三星与SK海力士加入OpenAI"星际之门"项目 10月2日，周四，韩国两大芯片龙头三星电子与SK海力士股价大涨。此前一天，两家公司与美国人工智能巨头OpenAI宣布合作，加入该公司的"Stargate （星门）"计划。 SK海力士一直是英伟达的主要存储供应商，而其竞争对手三星据报道正推进HBM4通过英伟达认证。 三星历来是存储市场的领军者，但在HBM领域被取得领先的SK海力士所威胁。调研机构Counterpoint Research在7月发布的报告显示，第二季度SK海力士 的存储营收已追平三星，二者正在全球存储市场争夺头把交椅。 三星股价上涨逾4%，创下自2021年1月以来新高；SK海力士大涨逾9%，为2000年以来的最高水平。 OpenAI在声明中表示，此次合作将"聚焦提升新一代AI所需先进存储芯片的供应，并在韩国扩大数据中心产能"。 这家ChatGPT的开发者称，两家韩国企业计划扩大先进存储芯片的产量，而这些芯片对其AI模型的运行至关重要。 消息发布之际，OpenAI首席执行官山姆·奥特曼在首尔与韩国总统李在明，以及三星与SK海力士的最高管理层会面。 OpenAI还签署了一系列协议，探索在韩国建设新一代 ...

SK海力士目前是英伟达AI加速器所需HBM芯片的全球领导者，而三星电子也正奋力追赶，力图成为主要供应商。这项与OpenAI的直接合作，将有助于巩固两 家公司在先进AI存储芯片领域的领先优势，使其在与美光科技等公司的全球竞争中占据更有利的位置。 这笔交易的背后，是全球范围内对AI算力需求的爆炸式增长。以OpenAI和英伟达为首的公司正在全球推动建设用于新一代人工智能工具的数据中心。就在上个 月，英伟达宣布将向OpenAI投资高达1000亿美元，以支持其新的数据中心和其他基础设施。 协议内容显示，此次合作远不止于芯片供应。 三星集团旗下的其他公司，包括三星SDS、三星物产和三星重工，也将与OpenAI合作探索未来技术，合作领域 涵盖浮动数据中心以及数据中心设计等。此外，SK集团旗下的SK电讯将与OpenAI联手，在韩国西南部共建一个专用的OpenAI数据中心。 OpenAI已与韩国两大芯片巨头三星电子和SK海力士达成初步协议，为其庞大的星际之门项目供应芯片及其他设备。 根据三星和海力士周三发布的声明，OpenAI CEO Sam Altman在首尔签署了一份意向书。该协议旨在将这两家在存储芯片领域占主导地位的公司 ...

电子 行业研究/行业周报 LPDDR4X 成品普涨，Q4NAND 或涨 5-10% ——存储芯片周度跟踪（2025.09.22-2025.09.26） ◼ 核心观点 HBM：美光科技 CEO 称 HBM 芯片供不应求，将成 2026 年存储板 块核心增长动力。根据财联社报道，美光科技 CEO Sanjay Mehrotra 在财报电话会上指出，预计全球存储芯片（尤其 HBM）供需不平衡 将加剧，因 DRAM 库存低于目标，而 NAND 库存持续下降；同时， 2026 年 HBM 产能已基本锁定，需求增长显著，2026 年 HBM 出货量 增速预计超整体 DRAM，成存储板块核心增长动力。 市场端：资源涨价叠加需求端询单动作频频，上周嵌入式 NAND 和 LPDDR4X 成品迎来普涨行情。根据 CFM 闪存市场报道，上游供应 端涨价的影响下，近期存储现货市场询单动作频频，资源涨价也已逐 渐传导至嵌入式成品端，上周 eMMC 和 UFS 普遍较大幅度调涨价 格。而 LPDDR4X 也因部分原厂暂停报价，并传出计划将大幅上调 LPDDR4X 资源价格，市场看涨情绪快速升温，基于一定的涨价预 期，高位横盘一月有余的 ...

以下文章来源于科创板日报 ，作者卞纯 科创板日报 . 专注科创板和科技创新，上海报业集团主管主办，界面财联社出品。 随着人工智能（AI）热潮蔓延至科技界原本不那么引人注目的角落， 本月，韩国两大芯片制造商的市值暴增逾千亿美元，而且这一迅猛的涨势 有望持续下去 。 本月以来，SK海力士凭借其在高带宽内存（HBM）芯片——这是人工智能应用的关键——领域的主导地位而股价飙升。投资者对SK海力士 规模更大的竞争对手三星电子将迎头赶上的押注也在上升。 与此同时，来自英伟达以外的人工智能处理器制造商的需求也在增长，由此带来的连锁效应正扩散至那些此前被视为前景黯淡的传统技术芯 片。 分析师们本季度已将这两家韩国芯片巨头的目标股价上调了约30%， 摩根士丹利预测，鉴于明年可能存储领域出现供需失衡，内存芯片行 业将迎来 "超级周期" 。 这些股票目前仍被视为相对便宜，来自海外基金的投资料将增加。 "我们预计外资将继续流入韩国芯片股，"Allspring Global Investments的投资组合经理Gary Tan表示。"围绕这些公司的乐观情 绪的最终推动力，仍然是美国大规模的人工智能基础设施建设，各大科技公司公布的重大资 ...

智通财经APP获悉，本月，韩国两大芯片制造商的股价大幅上涨，市值增加了逾 1000 亿美元。这种涨 势预计还将持续下去，因为人工智能热潮正在向科技领域中较为平淡的领域蔓延。分析师们本季度将韩 元汇率的目标值上调了约 30%。摩根士丹利预测，由于供需失衡的情况将在明年出现，内存市场将进 入"超级周期"。尽管如此，这些股票仍被认为价格相对较低，而且海外资金的投资活动预计也会增加。 由于 SK 海力士凭借其在HBM芯片领域的主导地位(这些芯片是人工智能应用的关键组成部分)而实现大 幅增长，而其更大的竞争对手三星电子也加大了追赶力度。与此同时，除了英伟达(NVDA.US)之外， 越来越多的 AI 处理器制造商的需求也在增加，这带来的连锁效应正蔓延到此前被认为已无市场的旧技 术芯片领域。 Allspring Global Investments投资组合经理Gary Tan表示："我们预计外国资金将继续流入韩国芯片类股 票。这些股票之所以备受看好，其根本原因在于美国大规模的 AI 基础设施建设正在推进，而这一进程 得到了各大科技公司发布的巨额资本支出公告的支撑。" 领先的人工智能处理器制造商英伟达继续大量采购最先进的 H ...

Market Overview - The initial value of the September Markit manufacturing and services PMI in the U.S. showed a decline but remained in the expansion zone, with easing price pressures, leading to a muted market reaction [1] - Following comments from Federal Reserve Chairman Jerome Powell regarding high equity valuations, major U.S. stock indices experienced a decline, with the S&P 500 down 0.6%, the Dow Jones down 0.2%, and the Nasdaq down 1% [3][6] Federal Reserve Commentary - Powell's remarks served as a public warning about current market valuations, impacting investor sentiment and shifting it from exuberance to tension [4] - He emphasized the need to monitor inflation and employment data without making any clear commitments regarding policy actions for October, indicating a cautious approach [4] Technology Sector Impact - Interest-sensitive technology stocks were hit hardest, with the index of the seven major tech companies declining; Amazon fell by 3.04%, Nvidia by 2.82%, and Tesla by 1.93% [5] Semiconductor Industry Insights - Micron Technology reported strong earnings, with Q4 revenue reaching $11.3 billion, exceeding market expectations of $11.2 billion, and earnings per share of $3.03, surpassing analyst predictions of $2.84 [8][10] - The demand for memory chips, driven by AI, is expected to keep supply tight into next year, with Micron's CEO highlighting the company's competitive product portfolio and unique position as the only domestic memory manufacturer [9] - Micron is also leading in HBM (High Bandwidth Memory) chips, with significant agreements in place for HBM3e chips and plans to introduce HBM4 chips, addressing market concerns about profit margins [12] Pricing Trends in Semiconductor Industry - The semiconductor industry is experiencing a price surge, with Micron's performance reflecting broader trends across the supply chain, driven by AI demand [13] - Advanced chip prices are rising, with TSMC indicating a 50% increase in 2nm process pricing compared to 3nm, and Apple securing a 50% share of TSMC's 2nm capacity [15] - Major competitors like Samsung and SK Hynix have also raised prices for memory and flash products, with DRAM prices increasing by up to 30% [15]

Group 1 - HBM chips have become the standard for AI computing, with their core advantage stemming from the vertical stacking structure of DRAM chips [2] - The main chip stacking technology currently is Thermal Compression Bonding (TCB), which faces limitations as the number of stacked layers exceeds 16, affecting yield and signal integrity [2][4] - Hybrid bonding technology emerges as a revolutionary solution, allowing for direct copper-to-copper bonding between DRAM chips, enhancing interconnect density without the limitations of bumps [2][4] Group 2 - The semiconductor industry is shifting towards small chips and 3D integrated chip (3DIC) technology due to the slowdown of Moore's Law, making packaging a key factor in driving AI chip performance [4][6] - According to Yole Group, the evolution of chip bonding technology is moving towards hybrid bonding as the ultimate goal, with a projected market growth for hybrid bonding equipment to reach $397 million by 2030 [6][9] Group 3 - Hybrid bonding technology offers significant advantages over TCB, including a 15x increase in interconnect density, 11.9x speed improvement, and over 100x energy efficiency performance [9][10] - Despite the higher infrastructure costs, the cost per interconnect is reduced by 10 times with hybrid bonding, and it can lower the HBM stack temperature by 20% [9][10] Group 4 - Currently, no company has successfully achieved mass production of hybrid bonding equipment due to challenges such as existing TCB machines being sufficient, high precision requirements, and the high cost of hybrid bonding machines [12][21] - Predictions indicate that by 2030, the cumulative installation of hybrid bonding equipment will range from 960 to 2000 units, reflecting a 7% increase from previous forecasts [12][14] Group 5 - Major players in the hybrid bonding equipment market include Besi, which has seen significant revenue growth and strategic partnerships, particularly with Applied Materials [21][22] - South Korean companies like Hanmi Semiconductor and Hanwha Semitech are key competitors in the hybrid bonding space, with Hanmi holding a dominant market share in TCB machines [23][24] Group 6 - LG Electronics is entering the hybrid bonding equipment market through a national project aimed at developing HBM hybrid bonding machines, indicating a strategic focus on semiconductor equipment [25][26] - Samsung is also developing its own hybrid bonding machines through its subsidiary SEMES, aiming to reduce reliance on external suppliers [27] Group 7 - In China, companies like Tuojing Technology and Qinghe Crystal Semiconductor are making strides in hybrid bonding equipment, with Qinghe announcing the launch of the world's first dual-mode hybrid bonding equipment [29] - Besi predicts that the hybrid bonding market will reach €1.2 billion by 2030, driven by the transition from TCB to hybrid bonding technology [29]

Group 1 - Morgan Stanley reaffirms "buy" rating for ASML with a target price of €822, indicating that the worst may be over for the company [1] - ASML's second-quarter report in July was disappointing, but subsequent news in August and September showed improvement, suggesting a potential recovery in the logic and memory markets [1][2] - Key indicators since summer indicate a gradual improvement, with no new negative news reported, and strong performance from major companies like Nvidia and Broadcom [2] Group 2 - The risk of weak performance guidance for ASML in 2026 is diminishing, while trends for 2027 are expected to be strong [3] - The market anticipates positive trends following ASML's revenue guidance announcement on October 15, which could support the stock price [3] - ASML's current stock price remains below the historical average P/E ratio of 30-35 times, indicating it is still attractive for investors [3]

Core Viewpoint - The article discusses the evolution of Base Die in HBM (High Bandwidth Memory) technology, highlighting the shift from DRAM manufacturing processes to foundry processes as companies prepare for the HBM4 era, driven by the increasing performance demands of AI servers [1][3][4]. Group 1: Base Die Evolution - The global memory companies are significantly improving the Base Die structure in preparation for the mass production of the next generation HBM4, as its role becomes increasingly critical due to the exponential growth in AI computing [3][4]. - Base Die is described as the "brain" of HBM, responsible for signal processing and power efficiency, which determines the overall performance and stability of HBM [3][4]. Group 2: Manufacturing Process Changes - Samsung and SK Hynix have decided to transition the Base Die to a foundry process to address heat and signal delay issues, enhancing energy efficiency and performance for high-performance computing [4][5]. - Micron Technology has chosen to continue using existing DRAM processes for Base Die production until HBM4 mass production, while utilizing TSMC's foundry for HBM4E, which may lead to long-term competitive disadvantages [4][5]. Group 3: Strategic Differences - Industry insiders suggest that Samsung and SK Hynix's rapid shift to foundry technology is to meet the demands of major GPU clients like NVIDIA and AMD, ensuring technological leadership [5]. - Micron's conservative strategy aims to leverage existing DRAM infrastructure for cost competitiveness, but this may negatively impact its competitiveness with performance-focused clients [5].

Core Viewpoint - The company is actively pursuing a strategic transformation by expanding into the semiconductor and AI integration sectors while maintaining stable operations in its traditional business. Group 1: Traditional Business Optimization - The company focuses on optimizing its traditional business structure, primarily in the bulk decoration service sector, serving leading real estate enterprises with a strong delivery capability and standardized operation system [2] - In the first half of 2025, the company's revenue from decoration services was 123.34 million yuan, with the top five clients accounting for 76.52% of total revenue [2] - The company has established new wholly-owned subsidiaries to enhance its comprehensive service capabilities, aiming to create an integrated "design-construction-decoration" model [2] Group 2: Semiconductor Industry Layout - The company has made significant investments in the semiconductor industry chain, focusing on advanced packaging, HBM storage, and ABF substrates, aligning with national self-sufficiency and the computing power era [3] - A new subsidiary, Micro Packaging Technology, has been established to focus on semiconductor packaging equipment, and another subsidiary, Zhongtian Data, is dedicated to AI integration services and big data applications [3] - The company has indirect stakes in projects like the high-end substrate project, which is expected to begin trial production in the third quarter, and is making progress in HBM chip domestic substitution [3] Group 3: Overall Strategic Transformation - The company is entering a new stage of strategic transformation, driven by internal improvements in traditional business, new business initiatives, and enhanced internal controls and management quality [4]