Core Viewpoint - Morgan Stanley maintains an "Overweight" rating on TSMC (TSM.US) with a target price of NT$1,275, citing significant increases in CoWoS capacity projections for 2026 and 2027 [1] Group 1: TSMC Capacity Expansion - TSMC's CoWoS capacity is expected to reach 95,000 wafers per month by the end of 2026 and further increase to 112,000 wafers per month by the end of 2027, which is a notable upgrade from previous forecasts [1] - The expansion is primarily driven by the planned capacity construction of the AP8 factory, expected to commence production in mid-2026, alongside sustained demand from key clients such as NVIDIA, Broadcom, and AMD [1] - From the second half of 2026, full-stack CoWoS packaging orders for non-AI products will begin to shift to OSAT partners, mainly ASE, allowing TSMC to focus more on the CoW (Chip-on-Wafer) segment [1] Group 2: Client Demand Insights - Broadcom is projected to have a CoWoS allocation of 185,000 wafers in 2026, representing a 93% year-over-year increase, driven mainly by demand from Google's TPU project [2] - Morgan Stanley estimates that TPU shipments will reach 2.5 million units in 2026, a 38% increase year-over-year, with additional contributions from Meta's first CoWoS-based ASIC (Athena) and OpenAI's ASIC projects [2] - NVIDIA's demand is expected to remain robust in the first half of 2026 due to increased chip sizes and strong B300 demand, with an estimated production of 2.9 million Rubin GPUs [3] Group 3: AMD's Growth Dynamics - AMD's CoWoS demand is expected to remain stable in 2025 but will grow in 2026 with the ramp-up of MI400/MI450 series and Venice CPU, which utilize HBM and CoWoS-S packaging [4] - The adoption rate of MI400/450 (using CoWoS-L) will be a key variable for AMD's growth trajectory, while the company plans to expand 2.5D/wafers-level fan-out packaging applications in gaming GPUs and high-end server/PC CPUs in 2026/27 [4]

Core Viewpoint - The historic alliance between Nvidia and Intel marks a significant shift in the competitive landscape of the semiconductor industry, focusing on AI infrastructure and personal computing products through a $5 billion investment from Nvidia into Intel [2][6][10]. Group 1: Partnership Details - Nvidia and Intel will collaborate on three main areas: integrating Nvidia's NVLink technology for seamless CPU-GPU connectivity, customizing x86 architecture CPUs for Nvidia's AI platform, and launching a new x86 SoC with integrated Nvidia RTX GPU for the PC consumer market [6][10]. - Nvidia's investment of $5 billion at a price of $23.28 per share has led to a nearly 30% surge in Intel's stock price, raising its market capitalization to $129.5 billion with a year-to-date increase of 47.53% [6][8]. Group 2: Strategic Implications - The partnership provides Intel with a crucial entry point into the AI market, which it has been striving to penetrate, while also enhancing its cash flow amidst financial pressures [10][11]. - Nvidia's role as a core player in the AI server market positions Intel's CPUs as potential standard components in Nvidia's system solutions, thereby increasing Intel's relevance in the AI ecosystem [10][11]. Group 3: Competitive Landscape - The collaboration alters the competitive dynamics, putting pressure on rivals such as AMD and ARM, as the combined strengths of Nvidia's GPUs and Intel's CPUs create a formidable alliance [14][15]. - Nvidia's CEO emphasized that the partnership does not hinder its ongoing development of ARM-based products, indicating a continued commitment to multiple architectures while leveraging Intel's x86 ecosystem for broader market penetration [16][17].

Core Viewpoint - The collaboration between Nvidia and Intel marks a significant strategic partnership aimed at expanding their market presence in the AI era, with Nvidia investing $5 billion in Intel to enhance their technological integration and market reach [5][30]. Group 1: Investment and Market Impact - Nvidia announced a $5 billion investment in Intel, equivalent to approximately 356 billion RMB, through stock acquisition at $23.28 per share [5]. - Following the announcement, Intel's stock price surged by 30% in pre-market trading, reaching its highest point in nearly a year [9]. - The partnership aims to tap into a market projected to be worth nearly $50 billion annually, with Nvidia focusing on expanding its data center market share, currently valued at $30 billion [12][15]. Group 2: Technological Integration - Nvidia plans to integrate its NVLink technology into Intel's systems to enhance Intel's competitiveness in the data center market, while Intel will incorporate Nvidia's GPUs into its X86 processors to penetrate the integrated graphics laptop segment, which has a market size exceeding $20 billion [11][12]. - Both companies have been collaborating on solutions and architectural designs for nearly a year, indicating a deep commitment to this partnership [14]. Group 3: Political and Strategic Context - The collaboration is seen as a strategic move for Intel, which has faced significant challenges, including leadership changes and production delays, and is now positioned to leverage Nvidia's support to revitalize its market standing [8][20]. - Nvidia's investment is also viewed as a way to strengthen its ties with the U.S. government, which has been supportive of Intel's initiatives to boost domestic chip manufacturing [22][24]. - The partnership may help Nvidia navigate the complexities of U.S.-China relations, as it seeks to balance its business interests with political pressures [25][26]. Group 4: Future Outlook - Nvidia's CEO expressed confidence that the investment will yield substantial returns in the future, highlighting the financial strength of Nvidia, which reported $56.79 billion in cash and cash equivalents as of the end of Q2 2026 [28]. - The collaboration is characterized as a "giant marriage" where both companies have distinct motivations, with Intel gaining a foothold in the AI market and Nvidia securing long-term political benefits [29][30].

文｜邱晓芬 编辑｜苏建勋 9月18日，英伟达官网上宣布了一则重磅消息：英伟达将向英特尔投资50亿美金（折合人民币356亿）。这一项投资将以收购英特尔股票的方式落地，以一 股23.28美元折算。 消息公布之后，北京时间9月19日凌晨一点，英伟达创始人黄仁勋和英特尔首席执行官陈立武召开了一场线上发布会。 发布会上，黄仁勋罕见换掉了他的皮衣，穿上了笔挺的西装。陈立武也用带有中式口音的英语，表示了他对此次合作的感激——"和黄仁勋相识超过30 年…….向黄仁勋致敬，感谢英伟达的支持、信任、信心。" △发布会截图 "瞄准年规模500亿的市场" 除了资金上的合作之外，"两英"未来将互相渗透进各自的疆域—— 在英伟达的疆域里，以英伟达在去年3月发布的NVLink 72系统为例，这一系统除了搭72颗英伟达Blackwell GPU之外，还需要36颗基于Arm架构的、英伟 达专门定制的Vera CPU。而英特尔的CPU由于通过PCIe进行连接，并不支持NVLink，在数据中心市场一直缺乏竞争力。 黄仁勋在发布会上表示，两家公司的技术团队围绕解决方案的讨论和架构设计，已经开展了将近一年的时间。 刚刚上任不到半年的陈立武直言，在他上 ...

Core Viewpoint - The article discusses the recent high interest in NVIDIA's new liquid cooling solution, specifically the microchannel lid, and its implications for the semiconductor industry [2][4]. Group 1: Investment Bank Perspectives - JP Morgan and Morgan Stanley provided detailed analyses of the microchannel lid, highlighting its efficiency in heat dissipation compared to traditional cooling methods [5]. - The microchannel lid integrates a heat spreader and cold plate, allowing for efficient heat transfer and cooling, which is crucial as chip power requirements increase [8][11]. - The adoption of the microchannel lid could increase the number of quick disconnects (QD) in VR series compute trays to at least 12, compared to 8 in the existing GB300 compute trays [12]. - In the short term, the impact on liquid cooling suppliers is limited, as a significant portion of NVIDIA's GPU shipments will still use traditional cold plates [13]. - Currently, ODMs are in the testing phase for the microchannel lid, with a decision expected in one to two months [14]. Group 2: Industry Perspectives - The microchannel lid concept was discussed in the industry as early as late August, with market speculation about its potential use in NVIDIA's Rubin GPU [15]. - Jentech, a key supplier for NVIDIA's lid products, is closely tied to NVIDIA's technology iterations and order fluctuations, which can influence its stock performance [16]. - The maturity of different cooling technologies ranks single-phase cold plates as significantly ahead, followed by dual-phase cold plates and immersion cooling, with microchannel lids lagging behind [18]. - Cold plate suppliers like AVC indicated that the microchannel lid may not be adopted until the release of the Rubin Ultra model, as current production timelines do not support its implementation [18]. - Companies are currently sending samples for the microchannel lid, but sample approval does not guarantee immediate procurement [19]. - Key players in the lid and cold plate sectors, such as Jentech and AVC, are conducting advanced research on microchannel lids, but it remains uncertain which company will dominate the market [21]. - Besides microchannel lids, 3D printing is also emerging as a cutting-edge research direction in the cooling field, offering high precision and customization capabilities [21].

Core Insights - AI-driven advancements are pushing the continuous iteration of liquid cooling technology, particularly with the introduction of the microchannel liquid cooling plate (MLCP) technology, which is expected to become a strategic material for cooling solutions in AI servers [1][2] Group 1: Technology Development - The power consumption of the Rubin GPU is expected to increase from 1.8kW to 2.3kW, exceeding the current cooling plate capacity, prompting NVIDIA to implement MLCP by the second half of 2026 [2] - MLCP technology integrates the metal cover on the chip with the liquid cooling plate, featuring microchannels that allow cooling liquid to flow directly over the chip, enhancing heat transfer efficiency and reducing volume [1] Group 2: Market Implications - The cost of MLCP is projected to be 3 to 5 times higher than existing cooling solutions, with manufacturing costs potentially increasing by 5 to 7 times compared to the current Blackwell cover if GPUs fully transition to MLCP [1][2] - Boyd, a supplier for NVIDIA, has delivered 5 million liquid cooling plates to large-scale data centers, indicating a growing demand for high-performance AI computing cooling solutions [2]

Core Viewpoint - NVIDIA has launched the Rubin CPX, a new GPU designed for large-scale contextual reasoning, which significantly enhances processing capabilities in AI applications [1][2] Group 1: Product Launch and Features - The Rubin CPX GPU is integrated into the NVIDIA MGX rack system, which includes 144 Rubin CPX GPUs, 144 Rubin GPUs, and 36 Vera CPUs, providing 8 EFLOPS of NVFP4 computing power [1][2] - The system features 100TB of high-speed memory and 1.7PB/s memory bandwidth, offering three times the attention mechanism processing capability compared to the NVIDIA GB300 NVL72 system [1][2] Group 2: Market Implications and Predictions - The value of PCBs supporting the CPX is expected to increase, alongside new PCB demand from the CX9 network card, leading to significant growth in the VR NVL144 CPX PCB value [2] - NVIDIA is advancing orthogonal backplane development and may adopt M9+Q materials, which could potentially double the PCB value in a single rack [2] - The technology for ASIC chip PCBs is evolving from high-layer to HDI, with future advancements expected from M8 to M9, continuously enhancing value [2] Group 3: Industry Trends and Demand - Micron has announced a price increase of 20%-30% for storage products, affecting all DDR4, DDR5, LPDDR4, and LPDDR5 products, with a halt in pricing expected for one week [2] - The U.S. Department of Commerce has added 23 Chinese entities, including 13 integrated circuit companies, to its entity list, which may create opportunities for domestic alternatives in analog chips [2] - The surge in downstream inference demand is driving strong growth in ASIC demand, with NVIDIA's technology upgrades boosting both PCB value and volume [2] Group 4: Sector Outlook - The demand for AI PCBs remains robust, with domestic PCB manufacturers likely to benefit from slow overseas expansion [4] - The overall outlook for various sectors is positive, with consumer electronics, PCB, semiconductor chips, and related industries showing stable to accelerating growth [4]

Investment Rating - The report maintains an "Overweight" investment rating for the electronic industry, indicating a positive outlook for the sector in the next 6 months [1]. Core Insights - The introduction of the Rubin CPX, which offers approximately 30 PFLOPS of computing power and is designed for "million-token" inference scenarios, marks a significant advancement in NVIDIA's product line [2][7]. - The launch of Rubin CPX is expected to accelerate global demand for computing power, particularly in AI applications that require long context generation, thus enhancing the value chain across various related sectors [3][7]. - The Rubin CPX system, which integrates with Rubin GPUs and Vera CPUs, aims to optimize resource utilization and reduce inference costs while improving response times [7]. Industry Trends - The report highlights a shift in the computing infrastructure towards a new phase characterized by "context and generation collaboration," driven by the demand for long-context inference and multimodal video generation [3][7]. - Companies within the supply chain, including those involved in PCB, copper cables, optical chips, and server manufacturing, are expected to benefit from the advancements brought by Rubin CPX [3][7].

Group 1 - Nvidia will hold its next GTC conference from March 16 to 19, 2026, in San Jose, USA, which is considered a significant event in the global AI sector [1] - The upcoming conference is expected to attract developers, entrepreneurs, and tech enthusiasts from around the world, indicating its growing importance [3] - Key highlights of the next conference will include the Rubin GPU and Vera CPU, which are rumored to outperform the current Blackwell chip and support large-scale AI training [4] Group 2 - Nvidia is expanding beyond hardware with initiatives like the AI factory operating system Dynamo, the robot model Isaac GR00T, and the physics engine Newton, aiming to create a comprehensive intelligent ecosystem [4] - Nvidia's stock has surged over 50% this year, with a market capitalization exceeding $2 trillion, making it one of the most valuable tech companies globally [4] - The rapid development of AI technology raises concerns about job displacement and the ability of the general public to keep pace, which may be addressed at the upcoming conference [4] Group 3 - Chinese tech companies like Huawei and Cambricon are rapidly developing their own AI chips, indicating a competitive landscape in the AI sector [4] - The competition in technology ultimately benefits humanity as a whole, highlighting the positive impact of innovation [4] - The anticipation for the next GTC conference suggests that it will be a significant event for technological advancements and industry insights [4]

Summary of Conference Call on Liquid Cooling Technology Industry Overview - The liquid cooling technology is emerging as a key driver in the energy sector, following light modules and PCBs, with a consensus that the liquid cooling segment is about to explode, presenting a vast market opportunity [1][2][3] Key Insights and Arguments - The liquid cooling market is expected to reach a scale of hundreds of billions by 2026, with the energy market potentially expanding to thousands of billions by 2027, indicating significant investment opportunities [2][4] - Taiwanese companies currently hold a competitive advantage in the liquid cooling product segment, particularly in the cold plate area, capturing 30%-40% of the value. However, A-share listed companies have opportunities in the remaining 60%-70% of the market, including CPU connectors and piping [1][5] - Domestic companies like Alibaba and ByteDance are actively investing in liquid cooling technology, with Invec being highlighted as a leading player with a comprehensive product range and significant market share in GPU and Ethereum sectors [1][6][7] Market Dynamics - The main market debate centers around the competitive landscape of the energy sector and whether A-share companies can secure a significant market share. As the liquid cooling industry continues to evolve, A-share companies are expected to gain competitive advantages [3][4] - The demand for liquid cooling technology is driven by the increasing power consumption of new generation chips, such as the GB300, which has a power consumption close to 2000 watts, necessitating advanced cooling solutions [9][10] Technological Developments - Current liquid cooling solutions primarily utilize water-based solutions, but future applications may shift towards fluorinated liquids due to their higher boiling points and stability compared to flammable organic silicone oils [10] - The development of the Robin chip and its associated equipment is underway, with production ramping up for related devices, indicating a growing need for advanced cooling technologies [8] Future Trends - The liquid cooling market is projected to grow rapidly, with estimates suggesting it could reach several hundred million RMB this year, close to a thousand billion next year, and potentially several thousand billion in the following years [4][13] - Domestic manufacturers across various segments, including cold plates, CDU piping, quick connectors, and data center operators, are expected to benefit from this growth trend, making them attractive investment targets [13] Additional Insights - The domestic computing power industry is undergoing rapid iteration, with significant capital expenditures reported by major players like Alibaba, which saw a 219% year-on-year increase in Q2 capital expenditures to 38.676 billion RMB [4][11] - The introduction of protocols like OSI2.0 by China Mobile is expected to accelerate the development of domestic super node clusters, further enhancing the liquid cooling market's potential [11][12]