分析师：甘坛焕 SAC:S1130525060003 gantanhuan@gjzq.com.cn 1 2026 年 1 月 27 日 产业研究周报 | 医药 AI 基建大时代， 研发转场， 医药格局将变 分析师: 赵海春 SAC:S1130514100001 zhaohc@gjzq.com.cn | 证券研究报告 | 核心要点 产业前沿 （1）前沿动向：AI 基建大时代，研发将从实验室转战 AI 平台，医药格局将变。 2026 年 1 月 21 日，在达沃斯世界经济论坛年会上，英伟达创始人黄仁勋做出重要 前瞻发言。对于医药医疗产业，要点有三：①AI 基建大时代来临。黄仁勋强调，AI 更应该被理解为一个多层系统。包括能源、芯片、云、人工智能模型本身以及应用 在内的五个层，这些层面必须协同扩展，并且已经引发了"人类历史上规模最大的基 础设施建设"。②研发从实验室转向 AI 平台。黄仁勋预测，药物研究将从传统实验 室转向人工智能平台，而制药巨头们已经开始迈出这一步。③AI 改变医药格局。我 们认为，随着各家药企 AI 融入度的改变，药企构架与产业格局将会受到 AI 应用这 一颠覆性的科技突破的影响而变化。（2）监 ...

Group 1 - The core viewpoint of the articles highlights the increasing demand for AI-driven cloud computing services, leading to a tightening supply-demand dynamic in the computing power sector, which is reflected in rising prices from major cloud providers [8][11]. - The "算力ETF" (Computing Power ETF) has seen significant inflows, with over 1.4 billion yuan accumulated in the last ten days, indicating strong investor interest in the sector [1]. - Major tech companies like Microsoft, Meta, and others have reported impressive cloud revenue, with Microsoft achieving $50 billion in cloud revenue for the second fiscal quarter, surpassing analyst expectations [4]. Group 2 - Clawdbot, an open-source AI assistant, has gained significant attention, with deployment support from Tencent Cloud and Alibaba Cloud, indicating a shift towards more personalized AI applications [3][6]. - The pricing strategies of leading cloud providers, such as Google and AWS, have been adjusted upwards, reflecting the increased costs associated with AI training and inference demands [8][9]. - The computing power supply chain is experiencing inflation, with projections indicating significant price increases for DRAM and NAND storage, as well as CPU products, driven by AI demand [9][10]. Group 3 - The rapid growth in token usage for AI models, particularly from OpenAI and Google, suggests a burgeoning market for AI applications across various sectors, with daily token calls reaching unprecedented levels [10]. - The anticipated rise in AI model applications is expected to further drive demand for computing resources, benefiting companies involved in hardware, cloud services, and AI development [11]. - The computing power ETF, 汇添富 (159273), is positioned to capture the growth opportunities in the AI-driven cloud computing market, covering a wide range of sectors including hardware, IT services, and data center operations [12].

责任编辑：刘万里 SF014 1月29日，外交部发言人郭嘉昆主持例行记者会。 法新社记者提问，据报道，中国已批准进口首批英伟达的H200人工智能芯片。外交部能否证实？对中 美关系来说，此举释放了什么信号？ 郭嘉昆 资料图。图源：外交部网站"我们已多次就有关问题作出回应，具体问题建议向中方主管部门进 行询问。"郭嘉昆说。 来源：北京日报客户端 记者：周经纬 郭嘉昆 资料图。图源：外交部网站"我们已多次就有关问题作出回应，具体问题建议向中方主管部门进 行询问。"郭嘉昆说。 来源：北京日报客户端 记者：周经纬 责任编辑：刘万里 SF014 1月29日，外交部发言人郭嘉昆主持例行记者会。 法新社记者提问，据报道，中国已批准进口首批英伟达的H200人工智能芯片。外交部能否证实？对中 美关系来说，此举释放了什么信号？ ...

仅 Meta(META.US)一家公司就披露了今年高达 1350 亿美元的支出雄心——这是商业领域规模最大的计 划支出之一。其供应商也做出了相应回应：周四，SK 海力士表示计划"大幅增加"资本支出，三星也表 示将加大存储芯片生产能力的投入。 与此同时，只要公司能够展示出增长，投资者就会继续回报这种雄心。微软(MSFT.US)在披露云计算服 务增长放缓后，股价下跌了 6.1%，而 Meta 则上涨了 6.6%。 本周，一系列行业巨头公布的财报强调了市场对 AI 硬件的需求是多么旺盛，并且这种势头很可能会延 续到 2026 年。 Meta、微软以及亚马逊(AMZN.US)和 Alphabet(GOOGL.US)等同行超大规模云服务商，正在推动全球范 围内对芯片、服务器和计算机的采购浪潮，这点燃了全球硬件供应商的热情，尤其是在亚洲。 智通财经APP注意到，全球最大的科技公司表现出对 AI 支出毫不减弱的迹象，这一创纪录的浪潮正推 动着三星电子和 SK 海力士等硬件供应商的发展。尽管人们对于人工智能需求是否具有持久性，以及能 否支撑起如此巨额的资本支出仍持怀疑态度。 存储芯片制造商正将生产线转向利润丰厚的高带宽内存( ...

今日市场分化明显，白酒、地产等传统蓝筹大涨，带领上证50指数大幅冲高，半导体、芯片等科技题材 走弱，科创50、创业板指跌幅明显。市场全天3.2万亿，较上个交易日放量超2000亿，两市下跌个股超 3500只。 据券商中国，值得注意的是，市场迎来了三大变局信号： 展望后市，东方证券认为，短期来看，股指震荡格局未变，但盘中受到调控迹象明显，大涨大跌皆不现 实，沪综指有望继续震荡上行，但春节前空间不大。 从个股看，两市上涨1803家，下跌3566家，107家涨幅持平。两市共89股涨停，共36股跌停。 截至收盘，沪指涨0.16%报4157.98点，成交14858亿元；深成指跌0.30%报14300.08点，成交17442亿 元。创业板指下跌0.57%，报3304.51点。 资金动向 今日主力资金重点抢筹广告营销、软件开发、白酒等板块，主力净流入居前的个股包括蓝色光标、贵州 茅台、科大讯飞等。 要闻回顾 1、世界黄金协会：2026年一季度中国市场黄金投资需求强劲态势或将延续 一是地产股和白酒齐飞，当板块轮动到地产和白酒这类补涨资产时，市场的结构性波动可能会加大。 二是包括黄金在内的大宗商品近期持续飙升，同样反映流动性的虚 ...

NVIDIA Corporation (NASDAQ:NVDA) is among Goldman Sachs’ top semiconductor stock picks. On January 26, NVIDIA Corporation (NASDAQ:NVDA) and CoreWeave expanded their long-standing collaboration with a new agreement to accelerate the build-out of more than 5 gigawatts of AI data centers, or AI factories, by 2030. This move, NVIDIA said, will support the growing demand for AI infrastructure. NVIDIA (NVDA) Invests $2B in CoreWeave, Expands AI Data Center Partnership As part of this expanded collaboration, NV ...

Core Viewpoint - The semiconductor industry is undergoing a historic transition as Nvidia has officially surpassed Apple to become TSMC's largest customer, driven by the surging demand for AI computing capabilities, indicating a fundamental restructuring of the global chip market [3][4]. Group 1: Nvidia's Rise and Market Dynamics - Nvidia's dominance as TSMC's top customer reflects a steep growth curve in AI-centric chip production, with Nvidia occupying more advanced process capacity than any other client [3][4]. - Nvidia's H100 and H200 data center GPUs, manufactured using TSMC's 4nm and 5nm processes, are priced over $30,000 each, significantly higher than smartphone processors, altering the semiconductor value chain's incentive structure [5][6]. - Over the past 18 months, Nvidia's wafer orders have grown exponentially, leading TSMC to expand its advanced packaging capacity to meet Nvidia's CoWoS requirements, which are critical for modern AI accelerators [6]. Group 2: Apple's Semiconductor Strategy - Apple's replacement by Nvidia does not indicate weakness but rather reflects a maturation of its product lineup and a more restrained hardware update strategy [7]. - Apple completed its transition to self-developed chips for its Mac product line in 2022, leading to a temporary spike in chip demand, which has since normalized [7]. - Despite a relative decline in its share of TSMC's revenue, Apple remains a key customer and is expected to continue using advanced processes for its upcoming A19 and M5 chips [7][8]. Group 3: TSMC's Capacity Allocation Challenges - TSMC faces increasing complexity in balancing the demands of major clients like Nvidia and Apple while also servicing other companies such as AMD and Qualcomm [9][10]. - TSMC's capital expenditure for 2024 exceeds $40 billion, aimed at expanding capacity in Taiwan, Arizona, Japan, and potentially Europe [10]. - The shift towards AI chip demand is influencing TSMC's technology development focus, necessitating a fundamental change in process optimization strategies to cater to high-performance computing [10]. Group 4: Competitive Landscape and Market Implications - Nvidia's strengthened relationship with TSMC occurs amid other foundries struggling to catch up in advanced processes, giving TSMC significant bargaining power [11]. - Geopolitical uncertainties complicate long-term capacity planning for Nvidia and TSMC, as prioritizing high-margin AI chip production may extend lead times and increase costs for other clients [11]. - Major tech companies like Amazon, Google, and Microsoft are accelerating their custom chip initiatives to reduce reliance on Nvidia and secure manufacturing resources amid tightening capacity [11]. Group 5: Future Trends in Semiconductor Manufacturing - The relationship between Nvidia, Apple, and TSMC is expected to continue evolving, with TSMC planning to enter mass production of 2nm processes by 2025 [12][13]. - The long-term trajectory of AI chip demand remains uncertain, with potential shifts in relative positions between Nvidia and Apple depending on market conditions and product innovations [12]. - TSMC's global expansion adds new dimensions to the relationship, as the willingness of Nvidia and Apple to shift core production to higher-cost, lower-risk U.S. facilities remains uncertain [13].

Core Insights - ASML is the only company globally producing extreme ultraviolet (EUV) lithography machines, holding a 90% market share in the lithography sector, and is expected to monopolize next-generation EUV technology [4][5][9] - The company's recent earnings report showed bookings exceeding analyst expectations, with EUV systems contributing significantly to its revenue [5][10] - ASML's advanced lithography machines are essential for the production of semiconductors, particularly for AI applications, with 99% of semiconductors being produced using ASML technology [5][10] Company Performance - ASML's net bookings for the fourth quarter of 2025 reached 13.2 billion euros, with EUV systems accounting for 7.4 billion euros [10] - The company sold 48 EUV systems in 2025, generating 11.6 billion euros in revenue [10] - ASML's share price increased by 36% last year and has surged another 32% since the beginning of 2023, reaching a valuation of half a trillion dollars [12] Market Dynamics - The price of ASML's advanced high NA EUV machines ranges from 320 million to 400 million euros, while the low NA EUV is priced around 220 million euros [11] - Companies like TSMC, Intel, and Samsung are currently experimenting with high NA EUV technology, which is expected to enter high-volume manufacturing by 2027-2028 [11][12] - Analysts predict continued stock price growth for ASML, with expected net sales for 2026 projected between 34 billion and 39 billion euros, surpassing the 32.7 billion euros achieved in 2025 [12]

Nvidia CEO Jensen Huang said on Thursday he hopes China will allow the U.S. technology giant to sell its powerful H200 artificial intelligence chip in the country and that the licence is being finalised. ...

Core Viewpoint - Nvidia's CEO Jensen Huang expresses hope for the Chinese government to approve the sale of the H200 AI chip in China, with the licensing process nearing completion [1][2]. Group 1: Nvidia's H200 Chip - The H200 is Nvidia's second most powerful AI chip and has become a significant point of contention in U.S.-China relations, with strong demand from Chinese firms [6]. - Nvidia has not yet received confirmation regarding the approval status from the Chinese government, which is still deliberating on the matter [5]. - The company is prepared to work with its manufacturing partner TSMC to manage supply and delivery if the H200 is approved [7]. Group 2: Chinese Market Dynamics - China has reportedly approved major tech companies like ByteDance, Alibaba, and Tencent to purchase over 400,000 H200 chips, but these approvals come with restrictive conditions [4]. - The Chinese government aims to balance the demands of its AI industry with the need to support its domestic semiconductor sector, which has contributed to the delay in import approvals [5]. - Despite U.S. approval for exports, the main barrier to shipments remains China's hesitation to allow imports of the H200 chip [6].