我们美国半导体团队的同事，在近期 4QCY25 财报电话会议关于前沿逻辑 / 代工及存储芯片支出强劲的评论 后，。 我们美国半导体团队的同事，在近期 4QCY25 财报电话会议关于前沿逻辑 / 代工及存储芯片支出强劲的评论 后，。 因此，我们将此次修正视为对欧洲半导体企业（ASML、ASMI 和 BESI）的积极信号。 因此，我们将此次修正视为对欧洲半导体企业（ASML、ASMI 和 BESI）的积极信号。 美国半导体团队目前预计，2026/2027/2028 年 WFE支出分别为（同比增长 + 20%/+21%/+9%），而此前预期为 124 美国半导体团队目前预计，2026/2027/2028 年 WFE支出分别为（同比增长 + 20%/+21%/+9%），而此前预期为 1240 亿美元 / 1320 亿美元 / 1440 亿美元。 我们注意到，美国团队更新的近期 WFE预测，是由存储芯片厂商在供需极度源头信息加微WUXL7713紧张环 境下的强劲产能扩张驱动的，尤其是用于 AI 加速器的先进存储芯片（如 HBM）。 尽管如此，我们强调，代工需求的强劲预期支撑了美国团队对中期的更高 WFE估计，因为台积电 ...

Core Insights - Nvidia's CEO Jensen Huang hosted a rare dinner for engineers from SK Hynix, highlighting the strategic importance of the next-generation high-bandwidth memory HBM4 for Nvidia's AI chip business [1][4] - The dinner signals Nvidia's view of HBM4 as a key differentiator for its upcoming AI accelerator, Vera Rubin, which is set to launch in the second half of this year [3][5] Group 1: HBM4 Specifications and Market Dynamics - Nvidia has set stringent specifications for HBM4, requiring a running speed of over 11 Gbps and a bandwidth exceeding 3.0 TB/s, which is over 30% higher than AMD's requirements [3][6] - SK Hynix has secured over 55% of the HBM supply allocation for this year, while Samsung and Micron hold approximately 20% each [3][6] - Samsung has begun shipping HBM4 products with a running speed of 11.7 Gbps and a bandwidth of 3.3 TB/s, marking a competitive entry into the HBM4 market [6][7] Group 2: Strategic Partnerships and Execution - Huang's personal involvement in the dinner is seen as a significant gesture of recognition and encouragement for SK Hynix, emphasizing the importance of timely delivery and performance optimization [4][9] - The upcoming months will be critical for SK Hynix to maintain its leading position in the competitive landscape of HBM4 supply [9]

Group 1 - Samsung Electronics has supplied Qualcomm with samples of the next-generation low-power DRAM, LPDDR6X, which is expected to be commercially available in the second half of 2027 [1] - The LPDDR6X samples were provided even before the previous generation, LPDDR6, has officially launched, indicating a tight development schedule for Qualcomm's chip [1][2] - Qualcomm's request for LPDDR6X samples is closely related to its development of next-generation semiconductors, particularly for its AI accelerator "AI250," set to launch in 2027 [1][2] Group 2 - LPDDR6X is expected to further enhance performance over its predecessor, LPDDR6, although specific specifications are not yet confirmed due to the JEDEC standard still being under development [2] - Qualcomm aims to reduce its reliance on mobile application processors (AP) by accelerating its AI accelerator business, with plans to launch the AI200 this year and the AI250 next year, both focusing on AI inference tasks [2][3] - The AI250 is anticipated to require over 1000GB of LPDDR capacity, while the AI200 is expected to use 768GB [3] Group 3 - LPDDR technology, traditionally used in smartphones, laptops, and automotive electronics, is now being adopted by AI semiconductor companies like NVIDIA for inference chips, expanding its application range [1][2] - The bandwidth of LPDDR6X is expected to significantly increase compared to LPDDR5X, enhancing AI computing performance [2]

关于两家公司在HBM供应方面谁领先，各方争论不休，而业界的关注点也转向了谁能率先完成准 备工作。鉴于韩国半导体公司已在HBM4竞争中占据优势，分析认为，重点应该放在长期供应稳定 性和产能上。 如果您希望可以时常见面，欢迎标星收藏哦~ 由于三星电子和 SK 海力士几乎成为英伟达下一代 AI 加速器中高带宽内存 4 (HBM4) 的唯一供应 商，业内人士和业外人士对供应"份额"和"第一"的说法仍然存在分歧。 据业内人士2月9日透露，三星电子和SK海力士正在协调其HBM4的出货计划，以配合英伟达下一 代AI加速器"Vera Rubin"的发布计划。英伟达预计将于下个月的GTC 2026技术大会上首次发布其 下一代AI加速器"Vera Rubin"。 Vera Rubin 将中央处理器 (CPU)"Vera"和图形处理器 (GPU)"Rubin"集成于单一系统中，与现有 的基于 Blackwell 的产品相比，其推理性能显著提升。由于其设计目标是运行大规模 AI 模型，因 此超高带宽内存 HBM4 几乎是必不可少的。 据报道，三星电子计划本月率先向客户交付HBM4显存。该公司强调其技术竞争力，并指出该产品 拥有业界领 ...

Core Viewpoint - AMD's stock price continues to decline, with a drop of nearly 16%, marking the largest decline since 2018 [1] Financial Performance - AMD reported Q4 revenue exceeding $10 billion, achieving over 30% year-on-year growth [1] - EPS growth has slowed but exceeded analyst expectations by nearly 16% [1] - Data center revenue reached a record $5.4 billion [1] - Revenue from MI308 chip sales in China was approximately $390 million [1] - Export restrictions are expected to impact 2025 with around $440 million in inventory and related costs [1] Guidance and Market Expectations - Q1 revenue guidance midpoint is $9.8 billion, slightly above consensus expectations but below the optimistic forecast of over $10 billion [1] - Analysts note that despite exceeding revenue and profit expectations, the overall Q1 guidance is perceived as slightly disappointing [1] - The market anticipates faster market share gains and higher short-term shipments in AI accelerators, with any conservative pace viewed as "below expectations" [1]

记者丨张嘉钰 编辑丨黎雨桐 2月4日，美股开盘涨跌不一，截至23:43，道指涨0.66%，标普500指数跌0.1%，纳指跌0.71%盘中一度跳水至1%。 | 名称 | 现价 | 涨跌 | 涨跌幅 | | --- | --- | --- | --- | | 道琼斯丁 平均 | 49567.49 | 326.50 | 0.66% | | 纳斯达克指数 | 23090.17 | -165.02 | -0.71% | | 标普500 | 6911.02 | -6.79 | -0.10% | | 纳斯达克100 | 25094.49 | -244.13 | -0.96% | | 万得美国科技七巨头指数 | 64950.28 | -311.81 | -0.48% | | 万得中概科技龙头指数 | 4082.22 | -121.14 | -2.88% | | 纳斯达克中国金龙指数 | 7508.31 | -101.70 | -1.34% | | 万得中概股100 | 2736.93 | -58.58 | -2.10% | | 万得美国科技指数(CNY) | 10802.17 | -124.15 | -1.14% | | 排 ...

Group 1 - AMD's Q4 revenue exceeded $10 billion, marking a record high with a year-over-year growth rate of over 30% [1] - EPS growth slowed but was approximately 16% higher than analyst expectations [1] - Data center revenue reached a record $5.4 billion, while MI308 chip sales in China generated about $390 million [1] Group 2 - AMD's Q1 revenue guidance midpoint is $9.8 billion, slightly above consensus expectations but below the optimistic forecast of over $10 billion [1] - Analysts noted that despite the strong revenue and profit performance, the market perceives the Q1 guidance as somewhat disappointing [1] - There are expectations for AMD to gain market share in AI accelerators with higher short-term shipment volumes, and any conservative approach may be viewed as "below expectations" [1]

Core Viewpoint - Goldman Sachs highlights that Baidu's chip division, Kunlun Chip, has officially submitted its listing application in Hong Kong, marking a significant step in unlocking asset value for Baidu [1][2]. Group 1: Listing and Ownership Structure - Kunlun Chip submitted its listing application to the Hong Kong Stock Exchange on January 1, 2026, through a global offering that includes both public and institutional placements [2]. - As of August 2025, Baidu holds a 59% controlling stake in Kunlun Chip, which is expected to remain a subsidiary post-listing, allowing Baidu to unlock market value while maintaining control over its core computing infrastructure [2]. Group 2: Financial Projections - Goldman Sachs projects that Kunlun Chip's revenue will reach approximately 6.5 billion RMB in 2026, with external sales expected to grow significantly, accounting for 14% of total cloud sales [7]. - The revenue structure is anticipated to shift towards more recurring and high-value subscription services, which could represent 43% of Baidu Cloud's total revenue by 2026 [7]. Group 3: Valuation Insights - If Kunlun Chip is valued similarly to Cambricon at a 40x price-to-sales ratio, Baidu's 59% stake could be worth up to $22 billion, representing 45% of Baidu's current market capitalization [1][12]. - A conservative valuation range for Kunlun Chip's 100% equity is estimated between $5 billion and $18 billion, translating to a value of $3 billion to $11 billion for Baidu's stake, which is 6% to 23% of its current market cap [12]. Group 4: Competitive Landscape - Comparisons with peers like Cambricon and Moore Threads indicate that these companies are trading at 40x to over 100x their projected 2026 sales, suggesting significant potential for Kunlun Chip's valuation [8]. - Goldman Sachs notes that Kunlun Chip's sales are expected to surpass those of Moore Threads and Wallen Technology by 2026, although it may lag behind Cambricon [8]. Group 5: Investment Implications and Catalysts - Goldman Sachs maintains a "Buy" rating on Baidu with a target price of $155, indicating confidence in Baidu's transition from traditional advertising to non-search business driven by AI [9]. - Key catalysts to watch include announcements regarding new orders or product upgrades for Kunlun Chip, updates on shareholder return policies, and progress on the listing in Hong Kong [13].

智通财经APP获悉，美国老牌芯片巨头英特尔(INTC.US)股价在周三美股盘前交易中走低，一度大跌超 5%。此前据两位知情人士向媒体透露，"AI芯片霸主"英伟达(NVDA.US)在实际评估了使用英特尔18A 先进制程技术如何制造其中一款AI芯片之后，并未继续向前推进，而是选择暂停对英特尔18A制程工艺 的测试。 全球两大AI芯片设计领军者(在芯片行业被称作fabless)——即专注于AI GPU算力体系的英伟达以及专注 于AI ASIC算力技术路线的博通(AVGO.US)早在三月初就有媒体曾曝出，这两大芯片设计领军者与英特 尔进行了高端芯片制造测试。 然而，自那以后，英特尔的18A芯片技术持续面临挑战，因为初期的那些客户测试结果不乐观导致了客 户们对于英特尔技术层面的不满，并且有可能继续将芯片代工/制造产能全线投入于英特尔在芯片制造 领域的最强竞争对手台积电(TSM.US)。 英特尔当前迫切需要大客户来撑起芯片代工业务 对于英特尔而言，掌握领先全球且极度适合量产的先进制程芯片制造技术可谓至关重要。这家美国老牌 芯片巨头甚至不惜购置更加昂贵的High-NA EUV光刻机——这也是阿斯麦所交付的全球第一台High ...

Core Insights - Google's TPU has gained significant recognition, leading to claims that it has surpassed NVIDIA in the market, prompting NVIDIA to assert its continued dominance in GPU technology [1] Market Share and Competition - Bank of America projects NVIDIA's market share will decline from approximately 85% to around 75% due to increased competition, particularly from Google's TPU, although this shift will occur gradually [2][6] - The current supply chain constraints and NVIDIA's scale advantages hinder competitors from rapidly capturing market share [7] AI Model Development - The competition in large language models (LLMs) is described as a long-term marathon, with recent releases from Google and Anthropic indicating a dynamic landscape [4] - Google's TPU has been integral to the training of its Gemini models, with plans to potentially lease TPU to Meta by 2026, which could intensify competition for existing GPU suppliers [4] Company Valuations - AMD's target price is set at $300, reflecting anticipated growth in AI and CPU market share [9] - Broadcom's target price is $400, supported by strong earnings growth and profitability in the semiconductor sector [10] - NVIDIA's target price is $275, justified by its leading position in the rapidly growing AI computing market [11]