Ac t i o n | Google TPU sales to grow 4x or ~$65B by FY27 — We model FY26/27 AI revenues of $63B/$102B driven by better-than-expected TPU ramps. We believe AVGO's five announced customers are Google, Meta, Bytedance, Anthropic and Fujitsu based on supply chain discussions, while its two additional engagements are with OpenAI and Apple. Gross margin dilution from Anthropic rack shipments — Given Anthropic's estimated ~$10B rack sales in 2H26 and ~$11B in 1H27 (per AVGO), which we assume ~55% gross margins, w ...

Core Insights - The trade agreement between the United States and Taiwan, signed on February 12, 2026, reduces tariffs from 20% to 15%, benefiting Taiwanese high-tech exports, especially in semiconductors, pharmaceuticals, and aircraft parts [1][10] - This agreement levels the playing field for Taiwan against other Asian countries, enhancing the competitive position of Taiwanese firms like Taiwan Semiconductor Manufacturing Company (TSMC) [2][10] - The deal is expected to create a favorable environment for Exchange-Traded Funds (ETFs) that focus on Taiwanese companies, allowing investors to benefit from the anticipated growth without the risks associated with individual stock selection [3] Trade Agreement Details - The agreement includes a commitment from Taiwan to purchase $84 billion in U.S. goods and a $250 billion investment pledge from Taiwanese firms into U.S. manufacturing [6] - Taiwan will remove or reduce 99% of tariff barriers on U.S. goods, providing preferential market access for U.S. exports [6] - This reciprocal trade framework aims to reduce the U.S. trade deficit while boosting the Taiwanese economy, particularly in the context of rising global demand for AI technologies [7][8] Economic Impact - Taiwan's economy grew by 8.6% in 2025, driven by a 78% increase in exports to the U.S., largely due to AI demand [8] - The reduction in tariffs is expected to lead to increased export volumes, allowing Taiwanese companies to capture efficiency gains [8] Investment Opportunities - The iShares MSCI Taiwan ETF (EWT) has net assets of $8.15 billion, with a 37.5% increase over the past year, and significant exposure to TSMC [11] - The Franklin FTSE Taiwan ETF (FLTW) has net assets of $706.3 million and has surged 46.2% over the past year, also heavily invested in TSMC [12] - The Fidelity Emerging Markets Multifactor ETF (FDEM) has net assets of $448.1 million, with TSMC as its top holding, and has gained 30.6% over the past year [13][14]

TSMC (2330.TW) Earnings Call Summary Company Overview - **Company**: TSMC (Taiwan Semiconductor Manufacturing Company) - **Market Cap**: NT$43,566,640 million (approximately US$1,380,482 million) [6][8] Key Industry Insights - **Revenue Growth**: TSMC is expected to achieve revenue growth of 20-25% in 2026, driven by strong demand from AI datacenters and edge AI applications [1][2] - **Long-term Revenue Target**: TSMC's revenue is projected to reach US$300 billion by 2030, aligning with a long-term CAGR of 20% [1][4] - **AI Revenue Growth**: TSMC anticipates a mid-40s% CAGR in AI revenue by 2029, with potential for upward revision during the upcoming earnings call [9][32] Financial Performance - **Recent Sales**: TSMC reported 4Q25 revenue of NT$1,046 billion, a 6% QoQ increase, surpassing guidance of a 1% decline [4] - **Earnings Projections**: - 2023A: Net Profit NT$838,498 million, EPS NT$32.33 - 2024A: Net Profit NT$1,173,268 million, EPS NT$45.24 - 2025E: Net Profit NT$1,670,190 million, EPS NT$64.40 - 2026E: Net Profit NT$2,200,192 million, EPS NT$84.84 - 2027E: Net Profit NT$2,805,252 million, EPS NT$108.17 [5][8] Capital Expenditure (Capex) - **Capex Guidance**: TSMC's capex for 2026 is expected to be around US$50 billion, with a focus on advanced nodes (N3/N2) and advanced packaging [3][11] - **Capacity Expansion**: TSMC plans to increase N3 capacity by 30% by early 2027 and aims for CoWoS capacity of 1.2-1.3 million wafers in 2026 and 1.8-2 million wafers in 2027 [3][11] Market Position and Competitive Landscape - **Customer Demand**: TSMC is expected to see strong demand from major customers like Nvidia and Google, particularly for AI chips and advanced packaging solutions [11][25] - **Pricing Strategy**: Anticipated price hikes of 5-10% due to technology mix and demand dynamics are expected to positively impact TSMC's gross margins [25][26] Valuation and Investment Outlook - **Target Price**: The target price for TSMC has been raised to NT$2,450 from NT$1,800, reflecting a 45.8% expected return [6][4] - **Investment Recommendation**: The recommendation remains a "Buy" based on strong growth prospects and favorable market conditions [1][4] Additional Insights - **Gross Margin Stability**: Despite potential impacts from foreign exchange and advanced node dilution, TSMC is expected to maintain a gross margin of around 60% throughout 2026 [4] - **Analyst Expectations**: Analysts expect TSMC to provide a positive outlook during the upcoming earnings call, with strong demand for its leading process nodes and advanced packaging business [32] This summary encapsulates the key points from TSMC's earnings call and provides a comprehensive overview of the company's financial health, market position, and future outlook.

Group 1: Market Reactions and Economic Indicators - Stock futures are muted as traders await a Supreme Court ruling on Trump's tariffs and December payrolls data, with S&P 500 futures up 0.1% and Nasdaq 100 contracts up 0.2% [1] - Mortgage stocks surged after President Trump announced a directive for the purchase of $200 billion in mortgage bonds, with LoanDepot (LDI) rising 16% and Rocket Companies (RKT) increasing by 5% [1][4] - The December jobs report and other economic indicators are scheduled for release, including October housing starts and University of Michigan sentiment [1][19] Group 2: Corporate News - Rio Tinto is in discussions to acquire Glencore, potentially creating the world's largest mining company with a combined market value exceeding $200 billion [2] - TSMC reported quarterly sales that surpassed estimates, raising expectations for sustained global AI spending in 2026 [2] - General Motors shares fell after announcing an additional $6 billion in charges related to cutbacks in electric vehicle and battery operations [2] Group 3: Stock Movements and Company Performance - Mag 7 stocks showed mixed performance in premarket trading, with Alphabet up 0.8% and Meta down 0.2% [3] - AXT Inc. saw a 14% decline after disappointing fourth-quarter revenue forecasts due to fewer export control permits from China [4] - Revolution Medicines gained 13% following reports of Merck's interest in acquiring the cancer drugmaker [4] Group 4: Economic Outlook and Market Sentiment - The S&P 500's early-year rally has slowed, with a shift away from AI stocks towards a broader range of tech players and sectors [5] - Traders are preparing for significant risk events, including the payrolls data and the Supreme Court ruling, which may impact global equities [6] - The market is anticipating a mixed reaction to the jobs data, with expectations for nonfarm payrolls to be around 70,000, in line with consensus [8][34]

Summary of Key Points from Conference Call Industry Overview - **Industry Focus**: Semiconductors, particularly in North America, with emphasis on AI and memory strength [1][7] - **Market Dynamics**: The semiconductor ecosystem is being tested by AI strength, with significant growth expected in AI-related revenues [1][2] Company-Specific Insights NVIDIA Corp. (NVDA) - **Market Position**: NVIDIA is expected to maintain a dominant market share despite overstated threats from competitors [2] - **Revenue Growth**: Projected revenue for FY27 is $329.829 billion, reflecting a 55% year-over-year change, with a gross margin of 74.6% [10] - **Product Demand**: There is significant customer anxiety regarding the ability to procure NVIDIA products, particularly the Vera Rubin model [2] - **Estimates Revision**: Estimates for NVIDIA's revenues have been raised, with a target price increase from $235 to $250, based on a new EPS estimate of $9.57 [9][66] Broadcom Inc. (AVGO) - **Revenue Projections**: Broadcom's ASIC revenue is forecasted to be $27.210 billion in FY2026 and $59.475 billion in FY2027, with a price target increase from $409 to $443 [13][15] - **Market Dynamics**: The company is experiencing strong demand for its products, particularly in AI and memory sectors, with a notable increase in TPU supply chain checks [13][17] - **Risks**: Potential risks include competition from Google's Tensor processor and architectural challenges faced by Meta in utilizing Broadcom's products [12][11] Memory Market Insights - **Supply Constraints**: There is an ongoing memory shortage, particularly in DDR5 and NAND markets, with high demand from cloud buyers and OEMs [16][18] - **Market Dynamics**: The memory supply has shifted to high-value users, leading to a stark supply tightness that is not artificial [16] - **Future Outlook**: The memory shortage is expected to persist, with potential risks of demand destruction as prices rise [17] General Market Trends - **AI Influence**: The strength of AI is impacting not only specialty back-end capacity but also front-end wafer supply and memory availability [20] - **Server Market**: The general-purpose server market is strong, with AMD gaining market share while Intel struggles to keep up [21] - **China Localization**: The Chinese semiconductor industry is striving for self-sufficiency, with significant investments in legacy technologies, although advanced technology remains constrained [24][25] Investment Recommendations - **NVIDIA**: Overweight rating with a price target of $250, reflecting strong growth potential in AI and data center revenues [59][66] - **Broadcom**: Overweight rating with a price target of $443, supported by strong AI growth and recovery in core semiconductor businesses [75][81] - **Micron Technology**: Top pick with a price target of $338, driven by improving DRAM fundamentals and AI demand [94][98] Conclusion - The semiconductor industry is poised for significant growth driven by AI, with both NVIDIA and Broadcom positioned to capitalize on this trend. However, supply constraints in memory and competition from emerging technologies present risks that investors should monitor closely.

Summary of Earnings Call for US Semiconductors Industry Overview - The semiconductor industry is experiencing varied earnings, with positive contributions from AI-related companies like AMD and stable performance in the PC and handset markets from Intel (INTC) and Qualcomm (QCOM) [1][10] - The analog recovery is anticipated to be delayed until next year, but an upturn is expected in the next 2-3 quarters due to low inventory levels [1][6] Key Insights AI and Capital Expenditure - Citi raised its 2025 global capital expenditure forecast from $397 billion to $415 billion, and for 2026 from $598 billion to $652 billion, reflecting increased capex plans from hyperscalers [3][17] - AI revenue is projected to grow significantly, with expectations of reaching $975 billion by 2030, indicating an 86% compound annual growth rate (CAGR) [3][17] PC and Handset Demand - PC demand has exceeded expectations, with Intel's client CPU business increasing by 8% quarter-over-quarter (QoQ) and AMD's by 10% QoQ in 3Q25 [4][23] - Handset demand remains stable, with Qualcomm reporting a 10% QoQ increase in its QCT Handset segment, driven by strong Android performance [4][28] Analog Market Dynamics - The industrial segment is showing signs of slowing growth, while the automotive sector is improving after a previous slowdown [5][50] - The overall analog business is expected to follow seasonal trends, with a potential upturn anticipated in the next 2-3 quarters due to low inventory and supply growth [6][37] Semiconductor Demand Trends - Demand from automotive, consumer, and communications markets is improving, while PC, handset, and industrial markets are stable [21] - The data center market remains robust, with AMD's data center revenue up 34% QoQ and Intel's up 5% QoQ in 3Q25 [33] Financial Performance Consensus Estimates - Consensus EPS estimates for 2026 declined by 12%, primarily due to a 49% reduction for Intel and a 16% reduction for Microchip Technology (MCHP) [2][11] - Aggregate sales estimates increased by 1% during earnings, with AMD showing a 5% increase in sales estimates, the highest among covered companies [14] Stock Recommendations - MCHP is highlighted as a top pick due to expected upside in estimates, with other recommended stocks including AVGO, ADI, MU, NXPI, and TXN [7][58] Additional Insights - The semiconductor sector is currently trading at a 37% premium to the S&P 500, reflecting confidence in future growth driven by AI [56] - The analog pricing has stabilized, which may lead to improved consensus estimates for analog companies [40][45] This summary encapsulates the key points from the earnings call, focusing on industry trends, company performance, and financial forecasts.

Summary of Global Server Market and AI Server Insights Industry Overview - The report focuses on the global server market, particularly the AI server segment, with projections extending to 2027E. The analysis includes various types of servers such as ASIC AI servers, AI training servers, AI inferencing servers, general servers, and HPC servers [1][14]. Key Insights Market Projections - The global server market is projected to grow significantly, with total revenues expected to reach **US$359 billion**, **US$474 billion**, and **US$563 billion** in 2025E, 2026E, and 2027E respectively, reflecting year-over-year growth rates of **42%**, **32%**, and **19%** [14][30]. - AI server shipments are expected to increase, with **10 million**, **14 million**, and **17 million** AI chips required in 2025E, 2026E, and 2027E, respectively. The contribution of ASIC shipments is projected to rise from **38%** in 2025E to **45%** in 2027E [1][8]. AI Server Segment Growth - High power AI servers (powered by **500 TFlops+ ASIC and GPUs**) are forecasted to see a **21%** and **39%** increase in volume estimates for 2025E and 2026E, respectively. Inferencing AI servers are expected to grow by **3%** and **5%** in the same years [3][11]. - Full rack AI servers are estimated to ship **19,000** units in 2025E, increasing to **67,000** units by 2027E, with a total addressable market (TAM) of **US$54 billion**, **US$157 billion**, and **US$232 billion** for 2025E, 2026E, and 2027E respectively [11][30]. Cloud Service Provider (CSP) Capital Expenditure - Leading US CSPs are projected to increase their capital expenditure (capex) by **67%**, **23%**, and **15%** year-over-year from 2025E to 2027E. In contrast, leading Chinese CSPs are expected to grow their capex by **55%**, **8%**, and **6%** in the same period [8][24][27]. AI Chip Demand - The demand for AI chips is expected to reach **9,990 thousand**, **13,631 thousand**, and **16,551 thousand** units for training and inferencing combined in 2025E, 2026E, and 2027E, respectively. The mix of GPU and ASIC chips is projected to shift, with GPUs making up **62%** in 2025E and decreasing to **55%** by 2027E, while ASICs will increase from **38%** to **45%** [3][20]. Additional Insights Revenue Breakdown - The revenue breakdown for AI servers indicates that training servers will generate **US$189.6 billion**, **US$294.6 billion**, and **US$370.6 billion** in 2025E, 2026E, and 2027E, respectively. High power AI servers are expected to generate **US$135.9 billion**, **US$137.9 billion**, and **US$139 billion** in the same years [30]. - General servers are projected to see revenues of **US$133.7 billion**, **US$140.2 billion**, and **US$146.5 billion** from 2025E to 2027E, indicating a recovery in the general server market [30]. Market Dynamics - The growth in AI applications and increased cloud capex are driving the demand for AI servers. OpenAI's user base has quadrupled to **700 million** weekly active users, indicating a broader adoption of AI technologies [3][11]. Competitive Landscape - The report highlights key players in the AI server market, including ODMs like Wiwynn, Wistron, and Hon Hai, as well as companies involved in silicon photonics and liquid cooling technologies [2]. This comprehensive analysis provides a detailed outlook on the global server market, emphasizing the significant growth potential in the AI server segment driven by technological advancements and increased cloud investments.

V i e w p o i n t | 16 Sep 2025 12:20:59 ET │ 10 pages Nicholas Lai +886-2-8726-9093 nicholas.lai@citi.com See Appendix A-1 for Analyst Certification, Important Disclosures and Research Analyst Affiliations. Taiwan Semiconductors and Electronic Components & Equipment US marketing feedback - Structural AI growth continues, prefer NVIDIA supply chain on smooth product cycle transition CITI'S TAKE We were in the US meeting investors over the past two weeks. We note that US investors appear more positive on AI' ...

Summary of Taiwan Electronics & Semiconductors Conference Call Industry Overview - The conference focused on the Taiwan electronics and semiconductor industry, highlighting the ongoing demand for AI technologies and the performance of key players in the semiconductor supply chain. Key Points and Arguments AI Demand and Semiconductor Growth - AI demand, particularly from the AWS supply chain, is showing strong momentum, benefiting companies like Accton and Wiwynn [1][2] - TSMC is expected to experience promising growth due to strong advanced node and AI momentum [1] - The launch of Apple's iPhone 17 series has resulted in muted year-over-year and month-over-month performance for most of the Apple supply chain, except for lens-makers like Largan and Genius [1] Preference for AI Supply Chain - Continued preference for the AI supply chain is noted, with expectations of increased demand for advanced nodes, chiplet adoption, and power consumption leading to higher average selling prices (ASP) for advanced foundries and OSAT [1] - Increasing demand for 800G networking is anticipated into the next year, along with higher consumption for data center power [1] Nvidia Supply Chain Insights - The market is becoming more bullish on AI ASICs, especially following Broadcom's strong outlook post-earnings call [2] - Nvidia is expected to double its rack shipments by 2026, reaching 50,000 or more, aided by better infrastructure development [2] AWS Supply Chain Transition - A model transition in the AWS supply chain is expected to begin in the coming months, with Marvell's CEO indicating a focus on inventory digestion [3] - Trainium 3 is anticipated to ramp up significantly in late 2Q26 or 2H26, with Alchip likely to benefit [3] Google TPU and TSMC/KYEC - TSMC and KYEC are expected to benefit from leading-edge advanced node wafer demand and final testing upside in the TPU space, although revenue contributions remain low in the near term [4] Mediatek's Progress - Mediatek's TPU development is slower than expected, but the company's engagement with Google remains strong, with potential upside into 2027/2028 [5] Semiconductor Performance Metrics - TSMC's July and August sales are tracking ahead of market expectations, with a high chance of beating its 3Q25 USD-based sales guidance of +8% quarter-over-quarter [7] - UMC is tracking inline, while Vanguard is lagging due to soft demand in DDIC [7] ASIC Design Services - Alchip is currently behind market expectations due to lackluster demand from its US IDM project, but a strong sales rebound is expected starting from 2H26 [8] Server and ODM Sales - Server players reported strong sales in August 2025, with Wiwynn and Accton exceeding forecasts due to robust AWS ASIC server demand [9] - AI server rack shipments are expected to pick up significantly in 2H25, transitioning smoothly from GB200 to GB300 [9] PCB/ABF Substrate Market - Non-AI product demand in the PCB sector is weak, with early order pull-ins affecting sales [10] - AI ASIC names are seeing robust sales momentum, with expectations of continued price hikes in the ABF substrate market due to inflation [10] Revenue Trends - Overall revenue trends for Taiwan's technology supply chains indicate mixed performance, with some companies outperforming while others lag behind [11][13] Additional Important Insights - The profitability of the ABF industry is expected to be well-protected from 4Q25, with ASP upside anticipated in 2026 [10] - The iPhone supply chain is experiencing muted momentum ahead of new model introductions, with specific lens-makers showing resilience [1][9] This summary encapsulates the key insights and trends discussed during the conference call, providing a comprehensive overview of the current state and future outlook of the Taiwan electronics and semiconductor industry.

Investment Rating - The report maintains a "Buy" rating for MediaTek and a "Neutral" rating for Alchip, while downgrading GUC to "Sell/H" [5][43][52]. Core Insights - The report emphasizes the increasing importance of system integration and interoperability in the semiconductor industry, particularly for AI ASICs and GPUs. It highlights that successful chip design alone is insufficient; cohesive and scalable systems are crucial for performance [1][9][18]. - The transition of AI chips from N5/4 to N3 nodes is expected to significantly increase average selling prices (ASP) and computing capabilities, despite a slowdown in overall AI chip shipment growth [2][16]. - Nvidia continues to dominate the AI server market, holding over 80% of market value, while Broadcom leads in the ASIC market driven by its Google TPUs [23][36]. Summary by Sections AI ASIC Market Dynamics - Broadcom holds the largest market share in the ASIC market, with stable trends expected for TPU v7 and growth anticipated for TPU v8 in late 2026 [3][36]. - The report estimates that AI ASIC shipments from major cloud service providers (CSPs) will reach approximately 4.5 million units in 2026, which is lower than consensus estimates [2][23]. System Integration and Design - The report stresses the need for advanced system design capabilities, including thermal management and network fabric design, to support the growing complexity of AI workloads [18][22]. - It notes that developing an HPC AI ASIC can take 3-4 years, necessitating partnerships with experienced semiconductor makers to expedite the process [20][22]. Company-Specific Insights - MediaTek is highlighted as a preferred choice due to its design expertise and expected revenue contributions from AI ASICs starting in late 2026 [5][43][45]. - Alchip's contribution is expected to begin in 2H26, with a cautious outlook due to potential redesign delays [52][53]. Financial Projections - The report provides revised financial estimates for MediaTek, projecting a revenue of NT$588.85 billion for 2025, with a YoY growth of 11% [46]. - Alchip's financial outlook is less optimistic, with a projected revenue of NT$41.96 billion for 2025, reflecting a decline [52][53].