▲ 戳蓝 色字关注我们！ "人生如旷野，而非轨道。每个人都有自己的时区，不必追赶他人的脚步。"——木心 你怎么看"98年的 姚顺雨 担任腾讯首席AI科学家"这件事。 但促使我做这篇文章的原因，是我一个朋友给我发的一张聊天记录截图。截图里是一位家长看到姚顺雨的新闻后，给孩子发的一长段微信。那段 话我看了真的挺有感触的。 这位家长说："儿子，你好好看看这文章，妈妈看完一晚没睡。这个姚顺雨也是同一年的吧？人家在腾讯做首席科学家，带几千人开发AI，你在 干 嘛 ？ 你 要 是 有 姚 顺 雨 一 半 的 科 研 功 力 ， 妈 妈 这 辈 子 是 不 是 都 做 梦 笑 醒 ？ 你 现 在 这 年 纪 正 是 定 胜 负 的 时 候 ， 你 不 努 力 ， 到 时 候 连 去 Jane Street、Two Sigma或者Anthropic这些顶级公司的门都敲不开……" 看到这段话，我突然意识到， 95后离开校园以后，依然逃不掉父母的"鸡娃焦虑"？只不过以前比的是中考、高考，现在比的是职级、薪资、 title。 所以今天我不想聊腾讯的AI战略布局，也不想分析姚顺雨的简历有多牛——清华姚班、普林斯顿博士、OpenAI核 ...

今天大盘整体上涨，截止到收盘，回到4.1星。 大中小盘股都上涨，中小盘股上涨更多一些。 价值风格微跌，成长风格微涨，波动不大。 港股今天只上午交易，略微上涨。 周三下午开始，港股等进入圣诞季。 文 | 银行螺丝钉 （转载请注明出处） 部分投资港股的基金，周四周五会暂停交易。 圣诞假期后恢复正常。 1. 最近A500的指数基金，成交量大幅提升。 A系列指数是A股这几年崛起速度非常快的一类指数。 这类指数目前有三个，A50、A100、A500。 三个指数都是龙头策略，挑选行业的龙头股。 A系列指数历史并不长，但龙头策略在A股历史比较悠久。 以前还有一个恒生A股龙头指数，在A股有十多年历史，业绩也不错。 也有科技龙头、消费龙头等行业龙头策略指数。 龙头策略略带有一些成长风格。 成长风格的策略指数，主要是龙头、质量、成长这三类。 但是之前的龙头策略指数规模都比较小，没发展起来。 2. 在A系列指数诞生后，这个情况发生了变化。 A系列指数基金出现一年多，规模增长很快，目前已经有了2000-3000亿以上的规模。 成为策略指数中规模最大的一类。 其次是红利类指数。 不过红利类指数的指数数量很多。A系列指数就只有A50、A ...

Analyst timothy.arcuri@ubs.com +1-415-352 5676 Natalia Winkler, CFA Analyst natalia.winkler@ubs.com +1-415-352 4626 ab 15 December 2025 Powered by YES UBS Evidence Lab Global Research Cloud Accelerator Study Blackwell Broadens Out, Pricing Holds Up Blackwell broadens out Given ongoing investor concerns about durability of AI demand, we revisit GPU cloud pricing and availability in collaboration with UBS Evidence Lab (>Access Dataset). We think this data illustrates the AI demand environment in general and t ...

Core Viewpoint - The healthy development of the AI industry requires abandoning "bubble anxiety" and "scale worship," focusing on core technology from a long-term perspective, and promoting practical commercialization [1] Group 1: Structural Bubble - The debate over the AI bubble in the U.S. is fundamentally about the imbalance between high investment and low returns, which manifests differently across hardware, software, and applications [2] - Nvidia, as a key player in the "computing power arms race," has seen its AI chip business revenue surge by 210% year-on-year in Q3 2025, with a gross margin of 78%, but its stock price and valuation are increasingly showing signs of a bubble [2][3] - Major tech companies like Microsoft, Amazon, and Google are expected to double their capital expenditures to over $470 billion by 2026, with nearly 60% directed towards Nvidia, amplifying the risk of over-investment in the industry [3] Group 2: Real Value - The current NASDAQ expected P/E ratio of 26 is relatively moderate compared to the 80 during the 2000 internet bubble, indicating that not all U.S. AI investments are a bubble [7] - Companies like Nvidia and Google have established strong positions in AI chips and large models, making their investments technically reasonable [7] - The revolutionary potential of AI for scientific research and industrial upgrades is real, as evidenced by initiatives like the "Genesis Plan" launched by the Trump administration [7] Group 3: Rationality and Overheating - In contrast to the U.S., China's AI investment is characterized by "excessive rationality and insufficient heat," with a lower overall bubble risk but some local areas needing caution [8] - Chinese companies are avoiding the U.S. path of "stacking computing power" and are making steady progress in domestic chip replacement [9] - However, there are signs of bubble risks in certain sectors, with some startups blindly following trends without core technology, leading to resource waste [9]

来源：竞合人工智能 摒弃"泡沫焦虑"与"规模崇拜"，以长期主义视角布局核心技术，以务实态度推进商业化落地，才是AI产 业健康发展的必由之路。 出品 | 竞合人工智能 作者 | 竞合 排版 | 洋洋 当博通单日暴跌12%、甲骨文回吐全年涨幅，英伟达遭大基金集体抛售，华尔街对AI泡沫的质疑从"是 否存在"转向"何时破裂"。这场始于ChatGPT的技术狂欢，在经历三年资本狂飙后，正迎来全球市场的 价值重估。美国AI投资究竟是理性布局还是非理性繁荣？中国AI产业又面临着投资不足与局部泡沫的 双重拷问？从软硬件到应用场景的全链条剖析，或许能揭示这场全球科技竞赛的真实图景。 01 结构性泡沫 美国AI的泡沫争议，本质是高投入与低回报的失衡困境，这种失衡在硬件、软件、应用三个维度呈现 出不同表征。硬件层面，"算力军备竞赛"导致资本支出失控，而作为算力核心的英伟达，既是这场竞赛 的最大受益者，也逐渐显露泡沫承压的迹象。一方面，英伟达凭借A100、H100等高端GPU构建了近乎 垄断的技术壁垒，2025年Q3 AI芯片业务营收同比激增210%，毛利率维持在78%的超高水平，全球超 90%的AI训练算力依赖其产品，订单排期已排至 ...

Core Viewpoint - The domestic GPU company Muxi Co., Ltd. is set to go public on the Sci-Tech Innovation Board with an issue price of 104.66 yuan per share, lower than the 114.28 yuan per share pricing of its competitor, Moer Thread, which will debut as the "first domestic GPU stock" on December 5 [2] Group 1: Product Development and Performance - Muxi's main GPU chip, the Xiyun C600 series, is positioned between NVIDIA's A100 and H100 in terms of performance, with risk mass production expected by the end of 2025 and formal production in the first half of 2026 [3] - The Xiyun C700 series, which is currently in the design and development phase, aims to support low-precision calculations and expand its application in various fields beyond the AI training and inference scenarios [4] - Muxi claims that its Xiyun C series products can be used for model pre-training, having successfully supported large models with billions of parameters [5] Group 2: Financial Performance and Projections - Muxi's revenue is projected to grow significantly from 424,000 yuan in 2022 to 743 million yuan in 2024, with a compound annual growth rate of 4074.52%, and 915 million yuan in the first half of 2025, reflecting a year-on-year growth of 404.51% [6] - The company expects to achieve annual revenue between 1.5 billion to 1.98 billion yuan in 2025, representing an increase of 101.86% to 166.46% compared to 2024, while net losses are projected to decrease significantly [7] - Muxi's management anticipates reaching breakeven as early as 2026, driven by stable gross margins and increased sales from new product developments [8]

【文/观察者网 阮佳琪】 急，很急，特别急——这就是英伟达首席执行官黄仁勋近来的真实写照。为了搞定对华芯片出口，最近一段时间 他都在忙活着向各方强调事态"十万火急"。 据《日经亚洲》4日报道，当地时间周三，在华盛顿智库美国战略与国际问题研究中心（CSIS）举办的一场活动 上，黄仁勋第N+1次发出警告称，如果美国企业放任华为等中国竞争对手抢占市场，中国将很快寻求向全球输出 其人工智能（AI）技术。 活动上，黄仁勋表示，美国限制英伟达芯片对华出口的举措，"实质上已拱手让出全球第二大人工智能市场"，这 将为华为等中国本土企业的技术成熟提供发展空间，最终使其具备在全球层面与美国企业抗衡的实力。 "中国市场的地位无可替代，我们不应将整个市场拱手相让，而应主动参与竞争。"他再次强调。 黄仁勋进一步指出，如果将中国市场完全让给其本土企业，还将为中国向其他国家输出先进技术创造便利条件。 路透社称，此次会面正值美国政府考量是否允许英伟达对华出售H200芯片的关键节点，这款AI芯片的性能相较于 英伟达当前旗舰产品落后一代。 早在11月21日，路透社就曾援引消息称，特朗普政府正考虑批准H200芯片的对华销售。此前，英伟达A100、 ...

Core Viewpoint - Nvidia has become a focal point in discussions regarding the value of AI and related stocks due to a series of stock sell-offs and accounting fraud allegations, prompting the company to respond to market skepticism with a detailed memo addressing twelve key concerns raised by investors [1][2]. Group 1: Nvidia's Response to Investor Concerns - Nvidia's investor relations team sent a seven-page memo to Wall Street analysts addressing various doubts, including a response to Michael Burry's criticism regarding stock-based compensation dilution and stock buybacks, clarifying that the total amount spent on share repurchases since 2018 is $91 billion, not the $112.5 billion claimed by Burry [2]. - The memo also refuted allegations comparing Nvidia's situation to historical accounting fraud cases, asserting that the company's core business fundamentals are strong and its financial reporting is transparent, emphasizing that it does not use special purpose entities to hide debt or inflate profits [2]. - Nvidia addressed concerns about the economic value of its hardware, stating that customers set GPU depreciation periods based on actual usage, with older models like the A100 still generating significant profits, contrary to claims that their economic lifespan is only 2 to 3 years [2]. Group 2: Analyst Insights on Financial Metrics - Analyst Ming-Chi Kuo stated that Nvidia's financial results align with industry norms, countering claims of anomalies in accounts receivable turnover days (Days Sales Outstanding, DSO) and inventory levels [3]. - Kuo explained that the increase in DSO from an average of 46 days (2020-2024) to 53 days in Q3 2026 is reasonable due to a significant rise in customer concentration from 23.8% to 65%, reflecting the bargaining power of major customers [4]. - Regarding inventory, Kuo clarified that the reported 32% increase in inventory for Q3 2026 is consistent with industry trends and that the majority of the inventory consists of work-in-progress items, indicating preparations for strong demand for the new Blackwell B300 chip [6].

Summary of Key Points from Conference Call Company: NVIDIA Financial Performance - NVIDIA's Q3 revenue increased by 205% year-over-year, reaching $57 billion, with a gross margin of 73.6% [1][2] - For Q4, the company expects a revenue midpoint of $65 billion, representing a 65% year-over-year growth, with an adjusted gross margin target of 75% [1][2] - NVIDIA anticipates maintaining a gross margin in the mid-70% range for fiscal year 2026, alleviating market concerns regarding potential impacts from storage price increases and ASIC advancements [1][2] Supply Chain and Manufacturing - NVIDIA is enhancing supply chain resilience by collaborating with TSMC to produce Blackwell wafers in the U.S. and expanding manufacturing capabilities with partners like Foxconn and Anker [1][4] - The company aims to ship 20 million units of Blackwell and Ruby series by 2026, targeting $50 billion in revenue, with potential upward adjustments [1][4] - New procurement agreements from Saudi Arabia are valued at approximately $10 billion, and additional collaboration with Anthropic is expected to generate $30-35 billion in demand [1][4] AI Market and Investment Strategy - NVIDIA's CEO Jensen Huang addressed concerns about AI bubbles and GPU depreciation cycles, asserting that a 5-6 year depreciation period for GPUs is reasonable and can extend with increased inference workloads [1][5] - The company maintains a proactive investment strategy in AI, ensuring that investments are contingent on the successful development of AI systems by partners [1][7] Market Demand and Supply Issues - NVIDIA faces a strong demand for GPUs, with many older models still in high demand, indicating a robust market environment [1][8] - Corwave reported a 95% renewal rate for H100 contracts, reflecting strong market demand for NVIDIA's products [1][8] Company: Google Product Launch and Market Impact - Google's release of the Gemini 3 model has positively impacted its stock and that of other cloud providers, indicating overall growth in the AI market [1][3][6] - Gemini 3 Pro achieved top scores in 19 out of 20 industry benchmarks, showcasing its superior performance in inference, multi-modal tasks, and programming capabilities [1][9] Industry Dynamics - The launch of Gemini 3 demonstrates the effectiveness of pre-training scaling laws and intensifies competition among major players in multi-modal technology [1][10] - Google's high pricing strategy for its top-tier models suggests that premium models can command higher fees, creating unique barriers and driving increased investment in computational power across the industry [1][10] Market Sentiment - Discussions around AI bubble risks have eased, with NVIDIA's financial results and Google's model launch contributing to a positive outlook for AI technology and its commercial applications [1][11][12]

Core Viewpoint - The article discusses the potential release of the H200 chip in China, highlighting its specifications and performance compared to domestic AI chips, while also considering the geopolitical context surrounding the decision [2][3][20]. Specifications and Performance - The H200 chip features significant improvements over the H100, including 141 GB of HBM3e memory and a memory bandwidth of 4.8 TB/s, compared to the H100's 80 GB and 3.35 TB/s [9]. - The H200's FP64 Tensor Core performance is 34 teraFLOPS, which is competitive with other high-end chips like the B200 and H100 [18]. Market Context - The article notes that the H200 is currently priced higher than the B200 in certain cloud service providers due to its suitability for high-precision computing and its scarcity [17]. - The usage of H200 in overseas cloud servers remains high, driven by legacy workloads that are difficult to migrate from older cards [19]. Geopolitical Considerations - The potential release of the H200 in China is contingent on the U.S. government's stance, particularly the influence of hawkish advisors [3][20]. - The article suggests that if the U.S. does allow the release of the H200, it is likely that China would follow suit [20].