陈俊兆 A0230124100001 chenjz@swsresearch.com を伟达拟整合 Groq 技术加 势,盛合晶微过会 20260301 电子行业周报 本期投资提示: 2026 年 03 月 01 日 证券分析师 杨海晏 A0230518070003 yanghy@swsresearch.com 袁航 A0230521100002 yuanhang@swsresearch.com 杨紫璇 A0230524070005 yangzx@swsresearch.com 研究支持 姜安然 A0230125070008 jiangar@swsresearch.com 联系人 请务必仔细阅读正文之后的各项信息披露与声明 Film 一周板块回顾：1）2 月 23 日至 2 月 27 日, A 股电子指数整体上涨。上证指数上涨 1.98%，深证综指上涨 3.10%，创业板上涨 1.05%，申万电子指数上涨 4.07%；海外 市场方面，费城半导体指数下跌 1.96%，纳斯达克指数下跌 0.95%，恒生科技下跌 1.41%，中国互联网 50 下跌 2.09%。重点个股涨幅前列：明阳电路（43.54%）、瑞 可达 （ ...

行 业 研 究 / 行 业 点 评 证券分析师 杨海晏 A0230518070003 yanghy@swsresearch.com 袁航 A0230521100002 yuanhang@swsresearch.com 杨紫璇 A0230524070005 yangzx@swsresearch.com 研究支持 姜安然 A0230125070008 jiangar@swsresearch.com 联系人 陈俊兆 A0230124100001 chenjz@swsresearch.com 2026 年 03 月 01 日 英伟达拟整合 Groq 技术加强推理 优势，盛合晶微过会 看好 ——20260301 电子行业周报 本期投资提示： 行 业 及 产 业 请务必仔细阅读正文之后的各项信息披露与声明 本研究报告仅通过邮件提供给 博时基金 博时基金管理有限公司(researchreport@bosera.com) 使用。1 证 券 研 究 报 告 电子 相关研究 - ⚫ 一周板块回顾：1）2 月 23 日至 2 月 27 日，A 股电子指数整体上涨。上证指数上涨 1.98%，深证综指上涨 3.10%，创业板上涨 1. ...

本报讯 （记者刘晓一）2025年，全球存储行业在AI需求爆发与供应紧张双重驱动下景气度上行，成为资本市场焦点。 2月27日晚间，国内存储芯片头部企业深圳市德明利技术股份有限公司（以下简称"德明利"）发布2025年年度报告，凭借 全栈存储解决方案优势，公司实现营收与盈利双提升。 年报显示，德明利2025年营业收入107.89亿元，同比增长126.07%，首次突破百亿元；扣非归母净利润6.88亿元，同比增长 120.77%，主业盈利持续增强。同时，公司拟向全体股东每10股派发现金红利4元，与投资者共享成长。 AI浪潮重塑存储行业增长逻辑，需求核心由传统消费电子转向企业级高性能存储，定制化、场景化能力成为关键。德明利 受益行业景气，更以全栈解决方案完成从标准产品供应商到场景化服务商的转型，抢占AI市场先机。 公司聚焦全链路存储解决方案，构建定制化交付体系，推动全线业务高增长。2025年，公司固态硬盘营收45.82亿元，同比 增长99.18%，占比48.20%，通过定制化方案进入头部互联网与服务器厂商供应链，QLCNAND技术实现商业化落地。嵌入式存 储营收36.63亿元，同比大增334.43%，与国产SoC平台深度 ...

智通财经APP讯，佰维存储(688525.SH)披露2025年年度业绩快报，2025年度实现营业总收入112.96亿 元，同比增长68.72%;归属于母公司所有者的净利润8.67亿元，同比增长437.56%。 公司在智能穿戴领域深耕多年，构建了差异化的竞争优势，公司ePOP等代表性存储产品具有低功耗、 快响应、轻薄小巧等优势，产品表现出色，已进Meta、Rokid、雷鸟创新、闪极等国内外知名AI/AR眼 镜厂商，Google、小天才、小米等国内外知名智能穿戴厂商供应链体系。2025年公司AI新兴端侧存储产 品收入约17.51亿元，其中，AI眼镜存储产品收入约9.6亿元。2026年随着AI眼镜的放量，公司与Meta等 重点客户的合作不断深入，将推动公司相关存储产品业务的持续高速增长。 ...

收购预期落空引发估值坍塌 根据公告，开普云原计划通过"支付现金+发行股份"的方式，购买深圳市金泰克半导体有限公司（下 称"金泰克"）持有的南宁泰克100%股权，并募集配套资金。就收购原因，开普云表示，公司在AI大模 型、AI智能体和行业应用信息化智能化领域的专长与南宁泰克在存储技术研发的实力相结合，可以推 动AI优化存储解决方案的开发。 收购事项延宕半年后，开普云（688228.SH）一纸公告宣布终止计划，引发股价大跌。 2月24日晚间，开普云公告称，因交易核心条款未能达成一致，公司决定终止收购南宁泰克半导体有限 公司（下称"南宁泰克"）100%股权的重大资产重组计划。市场此前对该笔收购能否达成已有疑虑，公 司股价自1月末见顶后便持续单边下行。 随着收购的正式落空，2月25日早盘，开普云股价大跌，盘中最大跌幅超18%，最终收跌14.78%，报收 115.68元，最近15个交易日累计下挫54%。为应对股价的持续下跌，公司同步抛出股份回购方案，拟以 不超过315元/股的价格，动用自有资金5000万元至1亿元进行回购，以维护公司价值及股东权益。 然而市场担忧的是，在资产注入预期不再、主营业务盈利能力孱弱的背景下，这 ...

在此轮存储涨价潮中，AI存储被视为最大的拉动因素。其中，全球三大存储芯片巨头三星、SK海力 士、美光，近三个月均将战略重心聚焦于AI高端存储领域，带动其股价同步走强。 综合多家韩国媒体消息，作为行业风向标，三星除推进最新一代HBM4芯片量产并洽谈高价定价外，还 将平泽P4工厂的投产时间从2027年一季度提前至2026年四季度，重点部署高性能DRAM与HBM产品， 抢占AI存储市场先机。据业界分析师估算，三星HBM4组件拟定价约700美元，较上一代HBM3E高出 20%—30%，其营业利润率预计可达50%—60%，盈利空间大幅提升。 中经记者谭伦北京报道 自2025年的存储芯片涨价大潮仍在延续。 2026年2月19日，韩国存储芯片巨头三星电子股价再度创下历史新高，盘中一度上涨逾5%，最高触及 190900韩元，今年以来累计涨幅已近50%，延续了2025年的迅猛涨势。 这一波狂飙背后，直接诱因是三星正在进行的涨价谈判。据韩媒披露，三星正与英伟达（NVIDIA）及 其他大客户敲定最新一代HBM4及超高性能存储芯片的供应协议，其报价较上一代产品高出至多30%。 在2025年存储价格已经翻倍的基础上，这种"利好叠加"让 ...

凭借在精密陶瓷领域的技术壁垒， TOTO 已卡位 AI 算力背后的关键一环 ——NAND 存储芯片制造，且该业务已贡献了公司 40% 的营业利润。 硬·AI 作者 | Kozmon 编辑 | 硬 AI 在东京股市的传统板块中，TOTO（东陶）长期被视为卫浴行业的防御性资产。但在伦敦激进对冲基金 Palliser Capital眼中，这家拥有百年历史的"马桶大王"，是一只被严重误读且低估的AI存储芯片概念股。 01 从烧制马桶到生产晶圆 这家公司最不为人知的杀手锏是"静电卡盘"。在半导体制造环节，这是一种利用静电力吸附并固定硅晶圆 的关键组件。随着AI浪潮推动NAND存储芯片向更高层数、更复杂结构演进，制造工艺中对"低温蚀刻"的 需求激增。 这正是TOTO的护城河所在。利用自1980年代以来在卫浴制造中积累的精密陶瓷技术，TOTO生产的静电 卡盘能够在极低温度下保持极高的稳定性。Palliser指出， TOTO 拥有长达五年的竞争优势，短期内竞争 对手难以望其项背。 尽管TOTO早已涉足此领域，但直到近年AI需求爆发，这项业务才从边缘走向舞台中央。Palliser强调， TOTO已悄然从一家传统的卫浴冠军，进化 ...

Core Insights - The storage chip sector is experiencing a continuous increase in market demand, with Jiangbolong revealing key developments in mSSD applications, UFS4.1 product collaborations, wafer supply assurance, and storage cycle outlooks during investor discussions [1] Group 1: mSSD Development - Jiangbolong's mSSD, utilizing Wafer-level System-in-Package (SiP) technology, integrates multiple components into a single package, offering significant manufacturing cost advantages while maintaining performance comparable to traditional SSDs [3] - The mSSD is being rapidly adopted by leading PC manufacturers, with analysts noting its potential as the standard storage solution for next-generation mobile computing platforms due to increasing demands from AI PCs for space and power efficiency [3] Group 2: UFS4.1 Product Capabilities - Jiangbolong is among a few global companies capable of developing UFS4.1 products at the chip level, with its self-developed controller offering superior performance in terms of process, read/write speed, and stability compared to comparable market products [3] - UFS4.1 is positioned as a premium storage option for Tier 1 clients' flagship smart terminal models, with mass shipments expected to commence soon [3][4] Group 3: Supply Chain Resilience - Jiangbolong has established deep cooperative relationships with major wafer manufacturers and leading smart terminal device companies, paving the way for large-scale commercialization of UFS4.1 [4] - The company has signed long-term supply agreements and commercial memorandums with global wafer suppliers, enhancing its supply chain resilience amid structural tensions in storage wafer supply [5] Group 4: Market Outlook - Jiangbolong anticipates a positive trend in storage prices, driven by structural changes in demand due to AI applications and the rapid expansion of AI infrastructure, coupled with a shortage of HDD supply [5] - The company highlights that the capital expenditure recovery of storage manufacturers will have limited short-term contributions to output growth due to the lag in capacity construction cycles [5]

Core Insights - Yingrun Technology has developed a series of products tailored for various AI scenarios and is proactively advancing PCIe 6.0 controller research to meet the future demands of high-density AI data center deployments [1][6] - The global storage chip market is experiencing structural expansion opportunities driven by explosive data growth due to AI technology, while also entering a critical phase of domestic substitution [1][4] - Despite progress in technology and products, domestic storage enterprises face multiple challenges in scaling production and achieving profitability [7] Company Overview - Yingrun Technology, established in 2017, focuses on independent research and innovation in storage technology, with core products including SSD controllers, solid-state drives, and various storage system solutions [1][3] - The company has launched 10 controller chips, covering consumer, industrial, and enterprise markets, achieving full-stack coverage from SATA to PCIe 5.0 in the enterprise SSD sector, with a maximum single-disk capacity of 64TB [3][4] Supply Chain Strategy - Yingrun Technology has adopted a strategy of deep integration with domestic supply chains, collaborating with upstream and downstream companies, including Yangtze Memory Technologies, to benefit from the increasing proportion of domestic chip procurement [4][6] - In 2024, the domestic module manufacturers' procurement ratio of domestic chips exceeded 55%, enhancing supply chain stability and mitigating geopolitical risks [4] Technological Development - The company is addressing the new requirements for storage performance, latency, and durability driven by AI applications by launching the "Dongting" series AI SSDs, featuring a flagship product with a random write latency as low as 4μs and a 4K random write performance of 1600K IOPS [6][7] - Yingrun Technology is exploring advanced process technologies and has laid out plans for PCIe 6.0 controller development to meet future AI data center needs [1][6] Industry Challenges - The domestic storage industry still faces technical shortcomings, with the yield rate of domestic DDR5 storage chips at approximately 65%, significantly lower than Samsung's 85% [5] - The transition from laboratory to mass production and from mass production to profitability presents significant challenges, particularly in high-reliability AI storage scenarios [7] - Achieving compatibility and long-term reliability in the enterprise storage market remains a barrier for domestic products to enter core data centers [7]

Funding Situation - Since February 2026, main funds have continuously net outflow, with a net outflow of 854 million yuan on February 4, and a higher outflow of 1.52 billion yuan on February 2, accounting for 25.27% of the day's trading volume [1] - Margin financing has also contracted, with a net repayment of 335 million yuan on February 2, indicating an increase in risk aversion among leveraged funds [1] Stock Price Situation - As of the close on February 11, the stock price was 132.97 yuan, down 21.11% from the high of 163 yuan on February 2, with a cumulative decline of 18.68% over 20 days [2] - The technical indicators have fallen below the moving average support, and the turnover rate has decreased from over 7% at the beginning of the month to 4.38%, indicating reduced market activity [2] - During the same period, the overall electronic sector declined by 1.09%, exacerbating the adjustment pressure on individual stocks [2] Company Fundamentals - The company's 2025 earnings forecast indicates a projected growth in net profit attributable to shareholders of 81.77%-134.78%, but the gross profit margin for the first three quarters was only 3.13%, with a year-on-year decline in net profit of 1.36%, reflecting a "revenue growth without profit growth" situation [3] - The current price-to-earnings ratio (TTM) is 238.58 times, which reflects high growth expectations in the AI storage sector, but the high valuation is more likely to trigger profit-taking in a weak market [3] Industry and Risk Analysis - Despite TrendForce's expectation of continued price increases in storage for Q1 2026 (with DRAM contract prices revised up by 90-95% quarter-on-quarter), there are market concerns about the risk of reaching a peak in the cycle [4] - The company has a large inventory (approximately 2.858 billion yuan as of the 2025 third-quarter report), and a price reversal could lead to impairment pressure, increasing investor caution [4] Recent Events - On January 19, significant shareholder reductions led to a single-day drop of 8.12%, compounded by some executives also reducing their holdings, which temporarily impacted market confidence [5] - Despite this, institutions remain generally optimistic, with four institutions giving buy ratings within 90 days [5]