Core Viewpoint - The U.S. stock market is experiencing a cautious pause after a strong rebound, with the S&P 500 index slightly retreating from its historical high as investors await key economic data [1] Group 1: Market Dynamics - The market is facing a tug-of-war between recalibrated corporate earnings expectations and an escalating memory chip supply crisis, which is reshaping the landscape of winners and losers in the tech sector [1] - The ongoing earnings season has highlighted supply chain tensions, particularly the surge in memory chip prices, which is threatening profit margins across various industries from consumer electronics to automotive manufacturing [1][4] - Investors are trying to assess the duration of this supply bottleneck and its long-term impacts on inflation and corporate earnings, leading to a pause in trading activity ahead of significant economic data releases [1] Group 2: Performance Disparities - The recent relentless rise in memory chip prices has created a clear divide in the stock market, with a 10% decline in the Bloomberg index tracking global consumer electronics manufacturers since late September, while memory manufacturers, including Samsung, have surged approximately 160% [3] - Companies that can secure supply, raise product prices, or redesign products to reduce memory usage are being evaluated by fund managers and analysts for their ability to navigate this pressure [3] Group 3: Corporate Warnings - Frequent warnings from companies regarding memory shortages and pricing issues are becoming common in earnings reports and conference calls, with Honda and Qualcomm highlighting supply risks that could hinder production [4] - PC manufacturers are facing the most significant impact, with Lenovo and Dell seeing stock declines of over 25% from their peak last October, as concerns about rising chip prices dampen PC demand [4] Group 4: AI-Driven Supercycle - The memory price surge has become a headline issue this earnings season, with concerns about the timeline of supply tightness being questioned, particularly due to massive investments in AI infrastructure by major U.S. companies [6] - The shift in capacity from traditional DRAM to high-bandwidth memory driven by AI has led to a "supercycle," breaking the typical boom-and-bust pattern of memory supply and demand [6] - Despite weak demand for end products like smartphones and cars, DRAM spot prices have skyrocketed over 600% in recent months, with memory manufacturers emerging as winners in the tech sector [6][7]

芯片领域，传出多则重磅消息！ 2月10日下午，芯片代工巨头台积电披露的数据显示，该公司1月份营收同比增长36.8%，创下数月来最快增 速，凸显出全球人工智能支出依然强劲。 周二美股盘前，台积电股价直线拉升，涨幅超过3%。 另外，有外媒报道称，美国特朗普政府计划在即将出台的芯片关税中，给予亚马逊、谷歌、微软等大型科技企 业豁免，以支持它们建设推动人工智能（AI）发展的数据中心。 台积电业绩大增 台积电今日（2月10日）公布1月营收报告，当月合并营收约为4012.55亿新台币，较上月增加了19.8%，较去年 同期增加了36.8%。 该数据披露后，台积电股价在美股盘前大幅拉升。截至券商中国记者发稿时，台积电股价 涨幅仍超过3%。如果在美股周二正式开盘后维持这一涨势，台积电的股价将再度刷新历史新高。 包括微软、亚马逊、Meta、甲骨文公司和Alphabet在内的科技巨头，正计划在2026年总计投入超过6000亿美元 的资本支出。对此，黄仁勋强调，"人类历史上规模最大的基础设施建设"正在展开，其背后动力是对算力 的"极度旺盛"的需求，AI公司和超大规模云服务商正利用这些算力来创造更多收入。 1月中旬，台积电发布的季度业 ...

Group 1 - A-shares experienced narrow fluctuations, with the Sci-Tech Innovation Board performing well, particularly in AI applications and computing power chips [1][25] - The Hong Kong stock market saw a significant rebound in the pharmaceutical sector, with the Hong Kong Innovation Drug ETF (520880) reaching a peak increase of 3.86% and closing up 2.9% [3][29] - The Hong Kong Medical ETF (159137) also performed well, gaining 2.1% and achieving six consecutive days of increases [3][29] Group 2 - The innovation drug sector is benefiting from a surge in business development (BD) overseas and commercial success, with major collaborations announced, including a partnership between Innovent Biologics and Eli Lilly worth up to $8.85 billion [6][29] - Over 70% of innovative drug companies are expected to achieve positive revenue growth in 2025, with notable companies like BeiGene reporting revenues exceeding 36 billion yuan [7][30] - The Chinese innovative drug industry is entering a phase of commercial success and internationalization, with a focus on the entire industry chain, including CXO and research services [7][30] Group 3 - ByteDance's recent releases, including the Seedance 2.0 video generation model and the Seedream 5.0 image generation model, have positively impacted the Sci-Tech Artificial Intelligence ETF (589520), which saw a price increase of 1.81% [10][31] - The ETF's composition includes significant holdings in domestic AI companies, with ByteDance accounting for over 29% of the index [11][35] - Analysts are optimistic about AI investment opportunities, citing increased capital expenditures from tech giants and a growing demand for AI applications [11][35] Group 4 - The semiconductor and storage chip sectors are experiencing a "super cycle," with companies like Chipone Technology reaching historical highs [15][38] - The global memory price is expected to surge by 80% to 90% in Q1 2026, driven by AI demand, with the storage market projected to reach $551.6 billion [17][39] - The semiconductor equipment industry is also on the rise, with total sales expected to grow by 26% in 2026, indicating a strong market outlook [17][40]

Core Insights - The AI industry is shifting from a training-centric model to an application-oriented inference model, marking a "revolution in reasoning" [1][3] - AI chip company Aixin Yuan Zhi Semiconductor has officially listed on the Hong Kong Stock Exchange, becoming the first Chinese edge AI chip stock [1][2] - Aixin Yuan Zhi has developed five generations of SoCs, focusing on terminal computing, smart vehicles, and edge AI inference [1][2] Industry Trends - The demand for inference computing power is expected to grow exponentially, surpassing training computing power as AI applications become more widespread [1][3] - The edge AI chip market is gaining significant attention due to its ability to meet cost, power consumption, and real-time response requirements [1][3][5] - The transition to edge computing is driven by privacy concerns and the need for low-latency processing, leading to new application scenarios [3][5] Company Performance - Aixin Yuan Zhi has achieved rapid revenue growth, with projections showing an increase from 50.2 million yuan in 2022 to 473 million yuan in 2024, representing a compound annual growth rate of over 200% [2] - The company has established itself as a top player in the visual edge AI inference chip market, holding a 6.8% market share and leading the mid-to-high-end segment with a 24.1% share [2][6] - As of September 30, 2025, Aixin Yuan Zhi has delivered over 165 million SoC chips, with significant contributions from its visual terminal computing and smart vehicle segments [6][7] Technological Advancements - Aixin Yuan Zhi's unique vertical integration strategy includes developing a complete technology stack from self-researched IP to core SoCs [6] - The company’s key technologies include the AXNeutron mixed-precision NPU and AXProton AI-ISP, which enhance performance and ensure high-quality input for inference tasks [6] - The company is positioned to capitalize on the rapidly expanding edge AI market, focusing on cost-effectiveness and system-level efficiency [5][6]

Core Viewpoint - Cisco has launched the Silicon One G300, a 102.4 Tbps switch chip designed for large-scale AI clusters, marking a significant step in transforming networks into AI innovation platforms [1][2]. Group 1: Product Features and Innovations - The Silicon One G300 chip supports high-performance, high-security, and high-reliability for distributed AI clusters, setting a new benchmark for AI backend networks [2]. - The Intelligent Collective Networking technology in G300 enhances performance and profitability by improving network utilization by 33% and reducing job completion time by 28% compared to non-optimized path selection [2]. - The new Cisco N9000 and Cisco 8000 systems, powered by the G300 chip, utilize liquid cooling technology and high-density optical modules to achieve significant efficiency improvements, including nearly 70% energy efficiency gains [5]. Group 2: Scalability and Programmability - Cisco Silicon One architecture is highly programmable, allowing upgrades post-deployment to support new network functions, thus protecting long-term infrastructure investments [3]. - The introduction of the P200 system product line expands capabilities for cloud, enterprise, and service providers, enhancing cross-data center interconnectivity and routing capabilities [6]. Group 3: Management and Observability - The Nexus One platform integrates chips, systems, optical modules, software, and programmable intelligence into a unified management solution, simplifying network operations across multiple sites [7]. - The integration of native Splunk platform capabilities into Nexus One allows for job-aware visibility, linking network telemetry data with AI workload behavior, crucial for environments with high compliance requirements [8].

记者丨杨坪 编辑丨巫燕玲 2026 年 2 月 9 日 ， 被 业 界 誉 为 "IC 设 计 海 归 第 一 人 " 的 科 学 家 杨 崇 和 ， 带 着 澜 起 科 技 （06809.HK）登上了港交所的上市大厅。 这是他继2013年登陆纳斯达克、2019年登陆科创板后，第三次带领企业完成IPO。 从电视机顶盒芯片起步，到做大内存接口芯片， 杨崇和一路将澜起科技推向内存互连芯片全 球第一的宝座，公司2024年市占率高达36.8%。 近年来，AI需求带动全球存储市场大爆发，澜起科技市值、业绩双突破。 在此轮港股IPO 中，澜起科技以每股106.89港元的上限定价，所得款项净额69.05亿港元，公开发售接获707.3 倍认购。 此外，澜起科技还引入了阵容强大的基石投资者，JPMIMI、UBS AM、Yunfeng Capital、 Alisoft China、华勤通讯、中邮理财等17名全球基石投资者合共认购3282.8万股，占发售总数 的49.82%。 上市首日，澜起科技大涨63.72%，公司总市值已超过2000亿港元。2月10日，澜起科技再涨 1.71%，报178港元/股，市值突破2400亿港元。 海归博 ...

2026年2月10日，边缘计算AI芯片企业爱芯元智半导体股份有限公司(以下简称"爱芯元智")(0600.HK)正式在香港交易所主板挂牌上市。据了解，爱芯元智 发行价为每股28.2港元，此次发售1.05亿股，募资总额29.61亿港元。 本报记者袁传玺 值得关注的是，爱芯元智自主研发的"爱芯通元"混合精度NPU与"爱芯智眸"AI-ISP，已形成具备高兼容性、高能效的AI推理技术平台。该平台支持 Transformer及CNN等主流AI模型架构，能够在资源受限的边缘与终端设备上实现高效AI推理，显著降低部署成本与延迟。 事实上，边缘计算是"云—边—端"协同体系的关键环节，市场正高速增长。根据灼识咨询的数据，全球边缘及终端AI推理芯片市场规模已于2024年达到 3792亿元，预计至2030年将扩增至16123亿元，年复合增长率达27.3%。 在行业高速发展的背景下，目前爱芯元智已成功开发并商业化多代SoC产品，覆盖数十种型号，截至2025年9月30日，公司累计SoC交付量突破1.65亿颗。 其中，视觉终端计算SoC超过1.57亿颗；边缘AI推理芯片8850系列自2023年推出后出货量快速增长，2024年超过10万颗； ...

知情人士透露，华尔街大型私人信贷公司之一阿波罗全球管理公司（Apollo Global Management） 即 将达成一笔交易：向一家投资工具提供约34 亿美元贷款，用于购买英伟达芯片，并租赁给刚与 SpaceX 合并的埃隆・马斯克旗下 xAI。 这笔交易将是阿波罗第二次大手笔投资用于向 xAI 租赁芯片的投资工具。去年 11 月，阿波罗已提供过 一笔类似的35 亿美元贷款。该人士称，阿波罗还计划对新设立的投资工具进行股权投资，该工具目标 是合计募集53 亿美元的股权与债务资金。 这笔投资最快将于本周敲定，目的是缓解 xAI 的部分资金压力。马斯克正雄心勃勃地打造全球最大规 模的数据中心，用于训练 AI 模型。长期投资马斯克旗下公司的Valor Equity Partners 正在安排此次交 易。 该投资工具是 Valor 更大规模计划的一部分：计划募集200 亿美元的股权与债务，用于采购 AI 芯片并部 署在 xAI 数据中心。 过去，包括格拉西亚斯在内的 Valor 合伙人曾在关键时期进驻马斯克旗下公司提供支持：包括特斯拉产 能危机期间、以及马斯克重组推特（后更名为 X）之后。 管理资产规模超9000 ...

来源：格隆汇APP Astera Labs((ALAB.US)盘前继续上涨近5%，该股此前2日累计大涨超30%。消息上，Astera Labs将于美 东时间2月10日盘后公布Q4财报，市场重点关注Astera Labs在亚马逊上调资本开支背景下，其Aries PCIe Retimer及Scorpio Switch产品线放量进展。GAAP净利润市场一致预期是5124万美元，同比增长107%， 环比下滑44%，公司指引上限是3660万美元。 ...

1. Report Industry Investment Rating - Not provided in the given content 2. Core Viewpoints of the Report - The report focuses on constructing multiple ETF portfolios, including macro industry, macro + momentum industry, core - satellite, and trinity style rotation portfolios, aiming to find better investment opportunities by combining macro factors, momentum factors, and style rotation [4][5]. - Different industries have different sensitivities to economic, liquidity, and credit factors. For example, traditional cycle industries are sensitive to the economy, TMT is sensitive to liquidity, and consumption is sensitive to credit [4]. 3. Summary According to Relevant Catalogs 3.1 ETF Portfolio Construction Methods 3.1.1 Based on Macro Method - Calculate macro - sensitivities of broad - based, industry - themed, and Smart Beta ETFs based on economic, liquidity, and credit variables. Combine with momentum indicators for complementary analysis [4]. - Traditional cycle industries are suitable for economic up - periods, TMT for weak - economy but loose - liquidity periods, and consumption for credit - expansion periods. State - owned enterprises and ESG - related themes have low sensitivities to liquidity and credit [4]. - Construct three ETF portfolios (macro industry, macro + momentum industry, and core - satellite) and adjust positions monthly [4]. 3.1.2 Trinity Style Rotation ETF Portfolio Construction - Build a medium - to - long - term style rotation model centered on macro - liquidity, and compare it with the CSI 300 index [5]. - Construct three types of models (growth/value rotation, market - cap, and quality models) by screening macro, fundamental, and market - sentiment factors. The model has 8 style - preference results [5]. - Select ETFs with high exposure to the target style, control industry exposure, and set allocation limits to get the final ETF allocation model [5]. 3.2 Macro Industry Portfolio - Select industry - themed ETFs with over 1 - year establishment and over 200 million current scale. Calculate sensitivity scores of economic, liquidity, and credit factors monthly, adjust scores according to the latest indicators, and sum them up. If liquidity and credit deviate significantly, remove the liquidity score. Select the top 6 industry - themed indices and corresponding largest - scale ETFs for equal - weight allocation [6][7]. - Currently, with falling economic leading indicators, loose liquidity, and tightened credit, the portfolio is biased towards TMT and consumption. The February positions are shown in Table 1 [8]. - The portfolio has large fluctuations and outperformed the benchmark significantly in January [11]. 3.3 Macro + Momentum Industry Portfolio - Combine macro and momentum methods to address the left - side nature of macro - based strategies (low win - rate but high odds). Use clustering to group industry - themed indices and select the highest - rising product in each group in the past 6 months for equal - weight allocation [12]. - The momentum - selected industries still have a high proportion of cyclical industries. The February positions are shown in Table 3 [16]. - The portfolio has performed well this year and outperformed the CSI 300 significantly in January [17]. 3.4 Core - Satellite Portfolio - Design a "core - satellite" portfolio with the CSI 300 as the core to address the high volatility and rapid industry rotation of industry - themed ETFs [19]. - Calculate macro - sensitivities for broad - based, industry - themed, and Smart Beta ETFs, construct three stock portfolios, and weight them at 50%, 30%, and 20% respectively [19]. - The current allocation of broad - based ETFs is biased towards the Sci - tech Innovation Board and the ChiNext. The portfolio has performed stably, outperforming the benchmark in most months except December, and had significant excess returns in January 2026 [23][24]. 3.5 Trinity Style Rotation ETF Portfolio - The model currently favors the small - cap growth - high - quality segment. The factor exposures and historical performance are shown in Table 7 [26]. - The February positions are shown in Table 9 [31]. - The portfolio has achieved certain excess returns, especially in some months such as August 2025 and January 2026 [29].