead to harmful outputs that may relate to sensitive personal information [3] - The importance of protecting personal identifiers, such as voice, is increasingly recognized due to the potential for deepfake technology to exploit these identifiers [3] Group 4: Protection of Personal Identifiers - The rise of deepfake technology allows for the creation of fraudulent audio and visual content, posing significant risks to individuals [4] - High-profile cases, such as the exploitation of Scarlett Johansson's voice by OpenAI, underscore the urgent need for legal protections against the misuse of personal identifiers [4] - The necessity for stricter regulations to prevent the infringement of personal rights through deepfake technology is becoming more apparent [4] Group 5: Virtual Digital Humans and Personal Rights - The emergence of virtual digital humans presents new challenges to the personal rights system, particularly regarding the use of real individuals' likenesses in creating virtual representations [5] - The commercial viability of virtual digital humans is being explored, but their interaction with the real world raises questions about potential violations of personal rights [5] - The determination of whether a virtual digital human infringes on an individual's rights hinges on the recognizable similarity to the real person, necessitating legal standards for assessment [5] Group 6: New Types of Personal Rights - Virtual digital humans can act as "virtual avatars," extending beyond traditional rights to encompass new forms of personal rights [6] - Legal interpretations are evolving to recognize that the use of real personal information in training AI companions can infringe upon various personal rights, including name and likeness rights [6] - The concept of a "virtual avatar" represents a composite of an individual's identity, necessitating the establishment of new legal protections for these emerging personal rights [6]

Group 1 - The core viewpoint is that the acceleration of marketization in the data element sector is supported by both policy and technological advancements, but the "lack of scenarios" remains a key bottleneck for industry development [1] - Industry experts at the 2025 China International Big Data Industry Expo emphasized that data itself is not an asset; it must be integrated into scenarios and solve real problems to be transformed into true value [1] - By creating replicable cases, promoting co-creation of technology and business, and building a secure and trustworthy environment, the value of existing data can be activated, leading to the emergence of new scenarios and business models [1] Group 2 - The future consensus is that "scenario-driven" approaches will transform data from "dormant resources" into "flowing value," which will genuinely support industrial upgrades and improve people's livelihoods [1]

网通车载两开花 核心产品稳占市场份额 裕太微表示，为更快推进网通和车载应用领域的拓展，公司于2024年成立了单独的网通事业部和车载事 业部。 其中，网通事业部主要由网通以太网物理层芯片、网通以太网交换机芯片、网通以太网网卡芯片组成， 覆盖所有除车载业务以外的其他所有的行业和应用领域。网通事业部的 2.5G 以太网物理层芯片在国内 运营商市场的 FTTR 和 XPON 产品领域，凭借卓越的性能和稳定的品质，2.5G 以太网物理层芯片产品 取得较高的市场份额。而在千兆以太网物理层芯片市场，广泛应用于国内显示屏、服务器以及机顶盒市 场，稳固地占据着国产千兆以太网物理层芯片市场份额的领先地位。百兆以太网物理层芯片产品在国内 工业伺服、机器人以及工业控制市场展现了强大的竞争力。通过 2.5G、千兆、百兆等多速率传输能 力，并结合 MII/RGMII/SGMII/QSGMII/高速 SerDes 等多种端口形态，可灵活适配运营商网络、工业控 制、数据中心及消费电子等领域的多样化应用需求。 而车载事业部专注于车载高速有线通信芯片的产品战略和规格制定、研发设计、市场推广和销售。面向 车载以太网和车载串行解串芯片的需求增长迅猛 ...