光模块：本周光模块指数+4.84%，本月以来光模块指数+4.14%。NV 业绩&指引双 beat。但从股价看市场反应温和， 我们认为主要系担心竞争格局、预期交易充分的原因。展望 3 月，NVIDIA GTC 大会即将开幕,我们预计有望展示 CPO 交换机、rubin ultra 等方案，或将给再催化光通信板块。 IDC：本周 IDC 指数+2.41%，本月以来，IDC 指数+2.44%。OpenRouter 数据显示，2026 年 2 月 9 日至 15 日当周，中 国模型的调用量为 4.12 万亿 Token，首次超过美国模型（2.94 万亿 Token）。中国领先的芯片制造商正努力在两年 内将采用尖端工艺技术制造的芯片产量提高五倍，以满足国内人工智能行业的需求。 核心数据更新： 电信业务量收增速逐步提升。2025 年电信业务收入累计完成 1.75 万亿元，同比增长 0.7%。按照上年不变价计算的电 信业务总量同比增长 9.1%。12 月我国光模块出口金额当月同比增加 0.9%；1-12 月累计同比降低 16%。4Q25 微软/谷 歌/Meta/亚马逊资本支出分别为 299 亿/279 亿/214 亿/39 ...

据Theinformation报道，Meta公司在多款芯片的研发上都遇到了问题。该公司放弃了一款内部代号为Iris的第二代训练芯 片。之后，该公司开始研发一款更先进的训练芯片，代号为Olympus，但现在这款芯片也已被放弃。 一位知情人士透露，Meta 最初计划与 Olympus 合作构建大型服务器集群，但高管最终认为，在与 OpenAI 和 Google 等老 牌竞争对手展开激烈竞争之际，这样做会给新模型的训练带来重大风险。例如，用于训练芯片的软件稳定性不如英伟达的 产品，而且 Olympus 复杂的设计也可能导致难以大规模生产。 与此同时，据报道，Meta Platforms 已经与谷歌签订一项价值数十亿美元的协议，租用谷歌的 AI 芯片（即张量处理单元） 来开发新的 AI 模型。这必将加剧谷歌和英伟达的芯片竞争。 Meta的定制芯片之旅 Meta 进军定制芯片领域是一项深思熟虑的战略，旨在克服现成 AI 加速器在技术和财务方面的局限性。首先从财务上看， Meta自研芯片应该是大有可为。 鉴于 Meta Platforms 在研发和资本支出方面投入了巨额资金——预计 2025 年，其研发投入约为 500 ...

Company Project Advancement - Google expects capital expenditures in 2026 to range between $175 billion and $185 billion, nearly doubling from $91.45 billion in 2025, primarily for investments in AI computing capabilities, cloud infrastructure, and strategic projects [2] Product Development Progress - Google plans to launch its first AI glasses in 2026, including versions with audio interaction and advanced models with lens display features, based on the Android XR platform and Gemini large model, marking its return to the smart glasses market since the failure of Google Glass in 2012 [3] Business and Technology Development - Google is negotiating a multi-billion dollar TPU supply deal with Meta, which plans to deploy Google's TPU in its own data centers starting in 2027, potentially altering the competitive landscape of the AI chip market [4] Company Business Status - Google's cloud business had an unfulfilled order amount of $240 billion at the end of Q4 2025, more than doubling year-over-year, supporting short-term growth prospects. The user base of the Gemini model continues to expand, with over 750 million monthly active users as of Q4 2025 [5]

OpenAI对英伟达最新的一些人工智能芯片并不满意，并且自去年以来一直在寻找替代方案，这可能会使这两家AI热潮中最受关注的公司之间的关系变得 更加复杂。 据媒体援引多位知情人士表示，OpenAI对英伟达最新的一些人工智能芯片并不满意，并且自去年以来一直在寻找替代方案，这可能会使这两家 AI热潮中最受关注的公司之间的关系变得更加复杂。 OpenAI这一战略转变，源于其对用于执行AI推理中特定环节芯片的重视程度不断提高。所谓推理，是指像支撑ChatGPT应用的AI模型在响应用 户问题和请求时所进行的计算过程。英伟达在训练大型AI模型所需的芯片领域仍占据主导地位，而推理正成为竞争的新战场。 分析称，OpenAI及其他公司决定在推理芯片市场寻找替代方案，标志着对英伟达AI主导地位的一次重大考验。 周一，英伟达收跌近2.9%。 当前，OpenAI和英伟达这两家公司仍在进行投资谈判： 去年9月，英伟达表示，计划向OpenAI投入高达1000亿美元，作为一项交易的一部分。该交易将使英伟达获得这家初创公司的股份，同时为OpenAI提 供购买先进芯片所需的资金。 在此期间，OpenAI已与AMD等公司达成协议，采购可与英伟达竞 ...

Core Viewpoint - US stock market saw a significant rise with the Dow Jones increasing by over 500 points, as investors shifted focus from recent declines in silver and Bitcoin to upcoming earnings reports from major tech companies and the January non-farm payroll data [1][6]. Group 1: Market Performance - The Dow Jones rose by 515.19 points, or 1.05%, closing at 49,407.66 points; the Nasdaq increased by 130.29 points, or 0.56%, to 23,592.11 points; and the S&P 500 gained 37.46 points, or 0.54%, ending at 6,976.49 points [3][8]. - Oracle announced plans to raise up to $50 billion to expand capacity for cloud customers, leading to a temporary stock increase of over 3%, but it ultimately closed down by 2.7% [3][8]. - Bitcoin fell below $80,000 for the first time since April 2025, indicating a risk reduction by investors following significant declines in gold and silver [3][8]. Group 2: Earnings Reports - Over 100 companies in the S&P 500 are set to release earnings this week, including Amazon and Alphabet, both of which saw stock price increases on Monday [6][10]. - Disney reported earnings that exceeded analyst expectations but warned of a decline in international visitors to domestic theme parks, resulting in a 7% drop in its stock price [10]. - Approximately 78% of the companies that have reported earnings so far have exceeded expectations, with overall earnings growth expected to reach the strongest level in four years [10]. Group 3: Nvidia and OpenAI - Nvidia's plan to invest $100 billion in OpenAI has stalled, with management expressing doubts about the deal and the lack of progress in negotiations [4][9]. - Concerns have arisen within Nvidia regarding OpenAI's business model and long-term competitive landscape, particularly as major companies like Microsoft and Google accelerate their own AI chip development [4][9]. - OpenAI is seeking new funding, with a valuation potentially reaching hundreds of billions, and Nvidia is reconsidering its investment strategy, possibly shifting to a smaller equity investment rather than the original infrastructure commitment [5][9].

北京时间2月2日晚，美股周一早盘转涨，道指上涨160点。市场关注贵金属与比特币的最新走势。超过 100家标普500指数成份公司将在本周公布财报。 道指涨164.96点，涨幅为0.34%，报49057.43点；纳指涨68.81点，涨幅为0.29%，报23530.62点；标普 500指数涨17.25点，涨幅为0.25%，报6956.28点。 比特币自2025年4月以来首次跌破80000美元，这表明投资者在周五黄金和白银大幅下跌后正在进一步降 低风险。 白银价格在过去12个月上涨了一倍多，其在上周五暴跌约30%。这标志着该金属创下了自1980年以来最 差的单日表现。黄金也重挫约9%。 目前比特币与贵金属价格均从各自的低点有所回升，这有助于缓解避险情绪。比特币最后交易价高于 77000美元，而现货黄金和现货白银均下跌超过1%。 随着人工智能（AI）相关问题浮现，华尔街也将注意力转向了英伟达。 据媒体援引知情人士消息报道，英伟达计划向OpenAI投入1000亿美元的计划已陷入停滞。随着英伟达 管理层对该交易表示怀疑，双方原定"数周内完成"的谈判至今无实质推进，合作已实质性搁置。 报道称，英伟达内部对OpenAI的商业模 ...

以下文章来源于锦缎 ，作者海星 来源丨锦缎 作者 | 海星 图源丨摄图网 锦缎 . 上市公司研究平台，专注价值发现、创造与传播 ASIC 趋势比预期还要猛 一年后的今天， ASIC 的发展速度远超预期。尤其近半年以来， ASIC 甚至逐渐成为 AI 竞争的胜负手：国内外大厂开年以来股 价表现最好的分别是百度、谷歌和阿里。 谷歌 TPU+ 自研模型 + 云 + 内部应用的王炸，已经让其立于不败之地；国内互联网大厂，近期被重估的只有自研 ASIC 芯片拆 分独立 IPO 的百度（计划拆分昆仑芯 IPO ）和阿里（计划拆分平头哥 IPO ）。 一年前，我们在《 DeepSeek 掀起算力革命，英伟达挑战加剧， ASIC 芯片悄然崛起 》一文中，更多的是看好 ASIC 带来类似 博通和晶圆代工的产业机会。 如今， ASIC 早已不只是单纯的产业趋势，更成为 AI 巨头必炼的内功。 1.AI 投资规划越大， ASIC 优势就越明显 ASIC （ Application Specific Integrated Circuit ，专用集成电路）与 GPU 的核心差异，在于 ASIC 是针对特定算法和应用 的 " 定制化优化 ...

Core Viewpoint - Microsoft has launched its second-generation AI chip, Maia 200, aimed at providing a cost-effective alternative to NVIDIA's AI GPU series for cloud AI training and inference tasks [1][3]. Group 1: Product Launch and Specifications - The Maia 200 chip, manufactured by TSMC, is designed for high-performance AI inference tasks and is being deployed in Microsoft's AI data centers [1][3]. - The chip features over 1.4 trillion transistors and is built on a 3nm process, offering more than 10 petaFLOPS of performance at FP4 precision and over 5 petaFLOPS at FP8 precision, all within a power consumption of 750 watts [5][6]. - Maia 200's performance per dollar is reported to be 30% better than Microsoft's current hardware, and it outperforms Amazon's Trainium by three times in FP4 performance [5][8]. Group 2: Competitive Landscape - The launch of Maia 200 positions Microsoft as a strong competitor against Amazon's Trainium and Google's TPU, with claims of superior performance in AI inference tasks [3][4]. - Major chip design companies like Marvell and Broadcom are increasingly focusing on developing custom AI ASIC solutions for cloud giants, indicating a competitive shift in the industry [2]. Group 3: Strategic Importance - The development of Maia 200 reflects Microsoft's serious commitment to in-house chip engineering, driven by the growing energy demands of large AI data centers and the need for cost-effective solutions [9]. - The AI ASIC technology route is becoming crucial for major tech companies, as they aim to enhance the cost-effectiveness and energy efficiency of their AI computing systems [10][11].

Group 1 - The communication ETF and software ETF showed recovery, with the communication ETF rising by 2.75% and the software ETF increasing by 1.52% after a significant decline [1] - NVIDIA's CEO Jensen Huang plans to visit China in January, focusing on key suppliers, highlighting the ongoing growth in the AI hardware sector, which has been thriving for over three years [3] - The market for optical modules is expected to double by 2026, with projections for 800G optical module shipments reaching 5-7 million units and 1.6T optical module shipments at 2.5-3 million units [3] Group 2 - Despite the booming hardware sector, the commercialization of AI software remains underdeveloped, with few companies achieving profitability in this area [3] - Alibaba announced a major update to its Qianwen app, transforming it into a digital assistant capable of executing tasks and connecting to real services, indicating a trend towards enhanced software capabilities [4] - The upcoming earnings disclosure period for overseas CSPs may lead to unexpected positive performance in the software sector, suggesting continued investment opportunities in AI [4]

Market Overview - The market experienced a rebound today, with all three major indices closing in the green. The Shanghai and Shenzhen stock exchanges recorded a total trading volume of 2.69 trillion yuan, an increase of 91 billion yuan compared to the previous trading day. The Shanghai Composite Index rose by 0.14%, the Shenzhen Component Index by 0.5%, and the ChiNext Index by 1.01% [1][12][13] - Despite a recent decrease in trading volume, market liquidity remains relatively high, indicating a robust market environment. However, there is significant upward pressure on the market, with accelerated sector rotation and a potential for structural themes in the short term [1][12][13] Military Industry - The military industry ETF saw a recovery today, opening higher and closing with a gain of 3.18% [1][13] - The domestic large aircraft C919 has commenced its first flight on the Guangzhou-Nanjing route, marking a significant step in expanding its operational network [3][15] - The C919 is expected to deliver 15 units in 2025, an increase of 2 units from 2024, indicating a substantial growth trajectory compared to 2023. The main bottleneck remains in production capacity [5][17][19] Commercial Aerospace - The commercial aerospace sector is gaining attention, with significant satellite deployment plans in place. The "Thousand Sails" constellation aims to deploy 1,300 satellites by 2030, while the "Galaxy" constellation plans to launch 5,000 satellites by the same year [6][18] - By the end of 2030, over 21,000 satellites are expected to be launched in China, with a demand for an additional 15,000 satellites projected between 2030 and 2035 [6][18] Artificial Intelligence and Software - The AI sector continues to thrive, with hardware components experiencing unprecedented growth. The market for optical modules is expected to double by 2026, with anticipated shipments of 5,000 to 7,000 units for 800G modules and 2,500 to 3,000 units for 1.6T modules [9][21] - Despite the booming hardware market, the software side of AI is still developing its commercial viability, with few companies achieving profitability. The integration of AI hardware and software is crucial for future growth [9][21][22] - Alibaba has announced a major update to its Qianwen app, transforming it into a digital assistant capable of executing tasks and connecting to real services, which may catalyze further developments in the software sector [10][22]