大摩闭门会：中国 AI GPU 前景展望以及台积电最新资本 支出预期; 上调阿里巴巴为互联网首选 260312_原文 2026 年 03 月 13 日 17:25 发言人 00:00 各位上午好，今天是 3 月 13 日星期五，欢迎来到摩根森周五的新经济板块热点前三在线直播。我是查理， 请注意本次的会议内容与您的问题将会被录音。在本次视频直播中我将会与 dac 以及 john 一起讨论最新的 两个趋势。 发言人 00:24 一个是中国 AI 的前景。Gary 的团队最近上调阿里巴巴成为互联网首选，其中一个很重要的原因就是 AIGP 国产化的推动。从我们半导体角度团队的角度来讲的话，我们也会提供一些从供应商从这个行业竞争上面 有什么样的 perspective。最后由我会再回到这个 global 的 AI 生命的手牌权。最近我们也上调了台积电 资本支出，就是说面向全球的 AI 的印刷刷式，我们还是蛮正面的。 发言人 01:06 好，我想因为九月刚好也上线了，那我们是不是还是从 90 开始？因为先讲一下这个客户端终端的需求的情 况，然后 dac 再讲一下怎么悲伤 china 和 AIKP 要形成我们对地面的看法。 ...

The Great AI Silicon Shortage AI 芯片大短缺 TSMC N3 Wafer Shortages, Memory Constraints, Datacenter Bottlenecks, Supply Chain Wars Winner 台积电 N3 晶圆短缺、内存限制、数据中心瓶颈、供应链战争赢家 IVAN CHIAM, MYRON XIE, RAY WANG, AND 3 OTHERS IVAN CHIAM、MYRON XIE、RAY WANG 以及其他 3 位作者 MAR 12, 2026 2026 年 3 月 12 日 ∙ PAID ∙ 付费内容 19 1 Share 分享 SemiAnalysis x Fluidstack Hackathon SemiAnalysis x Fluidstack 黑客松 We're hosting a hackathon ahead of GTC this year on Sunday the 15th, in partnership with Fluidstack, and would love to have you join us! ...

行政院已核定沙仑生态科学园区开发计画，规划全期土地面积为531公顷，完成后产值为2兆2,547亿 元，提供就业机会达3.5万人。 沙仑生态科学园区因开发范围涉及草鸮栖地，引发环保团体关注。 国科会南科管理局长郑秀绒表示，南科已先收集各界意见，并对外沟通，环保团体特别关切栖地破碎 化问题，园区开发规划已进一步优化。 她说，优化方案不仅单纯计算保留多少绿地或草生地，且更重视生态空间的完整性，例如滞洪池周边 森林等，将尽可能极大化保留，以降低对草鸮栖地影响。 近期南科管理局先透过专家会议及工作小组会议，广泛听取关心草鸮的学者与环保团体意见。相关讨 论成果与资料，目前已公开在南科管理局官网，供外界查阅。 公众号记得加星标⭐️，第一时间看推送不会错过。 市场传台积电1.2纳米制程预定地的沙仑生态科学园区，原订2025年底送交环评，因园区开发范围涉 及保育类猛禽草鸮栖地，国科会南科管理局已先行收集各界意见、并对外沟通，预估今年4月提出环 境影响说明书送交环境部。 年。根据估算，四座工厂首年合计将为台积电带来160 亿美元的营收，同时还将创造8000 至1 万个 就业机会。 目前尚未披露哪些客户打算采用台积电1.4 纳米制 ...

Core Viewpoint - The article highlights Taiwan Semiconductor Manufacturing Company (TSMC) as a key player poised to join the $2 trillion market cap club, benefiting from the increasing demand for artificial intelligence and advanced chip manufacturing [1][2][3]. Group 1: Market Position and Growth - TSMC is the largest contract chip manufacturer globally, accounting for nearly 70% of spending by major companies like Nvidia and Apple, with Samsung trailing at only 7% [5]. - TSMC's market share is expected to grow further due to its technological lead, with its 2nm process entering mass production by the end of 2025 [6]. - The company has raised prices on its advanced chipmaking processes by 3% to 10% starting in 2026, indicating strong pricing power and demand visibility through 2029 [9][10]. Group 2: Financial Performance and Projections - TSMC's stock reached a market cap of $1.8 trillion, up 97% over the past year, with expectations to surpass the $2 trillion mark soon [3]. - Management projects a 30% revenue growth in 2026, with a compound annual growth rate of 25% from 2025 to 2029, suggesting robust financial performance [11][12]. - The company anticipates earnings growth to outpace revenue growth due to strong demand for its 3nm and 2nm processes [13]. Group 3: Strategic Investments - TSMC plans to invest between $52 billion and $56 billion in capital expenditures this year, up from $40.9 billion last year, focusing on new facilities in Arizona to mitigate geopolitical risks [10]. - The company’s conservative outlook historically suggests potential for further upside in its financial projections [12]. - TSMC's ability to maintain high gross margins while ramping up next-generation processes positions it favorably for future growth [13].

分析师指出:"长期中断可能对支持芯片晶圆制造所需的稳定能源供应构成风险。"报告还强调了风险的集中性,因为仅 台积电(NYSE:TSM) 一家就 制造了全球90%的先进芯片,并消耗中国台湾总电力的约9-10%。 分析师认为,眼前的现实可能表现为成本的大幅上涨,而非完全停产。他们表示"这是一个值得关注的风险,关系到全球科技和AI芯片供应。"任何电 力波动都可能对全球AI和智能手机处理器供应产生直接连锁反应。 "硫磺短缺"与二阶中断 除了电力之外,报告还指出了一个不太明显但同样关键的风险:硫酸短缺。硫磺是石油精炼的副产品,目前在海湾地区陷入停滞,对于提取铜和钴等 金属至关重要,这些金属广泛用于芯片组件和电气化领域。 随着霍尔木兹海峡的关闭威胁到先进芯片制造所需的能源和化学原料供应,全球半导体供应链正面临一个新的危险瓶颈。根据摩根士丹利最新发布的"科技 简报"报告,这一中断对该板块构成双重威胁: 中国台湾晶圆代工厂即将面临的" 液化天然气 悬崖",以及可能破坏关键电池和芯片材料生产的"硫磺短缺"。 中国台湾的液化天然气储备与11天倒计时 全球科技行业的首要担忧仍然是中国台湾电网的稳定性,该电网严重依赖液化天然气(LN ...

Core Viewpoint - AI stocks are currently undervalued compared to late 2025, presenting solid buying opportunities as AI spending is expected to continue growing for many years [1] Group 1: Nvidia - Nvidia is the industry leader in AI, maintaining its dominance since 2023 with continuous innovation and premium pricing for its platform [3] - Current market cap is $4.4 trillion, with a stock price of $180.28, down 1.56% today [4] - Nvidia's Q4 2025 growth rate was 73%, with management expecting 77% growth in the next quarter, trading at 22 times forward earnings, indicating it is a strong buy [5] Group 2: Broadcom - Broadcom is positioning itself to challenge Nvidia by designing custom chips optimized for specific workloads, offering a more efficient and cost-effective solution [6] - The company expects its AI division to generate $100 billion in revenue by the end of 2027, which could dominate its overall business, currently making up less than half of its $68 billion total revenue [8] Group 3: Taiwan Semiconductor Manufacturing - Taiwan Semiconductor is a key player in the AI sector, manufacturing most logic chips for high-end devices, benefiting from increased spending on data centers by AI hyperscalers [9][10] Group 4: Microsoft - Microsoft has seen a 25% decline from its all-time high but is now trading at some of its lowest valuations in a decade, presenting a buying opportunity [11][13] Group 5: Alphabet - Alphabet has transitioned from an AI underperformer to a leader, with its stock price reflecting this recovery, currently trading at a premium of 26 times forward earnings [14] - The Google Cloud business has shown significant growth, with a 48% year-over-year revenue increase, indicating strong demand for its computing resources [16]

证券研究报告 | 行业周报 gszqdatemark 2026 03 15 年 月 日 电子 周观点：从 OFC 前瞻看光变革，把握光芯片与 CPO 机会 OFC 2026 大会前瞻：将于 2026 年 3 月 15 日至 3 月 19 日在美国加州举 行，恰逢 AI 数据中心光互联需求加速释放的关键节点。本轮光互连革命 从 800G 到 1.6T 的产品迭代、从可插拔到 CPO 的架构演进、从铜缆到光 学的短距替代，多条技术曲线同步推进，大会涵盖重点包括：1.6T 及 3.2T 光模块技术路径、CPO 与新型光学 I/O 架构、硅光子异构集成、可插拔方 案与 CPO 方案的路线之争，将对未来数年的光通信格局产生深远影响。 大会参与者汇聚了光芯片、光模块、光系统全产业链的核心玩家，包括 Coherent、Lumentum、英伟达、谷歌、Meta、微软、博通、Marvell、台 积电、英特尔、三星等，将定义光互联技术核心演进方向、发布从激光器、 调制器、光引擎到全栈光网络方案的全链条创新、更新硅光子代工平台进 展等。除了产业界，众多高校、科研院所也将参会，发表最新高速光调制 技术成果，根据论文摘要，围绕 EML ...

Industry Overview - Artificial intelligence (AI) is recognized as a highly disruptive technology, significantly enhancing productivity across various sectors [1] - PwC projects that AI adoption could contribute an additional 15 percentage points to global GDP by 2035 [1] - The AI market's revenue is expected to grow from $274 billion in 2023 to $5.3 trillion by 2035 [1] Company Analysis: Taiwan Semiconductor Manufacturing (TSMC) - TSMC is a leading manufacturer of semiconductors essential for AI applications, serving a wide range of industries including data centers, personal computers, and smartphones [4][7] - The global semiconductor market is projected to exceed $2.8 trillion in revenue over the next decade, driven by the demand for AI technologies [6] - TSMC anticipates a compound annual growth rate (CAGR) of mid-to-high 50% for its AI accelerator sales through 2029, with overall revenue expected to grow at a 25% CAGR [8] - Analysts are optimistic about TSMC's future, raising earnings expectations due to the ongoing growth in the semiconductor market [9] - TSMC holds a dominant market share of approximately 72% in the foundry segment, making it a prime investment opportunity [8][11] Company Analysis: Palantir Technologies - Palantir provides a software platform that facilitates the deployment of generative AI in business operations, significantly enhancing decision-making and process automation [12] - Following the launch of its Artificial Intelligence Platform (AIP) in April 2023, Palantir has seen substantial growth in its customer base and deal sizes [12][14] - The company reported a 34% increase in its customer count, reaching 954 by the end of 2025, and a significant rise in high-value contracts [14] - Palantir's remaining deal value surged to $11.2 billion, a threefold increase since the end of 2022, reflecting strong demand for its services [15] - The AI software platforms market is expected to grow at a 29% CAGR through 2034, positioning Palantir favorably to capitalize on this growth [17][18]

Group 1: Microsoft and AI Integration - Microsoft CEO Satya Nadella indicated that up to 30% of Microsoft's code was written by AI last year, with expectations that this percentage has increased [1] - Nadella stated that the amount of code written by AI is steadily increasing, with predictions that by 2030, 95% of the company's code might be AI-generated [2] Group 2: Taiwan Semiconductor Manufacturing - Taiwan produces 60% of the world's semiconductor chips and 90% of advanced semiconductor chips (7 nm or smaller), essential for high-end AI programs [4] - Taiwan Semiconductor Manufacturing (TSMC) is a "pure foundry" company with a dominant 72% market share in the semiconductor space as of Q3 2025, significantly ahead of Samsung at 7% [9] - TSMC is crucial for the semiconductor chip market and the AI data center hardware market, as it does not design chips but produces them on contract [7] Group 3: Nvidia and Market Dynamics - Nvidia controls 92% of the data center GPU market, with its GPUs being manufactured by TSMC, highlighting the interdependence between these companies [5] - Nvidia's top-of-the-line Blackwell chip is produced at a TSMC factory in Arizona, showcasing TSMC's role in supporting leading technology firms [6]

Core Viewpoint - Investing in artificial intelligence (AI) continues to present significant growth opportunities, with four key stocks highlighted as strong buys in March [1] Group 1: Microsoft - Microsoft is heavily investing in AI-related offerings, particularly through its Azure cloud platform, which is experiencing substantial growth [3] - Azure's revenue increased by 39% year-over-year in Q2 of fiscal year 2026, driven by partnerships, notably with OpenAI [3] - Despite a 25% decline from its all-time high, Microsoft's stock presents a buying opportunity for long-term investors [5] Group 2: Nvidia - Nvidia's stock is down approximately 11% from its all-time high, yet its valuation is considered depressed at 21.6 times forward earnings, making it cheaper than the broader market [6][8] - The company is expected to experience significant growth over the next five years due to increased AI spending, positioning it as a strong investment [8] Group 3: Broadcom - Broadcom is developing custom AI chips tailored for specific applications, which can outperform traditional GPUs at lower costs, indicating a competitive edge in the AI computing space [9] - The AI semiconductor division of Broadcom grew by 106% to $8.4 billion in Q1 of FY 2026, with expectations to exceed $100 billion in AI chip revenue by the end of 2027 [10] Group 4: Taiwan Semiconductor Manufacturing - Taiwan Semiconductor Manufacturing Company (TSMC) is the largest chip foundry globally, producing logic chips for various AI hyperscalers, making it a neutral and valuable partner in the industry [11] - TSMC anticipates a compound annual growth rate (CAGR) of nearly 60% for AI-related chips from 2024 to 2029, indicating strong growth potential [13]