如果日本能建起最先进的半导体工厂，半导体产业就能重获新生。日本人们都内心深处相信这个说法。 凭借巨额补贴、世界上最先进的制造设备以及大规模生产的最小晶体管，这一景象无疑值得成为一个国家级项目，而且似乎也象征着日本半导体产业的复 兴，该产业长期以来一直被认为"已经衰落"。 在这里，笔者想问一个问题：那家半导体工厂真的盈利吗？此外，在日本建设的尖端工厂真的能说是掌握在日本人手中吗？ 2026 年 2 月 5 日，有报道称，由台积电以及索尼、电装、丰田汽车公司等公司出资兴建的日本先进半导体制造株式会社（JASM） 熊本第二工厂（二 期，以下简称"P2"）决定采用 3nm 工艺生产半导体，而不是 7nm 或 4nm 工艺。这一消息似乎被视为日本半导体产业复兴的象征，但令人惊讶的是，很少 有人理解这一决定的真正意义。 让我直入主题。如果我们只看晶圆制造芯片的前端工艺，3nm（N3）工艺拥有巨大的盈利潜力。然而，最终决定这些利润的并非JASM P2，而是JASM的 母公司台积电。本文将基于盈利模型分析这种看似矛盾的结构，揭示日本半导体政策核心潜藏的重大风险。 为什么选择N3而不是N7或N4？ JASM P2最终跃升至N3 ...

Investment Rating - The report maintains a "Buy" rating for several companies within the advanced packaging and domestic semiconductor sectors, indicating a positive outlook for these industries [6][16]. Core Insights - The advanced packaging industry is experiencing increased demand due to ongoing AI investments, with significant growth expected in the hardware supply chain [5][6]. - Domestic packaging companies are poised to benefit from both overseas and local demand, with expectations of price increases in advanced packaging services [6][10]. - The report highlights specific companies to watch, including Longi Technology, Tongfu Microelectronics, and Yinxin Electronics, as well as materials suppliers like Huahai Chengke and Lianrui New Materials [6][10]. Summary by Sections Advanced Packaging Sector - The report emphasizes the strong demand for advanced packaging, particularly from TSMC's CoWoS packaging, which is currently in short supply, leading to increased orders for domestic companies [5][6]. - Taiwanese companies are reportedly raising prices for packaging services by 5-30%, reflecting a bullish market sentiment [5][6]. Domestic Semiconductor Opportunities - The report identifies a favorable cycle for domestic packaging companies driven by AI investments, with a projected increase in advanced process releases in 2026 [6][8]. - Major domestic players are expected to capture a significant share of the growing demand, with ByteDance and Alibaba projected to invest heavily in domestic computing power [7][8]. Supply Chain Dynamics - The report notes that supply chain constraints, particularly in IC substrates due to upstream material shortages, are creating bottlenecks, which could benefit leading companies like Shenzhen Circuit and Pengding Holdings [10][12]. - The overall supply-demand balance in the optical module industry is also highlighted, with recommendations to focus on companies like Zhongji Xuchuang that are proactively managing supply chain challenges [12]. Investment Recommendations - The report suggests a diversified investment approach across various sectors, including consumer hardware, domestic computing power, and overseas CSP/ASIC supply chains, with specific stock picks provided [14][16].

Core Insights - The article discusses the ongoing competition between GPGPU (General-Purpose Graphics Processing Units) and ASIC (Application-Specific Integrated Circuits) in the AI chip market, emphasizing the critical role of TSMC's CoWoS (Chip on Wafer on Substrate) advanced packaging capacity in determining the future landscape of AI computing power [1][10][36] - Huang Renxun predicts that data center revenues will reach $500 billion over the next six quarters, highlighting the significant financial stakes involved in this competition [1] Group 1: Competition Dynamics - The demand for computing power in AI is expanding, with advanced architectures, process technologies, and advanced packaging being the three key paths to progress [3][4] - NVIDIA has established itself as the leader in GPGPU through its CUDA ecosystem, while Google’s TPU represents a successful ASIC approach, showcasing the efficiency of custom architectures for specific algorithms [3][4] - The competition between GPGPU and ASIC is not merely about performance but also involves considerations of total cost of ownership (TCO) and the ability to optimize financial outcomes for large-scale users [27][28] Group 2: CoWoS Capacity and Supply Chain - TSMC's CoWoS capacity is projected to increase significantly, from approximately 12,000 wafers per month in December 2023 to an estimated 120,000 wafers per month by December 2026, equating to a total capacity of about 1.15 million wafers for AI chips [12][13] - The allocation of CoWoS capacity will be influenced by a complex interplay of technology, business, and geopolitical factors, with NVIDIA expected to secure nearly 60% of the capacity due to its early investments and strong demand [13][15] - The distribution of CoWoS wafers among major clients indicates that NVIDIA will receive around 660,000 wafers, while AMD and the ASIC camp will receive significantly less, highlighting the competitive advantage held by NVIDIA [16][20] Group 3: Performance and Revenue Implications - The performance of AI chips is closely tied to the area of the silicon interposer, with larger interposer areas allowing for more transistors and higher performance [25][26] - NVIDIA's GPUs are expected to command higher prices, with projections of $30,000 to $50,000 per unit, while ASICs like Google's TPU are priced significantly lower, impacting revenue dynamics in the AI chip market [26] - The article concludes that NVIDIA, leveraging its CoWoS capacity, is positioned to capture over 70% of the AI acceleration chip market revenue and more than 90% of the profits, reinforcing its dominant market position [26][34] Group 4: Future Outlook - The future of AI computing is likely to be a hybrid model combining both GPGPU and ASIC technologies, with cloud giants using NVIDIA GPUs for cutting-edge model training and self-developed ASICs for cost-sensitive large-scale inference [35] - The ongoing competition is characterized as a "boundary war," where both GPGPU and ASIC ecosystems will coexist, with TSMC as the ultimate beneficiary due to its critical role in providing CoWoS capacity [36]

Core Viewpoint - Morgan Stanley has raised its forecast for Google's TPU chip shipments for 2026 and 2027, expecting shipments to reach 3.7 million and 5 million units respectively, driven by TSMC's expanding CoWoS packaging capacity and strong market demand [1][3]. Group 1: Capacity Forecast Adjustments - Morgan Stanley has increased its CoWoS capacity forecasts for 2026 and 2027 by 8% and 13% respectively, reflecting TSMC's new capacity construction in the second half of 2026 and 2027 [1][3]. - TSMC's CoWoS capacity is expected to reach 115,000 wafers per month by the end of 2026, with external suppliers providing an additional 12,000 to 15,000 wafers per month [1][3]. Group 2: Demand Drivers - The increase in capacity is primarily driven by rising demand from the ASIC supply chain [1][3]. - The main shipments for 2026-2027 will come from TPU v7 (Ironwood) and v8 series (Broadcom's AX version and MediaTek's X version) [3]. Group 3: Company-Specific Insights - NVIDIA's CoWoS allocation for 2026 remains at 700,000 wafers, with slight adjustments in product mix due to HBM4 readiness issues [4]. - AMD's CoWoS forecast remains unchanged at 90,000 and 120,000 wafers for 2026 and 2027, respectively, with potential delays in the MI450 project [4]. - AWS's Trainium project has seen a slight reduction in expected shipments for 2026, now projected at 2.1 million units [5]. Group 4: Outsourcing Trends - The outsourcing ratio for packaging has increased, benefiting equipment suppliers [6][7]. - TSMC will focus on key GPU and AI ASIC projects, leaving smaller projects to packaging houses like ASE and Amkor [7]. - Equipment suppliers are expected to see a year-on-year increase in demand for CoWoS, with new capacity projected to grow by 40,000 to 50,000 wafers per month [7].

Core Viewpoint - TSMC has solidified its position as the core pillar of the global foundry market, capturing over 70% market share and achieving a revenue of $33.063 billion in Q3 2025, reflecting a 9.3% quarter-on-quarter growth [1][2]. Group 1: TSMC's Market Position - TSMC's revenue growth is significantly higher than its competitors, with a market share increase to 71% [1]. - The overall foundry market is growing, but TSMC is capturing the most substantial portion of this growth [1][2]. - Other foundry players, such as Samsung and SMIC, are struggling to close the gap in market share despite their revenue increases [2]. Group 2: Intel's Strategic Shift - Intel is aggressively transforming its strategy, focusing on advanced technology and ecosystem restructuring to regain competitiveness in the foundry market [3][4]. - The 14A process node is central to Intel's strategy, utilizing High-NA EUV technology to enhance power efficiency and chip density [4][5]. - Intel's EMIB technology is emerging as a viable alternative to TSMC's CoWoS packaging, addressing the demand for advanced packaging solutions [6][7]. Group 3: Customer Acquisition and ASIC Business - Intel has made significant strides in securing major clients, including Apple, which is expected to utilize Intel's 18A-P process for its M-series chips [8][9]. - Intel is establishing a dedicated ASIC department to provide customized chip solutions, leveraging its manufacturing capabilities to attract clients [10][11]. Group 4: Samsung's 2nm Strategy - Samsung is betting heavily on its 2nm process technology, aiming to turn its foundry business profitable by 2027 [12][13]. - The yield rate for Samsung's 2nm process has improved from 50% to a target of 70%, which is crucial for attracting major clients [13][21]. - Samsung has secured contracts with Tesla and Qualcomm, indicating a successful shift in its client base [14][15]. Group 5: UMC's Differentiation Strategy - UMC is focusing on mature processes and high-value applications, avoiding the high-risk advanced process competition [22][23]. - UMC has made significant progress in advanced packaging, securing a partnership with Qualcomm for high-performance chips [24]. - UMC is also entering the silicon photonics market through collaboration with IMEC, aiming to capture the next-generation high-speed connectivity applications [25][26]. Group 6: GlobalFoundries' Focus on Specialty Processes - GlobalFoundries is concentrating on mature and specialty processes, with a clear strategy to establish a unique position in specific markets [29][30]. - The acquisition of Advanced Micro Foundry enhances GlobalFoundries' capabilities in silicon photonics, positioning it as a leader in this field [31][32]. - GlobalFoundries is also acquiring MIPS to strengthen its computing capabilities, providing clients with ready-to-use IP modules [33][34]. Group 7: European Expansion and Local Manufacturing - GlobalFoundries plans to invest €1.1 billion to expand its Dresden facility, aiming to meet the growing demand for secure and differentiated technology in Europe [35][36]. - The company is also exploring partnerships for local manufacturing in the U.S. to address the increasing demand in key industries [27][37]. Conclusion - The foundry market is evolving, with TSMC maintaining a dominant position while competitors like Intel, Samsung, UMC, and GlobalFoundries are carving out their niches through strategic adjustments and technological advancements [38].

Group 1 - The core viewpoint of the articles highlights the positive performance of the semiconductor sector, particularly in advanced packaging and wafer manufacturing, driven by increased demand from GPU and ASIC customers for 2026-2027 [1] - The Shanghai Stock Exchange's Sci-Tech Innovation Board Semiconductor Materials and Equipment Index rose by 0.50%, with notable increases in component stocks such as Huahai Qingke (+5.19%) and Xinyi Chang (+4.74%) [1] - Dongwu Securities indicates that capital expenditure in wafer manufacturing is entering a new phase, with a dual expansion year for storage and advanced logic expected in 2026, supporting high levels of wafer foundry prosperity [1] Group 2 - The Sci-Tech Semiconductor ETF (588170) tracks the Semiconductor Materials and Equipment Index, focusing on semiconductor equipment (61%) and materials (23%), benefiting from the domestic substitution trend and AI-driven semiconductor demand [2] - The Semiconductor Materials ETF (562590) also emphasizes the importance of semiconductor equipment (61%) and materials (21%), targeting the upstream semiconductor sector [2]

Core Insights - The article emphasizes the strategic importance of SiC (Silicon Carbide) as a potential interposer material for advanced packaging solutions by Nvidia and TSMC, with a planned introduction by 2027, indicating a shift towards addressing thermal management challenges in AI computing [3][4]. - The competition in AI computing is shifting focus from transistor density to packaging and thermal management capabilities, highlighting the critical role of heat dissipation in chip performance [3][4]. - SiC is positioned as the optimal solution for CoWoS (Chip on Wafer on Substrate) interposers, balancing performance and feasibility, while traditional materials like diamond and glass fall short in practicality [3][4]. - The adoption of SiC in CoWoS could create a significant new market, particularly benefiting the mainland Chinese SiC industry due to aggressive investments in substrate capacity and cost advantages [3][4]. Group 1: SiC as Interposer Material - Nvidia and TSMC are considering SiC for future advanced packaging, with plans to implement it by 2027 [4]. - SiC is expected to address the thermal management issues associated with CoWoS packaging, which is crucial for the performance of AI chips [3][4]. - The transition to SiC interposers is seen as a strategic move to maintain competitive advantages in the semiconductor industry [3][4]. Group 2: Thermal Management Challenges - The article highlights the increasing power requirements of Nvidia's GPUs, necessitating improved cooling solutions to manage heat dissipation effectively [23][25]. - The CoWoS packaging technology is critical for high-performance computing, and any limitations in thermal management could hinder chip performance and reliability [25][37]. - SiC's high thermal conductivity (490 W/m·K) significantly outperforms silicon (130 W/m·K) and glass, making it a superior choice for managing heat in advanced packaging [105][106]. Group 3: Market Implications - The potential shift to SiC interposers could unlock a substantial new market, moving SiC from a niche power electronics market to a broader AI and data center infrastructure market [15][111]. - The mainland Chinese SiC industry is poised to benefit from this transition, leveraging its investments and production capabilities to capture a share of the global semiconductor supply chain [3][4][113]. - The expected growth in CoWoS capacity, projected at a compound annual growth rate of 35%, underscores the increasing demand for advanced packaging solutions [112].

Investment Rating - The report suggests a positive investment outlook for the SiC industry, indicating significant growth potential due to the anticipated adoption of SiC in advanced packaging technologies by major companies like NVIDIA and TSMC [6]. Core Insights - The report emphasizes the critical need for improved thermal management solutions in AI computing chips, particularly in the context of CoWoS packaging, which is currently facing challenges due to rising power demands [3][40]. - SiC is identified as a promising alternative material for CoWoS interposers, offering superior thermal conductivity and structural integrity compared to traditional silicon and glass materials [4][86]. - The potential for the Chinese mainland SiC industry to benefit significantly from the shift towards SiC interposers is highlighted, given its advantages in investment scale, production costs, and downstream support [5][6]. Summary by Sections 1. NVIDIA and TSMC's Consideration of SiC - NVIDIA plans to adopt 12-inch SiC substrates in its next-generation GPU packaging by 2027, indicating a strategic shift towards advanced materials for better performance [2][9]. 2. Need for CoWoS Thermal Management - The report discusses the increasing power requirements of AI chips, with NVIDIA's H100 GPU exceeding 700W, necessitating enhanced cooling solutions to manage heat effectively [15][22]. 3. SiC as a Preferred Interposer Material - SiC's thermal conductivity is 2-3 times that of silicon, making it an ideal candidate for interposers in CoWoS packaging, which is crucial for high-performance computing applications [4][86]. 4. Benefits for China's SiC Industry - If CoWoS adopts SiC interposers, the demand could exceed 230,000 12-inch SiC substrates by 2030, presenting a substantial opportunity for the Chinese SiC supply chain [5][6]. 5. Overview of SiC Substrate and Equipment Companies - Key beneficiaries of the SiC market growth include companies like Jingcheng Machinery, Jing Sheng Co., Tianyue Advanced, and others, which are positioned to capitalize on the anticipated demand for SiC substrates and related equipment [6]. 6. Investment Recommendations - The report recommends investing in companies involved in SiC substrate production and equipment manufacturing, as they are expected to benefit from the industry's transition towards SiC technology [6].

Core Insights - The global advanced packaging market is projected to reach $56.9 billion in 2025, with a year-on-year growth of 9.6%, and is expected to grow to $78.6 billion by 2028, reflecting a compound annual growth rate (CAGR) of 10.05% from 2022 to 2028 [3]. Industry Transformation - The logic of semiconductor packaging has shifted from merely providing protective casings to creating economic value, indicating a significant change in the industry dynamics [1][2]. - Advanced packaging is becoming a strategic focal point in the semiconductor supply chain, driven by the demand for AI and high-performance computing [2]. Market Dynamics - The demand for advanced packaging is largely fueled by AI applications, which have significantly increased the need for computing chips [2]. - The advanced packaging sales are expected to surpass traditional packaging for the first time in 2025, with consumer electronics and automotive electronics accounting for 85% of this market [2]. Innovation Directions - The industry is witnessing a surge in new technologies and materials, such as Chiplet technology, CoWoS packaging, and advanced substrates like silicon carbide and glass substrates [4]. - Key challenges in advanced packaging include efficient thermal management, heterogeneous integration of Chiplets, and the need for finer line widths and larger package sizes [5]. Equipment and Material Trends - The global advanced packaging equipment market is expected to reach $30 billion by 2030, with significant investments in hybrid bonding equipment and TSV etching machines [5]. - Glass substrates are emerging as a superior packaging material due to their better electrical and thermal performance, although they face challenges in production and reliability [5]. Strategic Development - The industry requires a collaborative ecosystem that integrates large, medium, and small enterprises to achieve high-quality development and innovation breakthroughs [6]. - There is a need for China to enhance its international influence by developing local standards and actively participating in global standard-setting to secure competitive advantages [6].

Core Insights - The AI semiconductor industry is expected to experience significant growth in 2026, with a shift in investment logic from upstream to downstream infrastructure [2][10] - The bottleneck in AI development has transitioned from chip manufacturing and packaging to downstream components such as data center space, power supply, and cooling systems [2][5] Upstream Capacity No Longer the Sole Bottleneck - Chip manufacturing and packaging have significantly expanded, alleviating previous supply concerns [4] - TSMC reported stronger-than-expected AI demand and a quick ramp-up in CoWoS capacity, indicating flexibility in the supply chain [4] - Despite ongoing tightness in advanced node wafer front-end capacity, AI semiconductors are prioritized over other applications like cryptocurrency ASICs [4] Bottleneck Shift - The current constraints are now focused on data center space, power availability, and supporting infrastructure, which have longer construction cycles than chip manufacturing [6] - The deployment of large-scale GPU clusters presents challenges in power consumption and heat dissipation, leading to a shift towards liquid cooling and high-voltage direct current (HVDC) solutions [6] Storage and Memory - AI workloads demand high-speed data storage and access, with companies like Meta opting for QLC NAND flash for cost efficiency [8] - The global demand for HBM (High Bandwidth Memory) is projected to surge, with NVIDIA expected to consume 54% of the total HBM by 2026 [8] Racks and Networking - OCP has introduced standardized blueprints for "AI Open Data Centers" and "AI Open Cluster Designs" to facilitate large-scale deployments [9] - Companies like Alibaba are focusing on pluggable optics for their cost-effectiveness and flexibility, while new technologies like CPO/NPO are gaining attention [9] Demand Forecast Indicates Explosive Growth for Downstream Components - Global cloud service capital expenditure is expected to grow by 31% in 2026, reaching $582 billion, significantly exceeding market expectations [11] - AI server capital expenditure could see approximately 70% year-over-year growth if its share in overall capital spending increases [11] AI Chip Demand Breakdown - NVIDIA is projected to dominate the CoWoS capacity consumption with a 59% share, followed by Broadcom, AMD, and AWS [12] - In AI computing wafer consumption, NVIDIA leads with a 55% share, followed by Google, AMD, and AWS [12] Investment Focus Shift - The signals from the OCP conference and industry data indicate a new direction for AI hardware investment, emphasizing the importance of downstream infrastructure [13] - Investors are encouraged to broaden their focus from individual chip companies to the entire data center ecosystem, identifying key players in power, cooling, storage, memory, and networking [13]