Summary of Semiconductor Advanced Packaging Industry Conference Call Industry Overview - The advanced packaging industry has become a key path to surpass Moore's Law, addressing physical bottlenecks such as high leakage power, exponential cost increases, and signal transmission losses in processes below 7nm [1][2][3] Core Technologies and Their Applications - **CoWoS-S**: Utilizes silicon interposer and TSV for high-performance interconnection, primarily used in flagship AI chips like NVIDIA H100/A100 and AMD MI300, but at a high cost [1][6] - **CoWoS-L**: Balances performance and cost through local interconnects, currently accounting for about 60% of TSMC's 2.5D packaging for Intel, and is the direction for future large AI chips and domestic companies like Huawei and Cambricon [1][6] - **CoPoS**: Replaces circular silicon interposer with rectangular panels, potentially increasing material utilization from 70%-75% to 100%. TSMC plans to trial production in 2026 and mass production in 2027, while domestic firms are in the research and prototyping phase [1][7] - **CoWoP**: Aims to eliminate the expensive substrate step by directly mounting chips onto PCBs, but is still in conceptual research due to engineering constraints [1][7] Industry Dynamics and Constraints - The necessity for advanced packaging arises from three main constraints: 1. **Physical Limits**: Quantum tunneling effects lead to significant leakage power increases as processes advance below 7nm and 5nm, making further miniaturization less cost-effective [2][3] 2. **Cost Constraints**: Increased complexity in processes raises overall costs exponentially due to more equipment, materials, and mask layers [2][3] 3. **Performance Bottlenecks**: Longer data and current transmission paths within chips lead to higher losses, hindering effective computational power release [2][3] Global and Domestic Players - Major global players in advanced packaging include TSMC, Intel, and Samsung, with OSAT firms like ASE also advancing their capabilities. Domestic firms like Changdian Technology are also positioning themselves in this space [4] Differences Between 2.5D and 3D Packaging - **2.5D Packaging**: Focuses on horizontal integration with multiple chips placed side by side on a silicon interposer, exemplified by CoWoS [5] - **3D Packaging**: Involves vertical stacking of chips, allowing for higher interconnection density and bandwidth, typically seen in HBM stacks [5] CoWoS Variants and Their Characteristics - **CoWoS-S**: High performance but high cost, used in flagship AI chips [6] - **CoWoS-R**: Uses organic RDL for flexibility and lower costs, suitable for cost-sensitive applications [6] - **CoWoS-L**: Aims for a balance between performance, cost, and size, suitable for future large AI chips [6] Future Trends and Directions - The penetration of CoWoS-L is expected to increase as domestic AI chip manufacturers like Huawei and Cambricon transition from CoWoS-S to CoWoS-L as their production volumes rise [6]

Summary of Key Points from Conference Call Records Industry and Company Involved - The discussion primarily revolves around the domestic computing power market in China, with a focus on ByteDance and its procurement strategies for computing power cards, particularly from domestic manufacturers like Huawei and Cambrian. Additionally, the impact of NVIDIA's acquisition of Groq and Meta's acquisition strategies are also discussed. Core Insights and Arguments - **ByteDance's Computing Power Demand**: ByteDance anticipates a significant increase in daily token consumption from 50 trillion at the end of 2025 to 400 trillion by 2026, indicating a surge in demand for computing power. However, the actual consumption of computing cards is expected to grow by 3 to 4 times due to parameter optimization effects [1][4]. - **Shift to Domestic Computing Cards**: With NVIDIA discontinuing older models, ByteDance plans to increase its procurement of domestic computing cards, particularly from Huawei's Ascend and Cambrian's new models. The expected total procurement amount for domestic computing cards is projected to be at least 600 to 700 billion yuan [1][4]. - **Market Trends for Domestic Computing Power**: The development trend for domestic computing power is confirmed to be positive and long-term. Major companies, including ByteDance, are already discussing procurement plans for 2026, with Cambrian showing excellent performance in model adaptation [1][5]. - **Growth Expectations for 2026**: The domestic computing power market is expected to grow by at least 100% in 2026, driven by ByteDance's strong demand for computing power. It is recommended to focus on domestic computing power as a foundational investment [1][6]. - **NVIDIA's Acquisition of Groq**: NVIDIA's acquisition of Groq enhances its technical capabilities, particularly with Groq's LPU architecture, which improves efficiency by computing directly on-chip without data transfer to storage. This acquisition helps NVIDIA address its shortcomings in TPU architecture and mitigates competition from Google [1][7]. - **Meta's Acquisition Strategy**: Meta's acquisition of Minus has not effectively enhanced its large model capabilities. Despite significant investments, Meta has fallen behind competitors like Alibaba and Deepseek, leading to a perception of strategic confusion and a decline in core competitiveness [1][2][8]. Other Important but Potentially Overlooked Content - **Performance of Domestic Manufacturers**: While Huawei's Ascend has shown average performance within ByteDance, Cambrian and other domestic manufacturers are expected to gain market share due to their superior adaptability and performance [1][4][5]. - **Investment Opportunities**: There is a recommendation to focus on quantifiable stocks related to domestic computing power, as the market outlook remains optimistic despite potential short-term fluctuations [1][5].

Core Insights - The global AI startup funding reached $110 billion in 2024, a 77% increase year-on-year, with projections to exceed $200 billion by 2025, nearly doubling the previous amount [1][2] - The AI sector in global stock markets has shown remarkable performance, with Nvidia's stock price increasing 11 times over three years, reaching a market capitalization of over $4 trillion, the highest globally [3] - AI is transitioning from a conceptual phase to a performance explosion phase, with its development speed surpassing expectations, comparable to the mobile internet era [6] AI Development and Investment - The AI revolution is crucial for nations, companies, and individuals, with significant investments from major companies like Facebook, Microsoft, Google, and Amazon, expected to reach $650 billion, $800 billion, $850 billion, and $1 trillion respectively by 2025 [10] - European and Japanese investments in AI are also substantial, with Europe planning to invest €20 billion and Japan projecting ¥196.9 billion, a 67% increase year-on-year [10] - Chinese companies are increasing their AI investments, with Alibaba investing ¥100 billion in the past year and planning to invest ¥380 billion over the next three years [10] AI Market Dynamics - The global AI landscape shows the U.S. leading, with China catching up, while Europe and Japan lag behind [12] - Nvidia dominates the high-end GPU chip market with a 90% share and a gross margin of 75% [12] - Major players in the AI model space include OpenAI's ChatGPT and Google's Gemini, which are leading globally [14] Application and Performance - AI applications are beginning to generate tangible revenue, with ChatGPT's weekly active users surpassing 800 million and projected revenue of $15 billion in 2025, a threefold increase from 2024 [10] - Companies like Microsoft and Tencent are integrating AI into their operations, with Microsoft reporting a revenue of $76.44 billion and a net profit of $27.2 billion in Q2 2025, reflecting an 18% and 24% year-on-year growth respectively [23] - The AI sector is expected to experience explosive growth as it penetrates various industries, with applications in autonomous driving, AI search, and AI design emerging [10] Investment Strategies - The AI sector is still in its early stages, with a development phase of 20%-30%, indicating significant growth potential [27] - A long-term investment strategy is recommended for AI leaders in the U.S. and Hong Kong, while short-term strategies may be more suitable for A-shares [30] - The market is anticipated to enter a new cycle of growth, potentially leading to a 5-10 year bull market as AI applications become more widespread [29]