2、50 倍全球产能有多难？ Terafab 计划落地奥斯汀，号称要在一个建筑里整合光刻、芯片生产、封装测试全流程，实现 AI 计 算全环节自主制造，还直奔 2 纳米先进制程，主打极致闭环、快速迭代。但这份野心，在产业数据 面前显得无比沉重。 周末马斯克宣布特斯拉、SpaceX、xAI 联手 Terafab 超级芯片工厂计划，目标年产能 1 太瓦算力芯 片，直接干到当前全球总产能的 50 倍，80% 产能送上太空，20% 供给特斯拉机器人与汽车。 这篇文章聊几个关键的问题。 1、马斯克为何非要自己造芯片？ 首先，马斯克为何非要自己造芯片？--还是算力太缺了。 当前全球芯片年产能约 20 吉瓦，仅能满足他未来需求的 2%，光是 Optimus 人形机器人，远期年产 10 亿台的目标就需要 100-200 GW算力芯片，更别说 xAI 的超大规模 AI 训练、SpaceX 的太空数据中 心，每一个都是算力吞金兽。 在马斯克看来，地球的算力天花板早已触顶：美国全国电网总容量才 0.5 太瓦，根本撑不起海量 AI、机器人同时运行；而且地面数据中心选址难、散热成本高，反观太空，太阳能效率是地面 5 倍，真空环境零散热成 ...

Core Insights - The main focus of GTC 2026 is not just on individual chip specifications but on whether NVIDIA can successfully mass-produce the Rubin platform, advance the Feynman architecture, and integrate optical interconnects, power supply, and liquid cooling to transition the AI industry from "buying GPUs" to "deploying AI factories" [1][4] Group 1: Event Overview - NVIDIA's GTC will take place from March 16 to 19 in San Jose, California, covering various fields including agent-based AI, AI factories, scientific AI, CUDA, high-performance inference, open models, physical AI, and quantum computing [2] - The Rubin platform is evolving from a single GPU product to an integrated AI supercomputing platform that includes CPU, GPU, interconnects, networking, and system components, enhancing the delivery unit from boards to complete cabinet systems [2] Group 2: Rubin Platform and Feynman Architecture - The Rubin platform is expected to enter mass production, with the potential unveiling of an enhanced version called Rubin Ultra, which will integrate 144 GPUs and achieve a network scale-up of up to 1.5PB/s, with a bidirectional interconnect bandwidth of 10.8TB/s per chip [2] - The Feynman architecture is anticipated to be one of the first chips utilizing TSMC's A16 process, integrating Groq's LPU hardware stack, with production expected to start in 2028 and customer shipments between 2029 and 2030 [3] Group 3: Infrastructure Transformation - The shift towards optical interconnects, high-voltage direct current (HVDC) power supply, and liquid cooling is driving the reconstruction of data center infrastructure, moving from traditional copper interconnects to higher bandwidth and lower loss optical connections [4] - The future of AI systems' scalability will depend not only on chip manufacturing capabilities but also on the efficient and stable delivery of power to each computing node, with liquid cooling becoming a standard configuration for high-power computing platforms [4]

Core Insights - M31 Technology has successfully completed silicon validation of its MIPI M-PHY v5.0 IP on a 4nm process and is advancing research on the 3nm process, indicating its capability to support UFS 4.1 technology [1][2] Group 1: Product Development - The MIPI M-PHY v5.0 IP achieves a maximum single-channel transmission rate of 23.32 Gbps in HS-G5 mode, doubling the performance of the previous HS-G4 generation [2] - The IP incorporates Adaptive EQ technology and supports multi-level signal transmission, maintaining excellent signal integrity and low bit error rate under high-speed conditions [2] - An optimized hibernate design in the MIPI M-PHY v5.0 IP effectively reduces system power consumption, balancing high performance with low power usage [2] Group 2: Ecosystem and Solutions - M31 has built a comprehensive UFS solution ecosystem, integrating PHY IP, JEDEC-compliant UFSHCI v4.1 controller IP, and UniPro control layer IP to simplify system integration [3] - The platform includes ISO 26262 functional safety design processes and certifications, meeting stringent safety and reliability requirements for automotive and high-reliability applications [3] - M31 aims to be a trusted technology partner in building next-generation high-speed storage platforms, leveraging its long-term R&D strength in high-speed interface IP [3]

Core Viewpoint - Semiconductor equipment manufacturer Applied Materials (AMAT.US) saw a significant increase of over 12% in early trading, reaching a historic high of $376.31 following strong quarterly earnings results [1] Financial Performance - For the first fiscal quarter ending in January, the company reported a profit increase of 70% year-over-year to $2.03 billion, with adjusted earnings per share (EPS) of $2.38, surpassing market expectations of $2.21 [1] - Revenue for the same period declined by 2% year-over-year to $7.01 billion, yet still exceeded market expectations of $6.87 billion [1] Future Outlook - For the second fiscal quarter, the company anticipates revenue between $7.15 billion and $8.15 billion, with adjusted EPS projected between $2.44 and $2.84, both of which are above market expectations of $7.01 billion and $2.28, respectively [1] - CEO Gary Dickerson highlighted that the acceleration of global industry investments in the AI computing sector is driving the company's performance towards a strong growth trajectory [1]

Core Viewpoint - Applied Materials (AMAT.US) experienced a nearly 13% increase in after-hours trading, reaching $370.83, following the announcement of strong first-quarter earnings driven by accelerated investments in AI computing [1] Financial Performance - For the first fiscal quarter ending in January, Applied Materials reported a profit increase of 70% year-over-year to $2.03 billion [1] - The adjusted earnings per share (EPS) were $2.38, surpassing market expectations of $2.21 [1] - Revenue declined by 2% year-over-year to $7.01 billion, yet it exceeded market expectations of $6.87 billion [1] Future Outlook - For the second fiscal quarter, Applied Materials projects revenue between $7.15 billion and $8.15 billion, with adjusted EPS expected to range from $2.44 to $2.84, both above market expectations of $7.01 billion and $2.28, respectively [1] - CEO Gary Dickerson highlighted that the acceleration of global industry investments in AI computing is driving the company's strong growth trajectory [1]

Core Viewpoint - The company focuses on four core technology areas: sensing, AI computing, connectivity, and security, providing differentiated products and services for smart terminals, IoT, and automotive electronics [1] Product Offerings - Main products include fingerprint recognition, touch control, audio, NFC/eSE/eSIM, optical sensing, health, and Bluetooth chips and solutions [1] - Storage chips are not part of the company's product range [1]

Core Insights - The article discusses the potential for a shift in U.S. Federal Reserve policy under the new chair, with expectations of interest rate cuts before any balance sheet reduction, which may exceed market expectations [2] - There is a notable interest in Chinese assets from global investors, with several foreign institutions expressing optimism about their performance [1] Market Analysis - The Dow Jones Industrial Average has historically surpassed the 50,000 mark, indicating a shift in market sentiment from growth stocks to value stocks, with increased allocations to high-dividend and defensive consumer staples [1] - Citic Securities suggests that February will see high volatility in the market, with a tilt towards risk assets, making equity assets a more favorable choice [1] Sector Research - Guotai Junan highlights the commercial aerospace industry, particularly the potential of SpaceX's plans to launch a million satellites, which could lower operational costs and make space a viable option for AI computing [7] - Huaxia Fund notes that the transformer industry is entering a high-growth cycle due to surging demand from AI data centers and global grid upgrades, with a focus on solid-state transformers as the next generation [8] - HSBC Jintrust emphasizes the attractiveness of the Hong Kong tech sector, despite recent adjustments influenced by external factors rather than fundamentals [9] Macro and Fixed Income - Tianfeng Securities anticipates that short-term deposit rates will remain stable, with potential downward pressure following the end of the quarter [14] - Morgan Stanley believes that changes in the Federal Reserve's leadership will not significantly alter the policy response function, with potential rate cuts later in the year [15] Asset Allocation - Guotai Fund advises a diversified asset allocation strategy, including A-shares, Hong Kong stocks, U.S. stocks, commodities, and bonds, with a focus on both growth and cash flow assets [3] - The article suggests a continued emphasis on "outbound + technology" strategies, particularly in sectors benefiting from cyclical recovery and AI [18]

Core Viewpoint - The article emphasizes the launch of a supercomputing internet platform that opens a billion-level testing resource pool for AI computing, aiming to revolutionize the traditional computing resource application model and enhance the accessibility and efficiency of computing power across various sectors [1][3]. Group 1: Supercomputing Internet Platform - The newly launched core node is the largest single domestic AI computing resource pool in the country, supporting large-scale AI computing scenarios such as trillion-parameter model training and AI for Science [3]. - The platform aims to address issues like uneven distribution of computing power, varying service levels, and the challenges of heterogeneous computing resource scheduling [3][5]. Group 2: User Categories and Benefits - Lightweight users, including novice researchers and startup teams, can access free resources such as 100 computing hours, 500GB of storage, and 10 million tokens, with additional rewards for feedback [3][4]. - Heavyweight users, such as those involved in large-scale model training and advanced material research, can apply for customized computing resources, including up to 100TB of storage and specialized support teams [4][5]. Group 3: Strategic Importance - The initiative aligns with national strategies for building interconnected computing nodes, indicating a significant step towards enhancing the value of computing applications [3][5]. - The comprehensive support for both lightweight and heavyweight users reflects a commitment to fostering innovation and collaboration within the AI computing ecosystem [5].

Financial Data and Key Metrics Changes - In 2025, the company achieved record revenue of $34.6 billion, a 34% year-over-year increase, with net income rising 42% to a record $2.5 billion and free cash flow nearly doubling to $2.1 billion [5][6][23] - For Q4 2025, revenue reached $10.3 billion, up 34% year-over-year, with gross margin at 57%, an increase of 290 basis points year-over-year [5][24] Business Segment Data and Key Metrics Changes - Data center segment revenue increased 39% year-over-year to a record $5.4 billion, driven by strong demand for EPYC processors and MI350 products [7][25] - Client and gaming segment revenue rose 37% year-over-year to $3.9 billion, with client revenue at a record $3.1 billion, up 34% year-over-year [16][26] - Embedded segment revenue grew 3% year-over-year to $950 million, reflecting demand across various end markets [18][26] Market Data and Key Metrics Changes - The company reported strong demand from hyperscalers in the cloud market, with EPYC cloud instances increasing over 50% year-over-year [8][9] - The enterprise market saw a significant shift towards EPYC adoption, with over 3,000 solutions optimized for major workloads [8][9] Company Strategy and Development Direction - The company is focused on capturing growth opportunities in high-performance computing and AI, with plans to grow data center segment revenue by over 60% annually over the next three to five years [15][20] - Strategic partnerships, such as with Tata Consultancy Services, aim to co-develop industry-specific AI solutions [12] Management's Comments on Operating Environment and Future Outlook - Management expressed confidence in the growth trajectory for 2026, driven by increased adoption of EPYC and Instinct products, alongside a return to growth in the embedded segment [22][30] - The company anticipates significant top-line and bottom-line growth, with a target of greater than 35% CAGR over the next three to five years [22] Other Important Information - The company closed the sale of its ZT Systems manufacturing business, which is reported separately in financial statements [27] - The company generated a record $2.3 billion in cash from operations during the quarter [28] Q&A Session Summary Question: Can you discuss customer engagements and demand for MI450 and Helios platforms? - Management confirmed strong customer engagements and progress on MI450 series development, with a ramp expected in the second half of the year [33][34] Question: Can you provide details on March guidance and data center GPU ramp? - Management indicated a sequential decline in client revenue but strong growth in data center, with expectations for GPU revenue to increase [38][40] Question: What are the assumptions for China MI308 sales beyond Q1? - Management forecasted $100 million in Q1 from MI308 sales to China but did not expect additional revenue due to the dynamic situation [45][46] Question: How is the company addressing server CPU capacity and growth trajectory? - Management noted increased supply capacity for server CPUs and strong demand, indicating a positive outlook for growth throughout 2026 [51][52] Question: What is the outlook for gross margins as new products ramp? - Management expressed optimism about gross margin progression, benefiting from favorable product mix and new product ramps [55][56]

香港聯合交易所有限公司與證券及期貨事務監察委員會對本申請版本的內容概不負責，對其準確性或完整 性亦不發表任何意見，並明確表示概不就因本申請版本全部或任何部分內容而產生或因倚賴該等內容而引 致的任何損失承擔任何責任。 ZHE JIANG HAI LIANG CO., LTD. 本申請版本乃根據香港聯合交易所有限公司（「聯交所」）╱證券及期貨事務監察委員會（「證監會」）的要求 而刊發，僅用作提供資訊予香港公眾人士。 本申請版本為草擬本，其內所載資料並不完整，亦可能會作出重大變動。 閣下閱覽本文件，即代表 閣 下知悉、接納並向浙江海亮股份有限公司（「本公司」）、本公司的聯席保薦人、整體協調人、顧問或承銷 團成員表示同意： 浙江海亮股份有限公司 倘於適當時候向香港公眾人士提出要約或邀請，有意投資者務請僅依據呈交香港公司註冊處處長註冊的本 公司招股章程作出投資決定；招股章程的文本將於發售期內向公眾刊發。 （於中華人民共和國註冊成立的股份有限公司） (a) 本文件僅為向香港公眾人士提供有關本公司的資料，概無任何其他目的；投資者不應根據本文件中 的資料作出任何投資決定； (b) 在聯交所網站登載本文件或其補充、修訂或更換附 ...