此次纳入采信的6项HiPi团体标准为《芯粒测试规范》系列标准的第1部分至第6部分，内容涵盖芯 粒互联接口兼容性测试、芯粒可测性设计及互联接口连通性测试、测试文件数据格式、测试设备与 工具、量产测试以及电磁兼容测试等方面。这套标准是HiPi联盟芯粒标准体系的重要组成部分，紧 密结合了国内产业基础和技术实践，明确了测试接口、方法以及覆盖要求，确保芯粒在出厂、封装 和系统集成的各个阶段均具备可测性与接口一致性。该标准体系不仅具有高度的适应性，还提供了 极具工程指导价值的实施方案，旨在为芯粒技术在高性能计算、人工智能、通信等领域的规模化应 用提供坚实的技术支撑。 当前，我国芯粒领域标准化工作持续推进。2022年，HiPi联盟提出芯粒标准体系1.0版本。2023 年，HiPi联盟发布芯粒互联接口团标。2024年，全国集成电路标准化技术委员会成立芯粒标准工 作组；芯粒互联接口标准列为国家集成电路领域标准稳链重点项目之一。2025年8月，国家市场监 督管理总局（国家标准化管理委员会）正式发布了5项《芯粒互联接口规范》（GB/T 46280.1- 2025～GB/T 46280.5-2025）系列推荐性国家标准，将于2026 ...

Core Viewpoint - The sharp rise in memory prices and subsequent panic buying have severely impacted the demand for entry-level and mid-range electronic products, leading to a significant reduction in orders from IC design companies and a polarized development in the foundry industry [1] Group 1: Memory Price Surge and Its Impact - The core issue stems from an imbalance in supply and demand in the memory market, with DRAM and 3D NAND flash memory prices experiencing significant increases due to panic buying from downstream customers [1][2] - The impact of rising memory prices is asymmetric, with entry-level and mid-range markets being the most affected, as manufacturers in these segments face dual pressures of supply tightness and rising chip prices [1][2] Group 2: Supply Chain Adjustments - Anticipating poor terminal sales, IC design companies are adjusting their orders to foundries, leading to a notable decline in orders for mid-range smartphone processors [2] - The AI boom is structurally affecting memory capacity, with data centers expected to consume 70% of global memory chip capacity by 2026, prompting manufacturers to shift focus to high-margin memory products for AI servers [2][3] Group 3: Competitive Landscape and Future Outlook - Major memory manufacturers are prioritizing AI-related memory production, limiting the capacity for standard consumer-grade memory, which puts entry-level products at a disadvantage in the resource competition [3] - The market outlook remains uncertain, with short-term price pressures on entry-level electronic products expected to persist as long as memory prices remain high, leading to subdued shipment volumes [3][4] - By 2026, a critical juncture is anticipated, with potential growth opportunities for the domestic industry as the demand for memory in data centers peaks, though the impact on the consumer market remains to be seen [4]

在打造重点支撑平台方面，《行动计划》明确，打造工业智能体创新中心。加快省级工业智能体创 新中心建设，争取国家级制造业创新中心布局。围绕工业场景"数字员工"需求，支持研发具备环境 感知、自主决策、动态适应能力的工业智能体，聚焦研发设计、生产制造、供应链管理等工业场 景，汇聚高水平智能体应用开发商，搭建工业智能体供需对接平台，构建自主可控技术基座，研发 工业智能体专用工具链，打造工业智能体开放共享生态，提升复杂工业场景下的智能体协作水平。 如果您希望可以时常见面，欢迎标星收藏哦~ 事关"人工智能+"，深圳最新发文。 今日（2月12日），据深圳工信微信公众号消息，深圳市工业和信息化局近日印发《深圳市"人工 智能+"先进制造业行动计划（2026—2027年）》（以下简称《行动计划》），从打造重点支撑平 台、赋能重点产业集群、强化工作保障三个方面提出了13条具体举措，抢抓智能化与工业化交汇 融合的历史机遇，加快人工智能技术与制造业全过程、全要素深度融合。 《行动计划》明确，推动人工智能技术应用于半导体产业链的关键环节，利用AI优化芯片设计、 软件代码等领域和环节的效率。以AI芯片为突破口做强半导体产业，面向AI手机、A ...

点这里加关注，锁定更多原创内容 全球市值最高的10家芯片公司 彭先生表示："人工智能正在重新定义硬件基础设施的基本要求，而Rambus凭借其行业领先的产品 和技术，正处于这一变革的核心，能够实现更高水平的性能。我很高兴加入Rambus董事会，并与 管理团队合作，支持公司的长期发展。" 彭先生拥有康奈尔大学电气工程硕士学位和伦斯勒理工学院电气工程学士学位。 （来源：半导体芯闻综合 ） *免责声明：文章内容系作者个人观点，半导体芯闻转载仅为了传达一种不同的观点，不代表半导体芯闻对该 观点赞同或支持，如果有任何异议，欢迎联系我们。 推荐阅读 10万亿，投向半导体 芯片巨头，市值大跌 黄仁勋：HBM是个技术奇迹 Jim Keller：RISC-V一定会胜出 如果您希望可以时常见面，欢迎标星收藏哦~ 先的芯片和硅 IP 提供商Rambus Inc.今天宣布任命 Victor Peng 为董事会成员，该任命将于 2026 年 2 月 12 日星期四生效。 彭先生是一位经验丰富的企业高管，在顶尖半导体公司拥有超过40年的领导经验。最近，彭先生 担任AMD（Advanced Micro Devices）总裁，领导AMD的嵌入 ...

公开资料显示，黄如，女，回族，1969年11月出生，福建南安人，中共党员，1997年7月参加工 作，理学博士，教授，中国科学院院士、发展中国家科学院院士。她还是第二十届中央候补委员。 黄如本科、硕士就读于东南大学电子工程系，后在北京大学计算机科学与技术系获博士学位。 1997年起，她先后任北京大学信息科学技术学院讲师、副教授（1999年）、教授（2002年），北 京大学博雅讲席教授（2016年）。 如果您希望可以时常见面，欢迎标星收藏哦~ 2月13日，新华社发布国务院任免国家工作人员信息。任命黄如（女）为国家发展和改革委员会副 主任（兼职）；任命陈卫为中国农业大学校长（副部长级）。免去孙其信的中国农业大学校长职 务。 （来源：内容转自澎湃新闻，作者： 岳怀让 ） *免责声明：文章内容系作者个人观点，半导体芯闻转载仅为了传达一种不同的观点，不代表半导体芯闻对该 观点赞同或支持，如果有任何异议，欢迎联系我们。 推荐阅读 10万亿，投向半导体 芯片巨头，市值大跌 黄仁勋：HBM是个技术奇迹 Jim Keller：RISC-V一定会胜出 全球市值最高的10家芯片公司 点这里加关注，锁定更多原创内容 喜欢我们的内容就点 ...

Core Viewpoint - The competition among the three major global memory semiconductor companies for HBM4 (the sixth generation high bandwidth memory) is intensifying, with Samsung Electronics leading the way in production ahead of SK Hynix and Micron, highlighting its advantages in R&D speed and performance [1][2]. Group 1: HBM4 Production and Market Dynamics - Samsung Electronics announced the mass production of HBM4 on the 12th, ahead of its competitors, emphasizing its commitment to exceeding JEDEC standards from the beginning of HBM4's development [2]. - The HBM4 memory will be installed in NVIDIA's upcoming "Rubin" AI accelerator, featuring 2048 I/O ports, which is double the number of the previous generation [2]. - SK Hynix and Micron are also set to begin mass production of HBM4 this month, with SK Hynix claiming to be the only company capable of simultaneously supplying both HBM3E and HBM4 [3][4]. - The term "mass production" used by these companies refers to "risk production," where chips are produced before receiving formal purchase orders, which may lead to market confusion [4][5]. Group 2: Supply Chain and Performance Concerns - The supply chain dynamics for HBM4 are expected to be significantly influenced by NVIDIA's supply strategy, which balances performance and supply stability [6]. - Samsung's HBM4 achieves a stable operating speed of 11.7 Gbps, setting a new industry standard, which is approximately 46% faster than the previous 8 Gbps standard [6]. - Despite Samsung's advancements, the company currently has a yield rate of about 60% for its 1c DRAM, which may decline further during post-processing [6][7]. - SK Hynix has secured the largest HBM4 allocation in negotiations with NVIDIA, but initial reliability assessments indicate challenges in meeting the 11 Gbps performance target [7]. Group 3: Future Expectations and Strategic Adjustments - Industry insiders suggest that NVIDIA may relax its performance requirements for HBM4 to ensure stable supply, given the current memory shortage [7]. - The ongoing investments in new and upgraded production facilities for 1c DRAM will take time to translate into actual capacity, impacting the overall supply chain [7].

Core Viewpoint - A bipartisan group of U.S. lawmakers is urging the Trump administration to impose a comprehensive ban on the export of semiconductor manufacturing equipment to China, citing the ineffectiveness of current targeted trade restrictions [2]. Group 1: Legislative Actions and Recommendations - The lawmakers argue that the current export control system has significant loopholes, as it only targets specific entities within China rather than implementing a nationwide ban [2]. - They emphasize the need for the U.S. government and its allies to enforce nationwide restrictions on critical semiconductor manufacturing equipment and components that China cannot currently produce domestically [2]. - The group highlights the importance of addressing the supply chain's vulnerabilities, particularly concerning equipment produced by Dutch companies, which has reportedly doubled from 2022 to 2023 and is expected to double again from 2023 to 2024 [3]. Group 2: Urgency and Strategic Concerns - The lawmakers express urgency in their recommendations, warning that if China continues to develop its own chip manufacturing technology without restrictions, it could render U.S. and allied export controls ineffective [3]. - They also call for strict actions against companies that provide after-sales maintenance services for existing controlled equipment in China, indicating a broader strategy to limit China's technological advancements [3].

Core Viewpoint - The conference held in Zhejiang Province marks a significant opportunity for the deployment of innovative ecosystems and showcases technological achievements, with Hangxin Technology winning the first prize in the Zhejiang Provincial Science and Technology Progress Award for its project on AI-enabled chip design and manufacturing integration [1][2]. Group 1: Award and Recognition - Hangxin Technology achieved a historic breakthrough by winning the first prize in the Zhejiang Provincial Science and Technology Progress Award for its project titled "Key Technologies and Scalable Applications of AI-Enabled Chip Design-Manufacturing Integration" [1][2]. - This award signifies Hangxin Technology's commitment to the chip sector and its success in overcoming critical core technologies, reflecting its strong innovation capabilities [2][3]. Group 2: Project Details - The project focuses on the critical area of Design Technology Co-Optimization (DTCO) in chip design and manufacturing, utilizing AI technology to overcome performance bottlenecks of traditional EDA tools, thereby enhancing design efficiency and reliability [2][3]. - The achievement supports the autonomy and industrial security of domestic EDA tools in high-end chip design and manufacturing, aligning with Zhejiang's technological innovation direction and the core requirements for the Provincial Science and Technology Progress Award [2][3]. Group 3: Future Outlook - Hangxin Technology plans to leverage this recognition to deepen its commitment to independent innovation, focusing on core technology breakthroughs in the chip field and accelerating the industrial application of semiconductors [4]. - The company aims to contribute to the integration of artificial intelligence and the chip industry, facilitating new breakthroughs in Zhejiang's innovation landscape [4].

据英国《金融时报》今日报道，本轮融资由比利时风投机构Hummingbird Ventures领投，交易完 成 后 ， 总 部 位 于 英 国 的 Olix Computing 估 值 超 10 亿 美 元 。 该 公 司 此 前 曾 从 Plural 、 Vertex Ventures、LocalGlobe 及 Entrepreneurs First 等机构获得未披露金额的融资。 如果您希望可以时常见面，欢迎标星收藏哦~ 英国初创企业Olix Computing正在研发集成光学组件的人工智能芯片，该公司已完成2.2 亿美元融 资。 Olix 的芯片针对AI 推理进行优化，推理是指 AI 模型完成训练后，在实际业务环境中运行的过 程。目前尚不清楚该处理器集成了哪些光学组件，以及具体应用方式。不过，该公司官网的一篇博 客文章提到，其芯片采用"创新型内存与互联架构"。这表明 Olix 正在利用光子组件构建互联模 块，即处理器中负责在电路之间传输数据的部分。 目前已有多家初创企业在研发光子互联技术。其中融资规模领先的Ayar Labs已开发出一款光学中 介 层 （ interposer ） ， 可 用 于 打 造 面 ...

Group 1 - Nvidia is a leading player in the AI wave, dominating the market with its GPU architecture and CUDA ecosystem, which are crucial for AI training and inference [1] - The rise of Nvidia is not an isolated event; it is part of a complex ecosystem involving wafer manufacturing, advanced packaging, storage bandwidth, network interconnects, power and cooling, materials, and equipment [1] - The AI infrastructure competition has shifted from a focus on computing power to overcoming connectivity bottlenecks, highlighting the importance of interconnect technologies [8] Group 2 - Credo, a lesser-known company, has seen its stock price surge due to its active electrical cables (AEC) that are essential for connecting GPUs in large AI data centers [2][4] - Credo holds an 88% market share in the AEC market, with revenues expected to double again by 2026, reaching nearly $1 billion [4] - The demand for optical fibers is increasing as they offer superior speed and energy efficiency compared to copper cables, with Corning emerging as a key player in this market [5][7] Group 3 - Lumentum is transitioning from a telecom supplier to a core enabler for AI data centers, with significant revenue growth driven by strong demand for optical devices [11][13] - Coherent is also innovating in the optical transceiver space, with a focus on high-speed devices that are essential for AI and machine learning applications [15] - SiTime specializes in precision clocks that are critical for synchronizing AI servers and networks, experiencing substantial revenue growth due to increased demand from data center clients [19][21] Group 4 - The supply chain supporting Nvidia's empire is both powerful and fragile, characterized by high technical barriers and single-point dependencies [27][28] - Companies like Corning, SiTime, and Nittobo are crucial for the reliability of AI systems, with their products being essential for maintaining system performance under extreme conditions [26][25] - The future of these companies is uncertain, as they face potential demand fluctuations and the risk of supply chain disruptions [27][28]