IT之家 12 月 3 日消息，研调机构 TrendForce 集邦咨询今日表示，五大 NAND 原厂今年三季度相关业务营收实现 16.5% 环比增长，总额接近 171 亿美元 （IT之家注：现汇率约合 1210.36 亿元人民币），其中 KIOXIA 铠侠营收相较二季度大幅扩张 33.1%。 | Rankings | Company | Revenue (US$M) | | | Market Share (%) | | --- | --- | --- | --- | --- | --- | | | | 3Q25 | QoQ (%) | 3Q25 | 2Q25 | | 1 | Samsung | 6,000.0 | 15.4% | 32.3% | 32.9% | | 2 | SK Group (SK hynix + Solidigm) | 3,525.6 | 5.7% | 19.0% | 21.1% | | 3 | Kioxia | 2,841.0 | 33.1% | 15.3% | 13.3% | | 4 | Micron | 2,423.0 | 15.4% | 13.0% | 13.5% | | 5 | ...

目前，OpenAI的采购仍主要通过微软、甲骨文等云合作伙伴间接完成，而不是通过9月意向书中规划的直接采购路径。 OpenAI"确实想改为直接采购"，Kress说， 英伟达首席财务官Colette Kress周二表示，公司预计到2026年的Blackwell和Rubin人工智能芯片5000亿美元预订额中，不包含目前与OpenAI就下 一阶段合作正在进行的任何工作。她同时透露，英伟达与OpenAI在9月公布的最多1000亿美元投资协议仍未达成最终版本，OpenAI有意转向直接采 购，双方仍在谈判最终协议。 英伟达首席财务官Colette Kress周二表示，到2026年Blackwell和Rubin人工智能芯片的5000亿美元预订额中，不包括该公司目前正就OpenAI下一阶段 协议所做的任何工作，并且透露英伟达与OpenAI之间此前公布1000亿美元的投资协议仍未最终敲定。 英伟达9月公布了一份向OpenAI投资的意向书，根据协议内容，英伟达将为OpenAI部署至少10吉瓦的英伟达系统，其规模足以为超过800万户美国家 庭供电。英伟达承诺随着每批次系统上线，将向OpenAI投资最多1000亿美元。这一消息曾推动A ...

点击小程序查看报告原文 Abstract 摘要 我们在再工业化开篇系列（一）、（二）中，聚焦美国产业回流进程，着重探讨美国在全球制造业版图中的定位，剖析其制造业"空心化"现象与供应链重 构挑战，并结合我们9月美国实地调研提供一线观察。 在本文中，我们通过宏观、微观多维数据分析当前美国基建和制造业景气度变化趋势，建议关注2026 年三条投资主线。主线一： 大型公共项目驱动，硬 基建建筑投资旺盛； 主线二： 时隔三年自动化设备周期触底回升，代表投产活动的工业设备投资指标回暖； 主线三： AI基础设施建设潮下的数据中心产 业链。 T正文ext 美国再工业化背景下，聚焦三条投资主线 在美国大型基建项目长周期建设、制造业回流、AI基建多重催化的背景下，我们建议关注2026年三条投资主线： 大型公共项目驱动，硬基建建筑投资旺盛，我们预计2026年非住宅其他领域如制造业、软基建的建筑投资也将全面回暖： 推动基建（如交通、电网）、 制造业厂房等投资， 受益环节包括工程基建EPC、工程机械、设备租赁、工业自动化及电气设备等产业链。 美国制造业回流带来再工业化需求，时隔三年自动化设备周期触底回升，代表工业生产活动的工业设备投资指 ...

Core Viewpoint - The article discusses Alibaba's organizational transformation from a decentralized "federal" structure to a more centralized model, emphasizing the need for unified infrastructure to compete effectively in the market. This shift is evidenced by significant changes in financial performance and strategic focus, particularly in the realm of instant retail and AI infrastructure. Group 1: Organizational Structure and Strategy - The nature of an organization's structure directly influences its product system, with a decentralized structure leading to modular and incompatible products [1] - Alibaba's "1+6+N" organizational reform aimed to create a federal system, but this approach has effectively ended, leading to a return to a centralized model [1][2] - The shift to a centralized structure is seen as necessary for addressing the challenges of building integrated physical and digital infrastructures [1][4] Group 2: Financial Performance - Taotian Group's adjusted EBITA dropped from 44.3 billion to 10.5 billion yuan, a decline of 76% year-over-year [2] - The significant profit drop is interpreted as a strategic decision to reallocate resources rather than a mere consequence of increased competition [3][4] - The centralized decision-making process allows for the redistribution of profits from profitable units to support critical areas like instant retail [4][7] Group 3: Instant Retail and AI Infrastructure - Instant retail is framed as a foundational infrastructure rather than just a business, focusing on long-term strategic investments rather than immediate profitability [6][8] - The integration of various business units is crucial for operational efficiency, enabling seamless collaboration across logistics, supply chain, and sales [8] - Alibaba's cloud intelligence group reported a 34% revenue growth, with AI-related products experiencing triple-digit growth for five consecutive quarters [8][9] Group 4: Capital Expenditure and Strategic Focus - High capital expenditure (CapEx) of 31.4 billion yuan indicates a strategic commitment to AI infrastructure, which is seen as a long-term investment [9][10] - The centralized approach to AI infrastructure is necessary to avoid data fragmentation and inefficiencies across different business units [10][11] - The transition from an "application factory" to a "smart power plant" reflects a strategic pivot towards building a unified AI capability [12] Group 5: Implications of Centralization - The centralization process results in the marginalization of non-core businesses, with a 25% revenue decline in "other" business segments [13][14] - The focus on core areas like e-commerce and cloud services signals a shift away from diversified operations towards a more streamlined infrastructure model [14][16] - The article highlights the challenges of executing this transformation, including potential internal confusion and inefficiencies during the transition [16][17]

2025 年 11 月 21 日 招银国际环球市场 | 睿智投资 | 行业研究 英伟达（NVDA US，未评级）再次交出了超预期的季度业绩。公司三季度收入大 幅增长至 570 亿美元，同比增长 62%，环比增长 22%，超出彭博一致预期/公司 指引 3%/6%。公司本季度收入环比增加 100 亿美元，显著超出 40-50 亿美元的历 史正常水平。Non-GAAP 毛利率为 73.6%，略低于彭博预期但小幅超出公司指 引。毛利率环比提升 1.0 个百分点，主要得益于数据中心产品组合升级、交货周 期缩短以及成本结构优化。管理层给出了强劲的四季度指引：预计收入将达到 650 亿美元，同比增长 65%，环比增长 14%，高于彭博预期的 620 亿美元。这意 味着公司下季度将实现 80 亿美元的大幅环比增长。四季度 non-GAAP 毛利率指 引为 75.0%，符合公司此前预计毛利率将在年底达到 75%左右的目标。对于 2027 财年，管理层预计毛利率将持续维持在 75%。 值得注意的是，英伟达在多重挑战下仍交出了出色的业绩，例如：1）受出口管 制，中国市场的收入贡献几乎忽略不计；2）市场对 AI 行业是否存在泡沫产生 ...

Investment Rating - The report maintains a "Buy" rating for the company with a target price of 100.0 RMB [6][5]. Core Insights - The company is poised to capitalize on the strong demand for AI infrastructure, achieving a balance between volume and profitability. The revenue from GPU AI servers is expected to continue its rapid growth, with Q3 revenue increasing over five times year-on-year and a strong outlook for Q4 [1][2]. - The company anticipates improvements in gross margin due to the ramp-up of new products, enhanced yield rates, and increased self-supply value of components. The outlook for Q4 remains optimistic, with expectations of significant revenue growth [1][3]. - The AI server and 800G switch businesses are experiencing exponential growth, with CSP AI server revenue expected to increase over four times year-on-year in Q4, and the 800G switch revenue projected to grow over 4.5 times [2][4]. Summary by Sections Revenue Growth - The company reported a year-on-year revenue increase of over three times for CSP AI servers and over five times for GPU AI servers in Q3. For Q4, CSP AI server revenue is expected to grow over four times year-on-year, while GPU AI server revenue may also exceed five times growth [2][4]. - The overall cloud computing business, including AI servers, is projected to grow by 83.6%, 103.3%, and 24.5% in 2025, 2026, and 2027 respectively, with its contribution to total revenue increasing significantly [2][15]. Profitability and Margin - The gross margin for the first three quarters of 2025 was 6.8%, consistent with the previous year. The company expects this margin to improve in Q4 due to successful product launches and cost efficiencies [3][5]. - Factors supporting margin improvement include successful mass production of new products, increased self-supply value of key components, and strong customer partnerships that enhance profitability [3][4]. Market Outlook - The report highlights a robust investment outlook for AI infrastructure, with North America's top five CSPs expected to invest over 390 billion USD in data centers by 2025. Global AI infrastructure capital expenditure is projected to exceed 1 trillion USD by 2030 [4][5]. - The company anticipates that the demand for 800G switches will continue to grow, with expectations of doubling shipment volumes annually, indicating a strong market for AI cluster construction [4][5]. Financial Projections - The company forecasts net profits of 353.2 billion RMB, 550.1 billion RMB, and 627.8 billion RMB for 2025, 2026, and 2027 respectively, with corresponding EPS of 1.78, 2.77, and 3.16 RMB [5][10]. - The projected revenue for 2025 is 892.2 billion RMB, with a year-on-year growth rate of 46.5% [10][15].

Core Viewpoint - Barclays Bank has downgraded Oracle's debt rating to "underweight," warning that its credit rating could potentially fall to BBB-, which is close to junk bond territory [2][4]. Financial Performance - Oracle's total interest-bearing debt has doubled over the past decade to $111.6 billion, with a debt-to-equity ratio of 500%, significantly higher than Amazon's 50% and Microsoft's 30% [4]. - The company has over $100 billion in unrecognized lease commitments off its balance sheet, raising concerns about its leverage levels [4]. - Oracle's cash reserves, currently around $11 billion, may be exhausted by November 2026, leading to refinancing needs [3][4]. Market Reaction - Following the downgrade, Oracle's stock opened down over 2%, reflecting immediate market concerns [1][2]. - The anxiety surrounding AI investments has also affected other tech giants like Google and Meta, which saw their stock prices decline sharply [6][9]. Industry Context - The bond spreads for hyperscale cloud service providers, including Oracle, Google, and Meta, have widened to 78 basis points, the highest level since market turmoil in April [9]. - Investors are increasingly worried about the rising leverage levels and potential overcapacity in the AI infrastructure market, despite these companies having substantial cash reserves [9].

Core Viewpoint - The market reaction to Meta's Q3 earnings report is considered excessive, primarily driven by concerns over capital expenditures and stock buyback policies, while the company's fundamentals remain strong [1][2]. Financial Performance - Meta reported Q3 revenue of $51.24 billion, a year-over-year increase of 26%, with Q4 guidance set between $56 billion and $59 billion [1][13]. - Operating cash flow for the past three quarters reached $79.6 billion, also reflecting a 26% year-over-year growth [3]. - Capital expenditures for the same period amounted to $48.3 billion, representing approximately 61% of operating cash flow [4]. Advertising Business - Key metrics in the advertising business showed improvement, with impressions up 14% and average ad prices increasing by 10%, leading to a 26% growth in overall ad revenue [2][11]. - The strong growth in advertising revenue is expected to continue, with projections indicating a growth rate above 20% as long as this trend persists [2][9]. Capital Expenditure and Stock Buybacks - Meta's capital expenditure guidance for 2025 was raised to a range of $70 billion to $72 billion, with expectations for significant growth in 2026 [4][5]. - The company has repurchased $26.32 billion in stock over the past three quarters, with an additional $25.03 billion remaining in authorized buybacks [5]. Market Valuation - Meta is currently the lowest-valued large-cap tech company, with a forward P/E ratio of 25.6 [2][18]. - The stock has seen a decline of 15.6% since the last analysis, leading to a potential buying opportunity as the stock is viewed as oversold [1][18]. Future Outlook - Analysts maintain an optimistic outlook for Meta, predicting the stock will stabilize around $600 before potentially reaching historical highs within 3-6 months [1][2]. - The company is expected to continue benefiting from its strong advertising business and the ongoing development of its AI infrastructure [1][9].

Group 1: Technology Sector - Global technology stocks have experienced increased volatility, with the MSCI Information Technology Index rising by 5.6% from October 11 to November 10, outperforming the MSCI Global Index which increased by 3.4% [1] - The valuation of US technology stocks remains high, with the Shenyin Wanguo Electronics and Semiconductor Indexes showing a month-on-month price-to-earnings ratio change of -10% and +13% respectively [1] - Storage prices are expected to continue rising, with strong DRAM prices anticipated to last at least until Q3 2026, and NAND prices expected to remain robust until at least Q3 2026, an extension from previous expectations of H1 2026 [1] Group 2: Semiconductor Industry - In September, China's semiconductor manufacturing equipment imports reached $5.76 billion, marking a year-on-year increase of 35%, continuing a trend of significant growth for four consecutive months [2] - TSMC reported a 17% year-on-year revenue growth in October, indicating strong performance in the semiconductor sector [2] - Investment recommendations suggest focusing on domestic semiconductor equipment and core targets for domestic substitution, as AI infrastructure construction in both overseas and mainland China is expected to continue growing rapidly through 2026 [2] Group 3: Automotive Sector - In October, retail sales of passenger vehicles in China slightly decreased by 0.8% year-on-year, totaling 2.24 million units, while cumulative sales from January to October increased by 7.9% year-on-year to 19.25 million units [3][6] - The penetration rate of new energy vehicles (NEVs) reached 57.2% in October, with domestic brands increasing their market share to 70.8% in the NEV segment [3][6] - Passenger vehicle exports continued to show strong growth, with a total of 568,000 units exported in October, representing a year-on-year increase of 27.7% [4][6] - The share of new energy passenger vehicle exports rose to 44.2%, with 250,000 units exported, reflecting a year-on-year increase of 104% [4][6] - Investment insights suggest that the adjustment of new energy vehicle purchase tax exemptions in 2026 may stimulate consumer purchases towards the end of the year, maintaining high sales momentum [6]

Core Insights - The Shanghai Stock Exchange Sci-Tech Innovation Board Chip Index has shown a strong increase of 1.73%, with notable gains from stocks such as Source Technology (up 11.59%) and Haiguang Information (up 5.09%) [1][4] - The Sci-Tech Chip ETF (588200) has risen by 42.34% over the past three months, indicating significant investor interest and market momentum [1][4] Market Performance - The Sci-Tech Chip ETF experienced a turnover rate of 2.79% with a transaction volume of 1.14 billion yuan [4] - Over the past two weeks, the ETF's scale has increased by 255 million yuan, ranking first among comparable funds [4] - In the past week, the ETF's shares grew by 39.6 million, also leading in new share issuance among similar funds [4] - The ETF recorded a net inflow of 34.36 million yuan recently, with a total of 847 million yuan net inflow over the last five trading days [4] Historical Performance - As of November 5, the Sci-Tech Chip ETF has achieved a net value increase of 101.46% over the past two years, ranking 31st out of 2380 index stock funds [4] - The ETF's highest single-month return since inception was 35.07%, with the longest streak of consecutive monthly gains being four months and a maximum cumulative increase of 74.17% [4] Top Holdings - The top ten weighted stocks in the Sci-Tech Chip Index account for 60.55% of the index, with Haiguang Information and Cambricon leading the list [4][7] - The top stocks include Haiguang Information (11.09%), Cambricon (8.59%), and SMIC (9.58%) among others [7] Industry Outlook - Datong Securities maintains a positive outlook on the cyclical opportunities arising from storage price increases, driven by AI infrastructure development [5] - The demand for storage and computing power is expected to remain strong, leading to a tight supply-demand balance in the storage industry [5] - Semiconductor equipment demand is anticipated to rise as manufacturers increase capital expenditures to meet growing storage needs, particularly in critical areas such as etching and deposition [5]