Summary of Key Points from the Conference Call Industry Overview - **Industry Focus**: Greater China Semiconductors, specifically related to AI supply chain updates and CoWoS (Chip on Wafer on Substrate) expansion in China [1][2] Core Insights and Arguments - **Investment Recommendations**: - **Overweight (OW)**: TSMC (Top Pick), SMIC, Aspeed, MediaTek, Alchip, GUC, KYEC, ASE, FOCI, ASMPT, AllRing [5] - **Memory Stocks**: Winbond (Top Pick), Phison, Nanya Tech, APMemory, GigaDevice, Macronix [5] - **Underweight (UW)**: UMC, Hua Hong, Vanguard, WIN Semi, OmniVision, ASMedia [5] - **Long-term Demand Drivers**: - **Tech Inflation**: Anticipated price elasticity affecting demand for tech products, with rising costs in wafers, OSAT, and memory impacting margins for chip designers into 2026 [5] - **AI Cannibalization**: AI is expected to replace some human jobs, leading to demand weakness and a shift in the semiconductor supply chain prioritizing AI semiconductors over non-AI [5] - **Tech Diffusion**: A resurgence in AI semiconductor demand driven by generative AI, expanding into various sectors like robotics and AI glasses [5] - **China AI Development**: - DeepSeek is expected to stimulate demand for inferencing AI, raising questions about the sufficiency of domestic GPU supply [5] - The potential dilution of domestic GPU supply chains due to Nvidia H200 shipments [5] Valuation Comparisons - **Key Metrics**: - TSMC's P/E ratio is projected to be 32.8 in 2024, decreasing to 19.4 by 2026, with an expected EPS growth of 1.5% in 2024 and 1.8% in 2025 [6] - SMIC's P/E is not measurable (NM) due to negative growth projections, while UMC is expected to see a decline in EPS growth [6] - **Memory Sector**: - GigaDevice is projected to have a P/E of 76.5 in 2025, with significant growth expected in the following years [7] - Winbond's stock is expected to rise by 32% with a projected EPS growth of 6.3% in 2024 [7] Market Trends - **AI Semiconductor Growth**: - AI semiconductors are projected to account for approximately 34% of TSMC's revenue by 2027 [16] - The global semiconductor market size is expected to reach $1 trillion by 2030, with cloud AI semiconductor TAM projected to grow to $235 billion by 2025 [75] - **Cloud Capex**: - Major cloud service providers (CSPs) are expected to increase capital expenditures significantly, with a projected $621 billion in spending in 2026 [82] - Nvidia's CEO estimates global cloud capex could reach $1 trillion by 2028 [84] Additional Insights - **CoWoS Capacity Expansion**: TSMC may expand its CoWoS capacity to 125k wafers per month by 2026 due to strong AI demand [108] - **Monthly Token Processing**: Data indicates that AI inference demand is on the rise, suggesting a growing market for AI semiconductors [89] This summary encapsulates the critical insights and projections from the conference call, focusing on the semiconductor industry, particularly in the context of AI and its implications for investment strategies.

Summary of Key Points from the Conference Call Industry Overview - The conference call primarily discusses the **AI semiconductor supply chain**, focusing on companies like **Oracle**, **Nvidia**, **Broadcom**, and **TSMC** [1][2][3][4]. Core Insights and Arguments 1. **Oracle's Revenue Backlog**: Oracle announced a significant revenue backlog, which positively impacted Nvidia and its supply chain stocks, including TSMC and KYEC. The company expects its fiscal year 2026 capital expenditures (CapEx) to be around **$35 billion**, primarily for revenue-generating equipment [2][11]. 2. **AI ASIC Demand**: Broadcom reported **$10 billion** in revenue from a fourth customer, identified as OpenAI, indicating strong demand for AI ASICs. This revenue is based on server racks, with less than 100% revenue capture noted [2][9]. 3. **TSMC's Moore's Law 2.0**: TSMC's Deputy Co-COO discussed the transition to Moore's Law 2.0, emphasizing the importance of system integration over mere chip scaling. This shift is expected to enhance the production timelines for Nvidia's Rubin GPU chip, projected to enter mass production in **2Q26** [3][20]. 4. **Cloud CapEx Growth**: The global cloud CapEx is expected to reach **$582 billion** in 2026, reflecting a **31% year-over-year growth**, which is significantly higher than the consensus estimate of **16%** [12][13]. 5. **AI Server CapEx**: The implied AI server CapEx is projected to grow by approximately **70% year-over-year** in 2026, driven by increased spending on short-lived assets [13]. Additional Important Insights 1. **Stargate Project**: A joint venture between OpenAI, Softbank, and Oracle aims to develop infrastructure for next-gen AI, with indications of strong engagement in the supply chain, particularly in server racks [9]. 2. **Chip Consumption**: The consumption of Blackwell chips is expected to reduce inventory risks in the AI semiconductor supply chain, with **28,000 GB200 NVL72 racks** translating to around **2 million units** of chip consumption [10]. 3. **Market Dynamics for KYEC**: KYEC's target price was raised from **NT$158 to NT$188** due to expected strong revenue growth from AI CapEx hikes and market share gains in ASIC and chip probing [71][72]. 4. **CPO Technology**: The adoption of Co-Packaged Optics (CPO) technology is anticipated to start production in **2H26**, with discussions on its coexistence with copper cables [24]. Conclusion The conference call highlighted a robust outlook for the AI semiconductor industry, driven by significant investments from major players like Oracle and Nvidia. The transition to Moore's Law 2.0 and the expected growth in cloud CapEx and AI server CapEx indicate a strong demand trajectory for AI-related technologies and infrastructure.