（Re uters） 中国股市科技股持续表现强劲，市场越来越确信，中国人工智能不借助英伟达等美国企业 的力量也能发展。在年初的"DeepSeek冲击"中，导致美国科技股暴跌。虽然目前美国降息的预 期支撑着股市，但DeepSeek 极有可能再次掀起波澜…… 在中国的股票市场，科技股持续表现强劲。市场越来越确信，中国人工智能（AI）不借助英伟达等美 国企业的力量也能发展。契机是上周悄然发布的中国新兴AI企业DeepSeek的新型大语言模型 （LLM）。 美国总统特朗普表示将建立英伟达向美国政府上缴对华销售收入15%的机制，由于面向中国的人工智能 芯片"H20"相关产品被报道将停止生产，受此影响，对中国AI寄予厚望。 中国人工智能开发能否摆脱英伟达？ DeepSeek于8月21日发布了新LLM"V3.1"。打破3个月的沉默发布的并不是期待已久的"R2"。有报道 称，R2在华为半导体上未能达到期待的能力，开发被推迟。V3.1可以说是R2之前的"过渡"。 市场之所以逐渐关注，是因为DeepSeek的官方说明表示DeepSeek-V3.1 使用了 UE8M0 FP8 Scale的参数 精度。在评论区，DeepSeek在置 ...

今天，我们就会从这些维度，带你拨开数据的表面，看到背后真正的逻辑。本次分析，将是全网独一无二的深度解读。如果你想领先市场一步，千万不要 错过任何一个细节。 英伟达未来三大催化剂 英伟达如今的市值已经飙到 4.34 万亿美元（截止上周五），比日本一整年的GDP还要高。 英伟达之所以这么牛绝不仅仅因为GPU卖得好，那只是表层的故事。真正让资本市场疯狂的是，它数次财报里透露出来的对未来AI时代垄断地位的巨大 预期。 我们美股投资网是唯一敢在英伟达财报前，明确亮明自己观点的频道！从2024年2月19日至今，我们已经连续做了六次英伟达财报前瞻，其中 5次全部正 确。 这一次我们也参考了大量的资料，力求用数据说话！ 我们这次对英伟达财报判断是：大概率超预期。但重点在于，英伟达的股价已经提前消化了大量利好，想要在这个位置来一波大涨，难度比之前可要大得 多。换句话说，涨没问题，但不期望会暴涨。 曾成功投资过脸书推特Robinhood和Coinbase的美国顶级的风险投资（a16z）表示过，英伟达的优势，并非只是GPU性能，还有其围绕芯片，在网络、内 存、供应链及产业生态上建立的护城河。 问题是，资本市场的逻辑残酷——没有永远 ...

群友2： 阿里和字节有芯片，目前可能不如寒武纪。 群友1： 芯片最大的挑战不是芯片本身， 是能不能造出一个生态，类似于大家都以为英伟达是个纯芯片公司， 但是忽略了他的生态， 类似CUDA 和 TensorFlow 这类东西可以兼容各种平台， 其实英伟达是个算力模拟公司。寒武纪基本没有软件生态能力， 就硬件来说，不和英伟达比 ，和华为海思就 差距很大。 这个只能算是远期的畅想（本质是不严谨的），走到那步还有很长很长的路，咱们也不可能帮李想去做决策和判断。 TOP2:这块能细化展开一下吗？ 群友1： 其实特斯拉给了另外一个很好的思路， 是芯片。 大算力芯片， AI6 基本可以替代Dojo了， 也就是说理想如果神经网络芯片+软件能力很强， 基 本就可以锚定寒武纪，做超算。 但本质还是得从业务和商业角度出发。 群友1： 抱歉，我对半导体研究不多，没法给出一个详细有力的论述。 但特斯拉最近把Dojo叫停了 （本身用于云端计算的大算力芯片，本来想一半用英 伟达一半用Dojo），其实不是放弃，大概率是Musk 发现这个2nm 的 AI6芯片既能上车还能组大集群做云端算力了。现在国家明显就是缺大算力芯片（看 最近的寒武纪就知 ...

万亿市值的AI巨头英伟达，它的创业历程，远比它的芯片技术更值得深思。 很多人看到的是技术红利，却没看见背后数次生死边界的挣扎：产品方向错误、主流标准冲突、资金链几近断裂…… 而黄仁勋带领团队走出绝境的方式，不是靠运气，而是依靠清醒的自我认知、坚定的技术路线，以及一套极度扁平、高度透明的反常规管理机制。 他顶住压力投入6亿美元开发CUDA，被外界认为"战略浪费"，却为英伟达今天成为"AI霸主"埋下关键伏笔。 今天这篇文章梳理了英伟达跨越32年的关键抉择与底层逻辑： 如何面对错误并及时止损？ 怎样在技术路线之争中保持战略定力？ 如何用"使命驱动"替代"层级管理"，打造出极致敏捷的组织？ 希望今天的分享，对你有所启发。 一、面对生死边缘，初生的英伟达差点夭折 1993年，30岁的黄仁勋与两位资深的微芯片设计师克里斯·马拉考斯基和柯蒂斯·普里姆共同创立了英伟达。 尽管黄仁勋是三个人里最年轻的，但他的学习能力让马拉考斯基和普里姆深信他更适合担任CEO。 他们三人的创业初衷是设计一款能让竞争对手"嫉妒得眼睛发绿"的图形芯片。公司最初命名为NVision，但后来他们发现这个名字已被一个卫生纸品牌使 用。 于是，黄仁勋建议使 ...

Core Viewpoint - AMD, under the leadership of Lisa Su, is positioning itself as a key player in the AI chip market, aiming to surpass Nvidia's dominance while navigating the complexities of U.S.-China relations regarding semiconductor exports [3][5][7]. Group 1: Company Performance and Strategy - Since Lisa Su became CEO in 2014, AMD's market capitalization has surged from approximately $2 billion to nearly $300 billion, showcasing a remarkable turnaround [5]. - AMD has successfully doubled its data center revenue from $6 billion in 2022 to $12.6 billion in 2023, indicating strong growth in high-performance computing [6][16]. - The company has adopted chiplet technology, which has proven to be highly beneficial, and launched the world's first 7nm data center GPU, enhancing its competitive edge [6]. Group 2: Competitive Landscape - AMD's market size remains significantly smaller than Nvidia's, which has a market capitalization of $4.4 trillion, highlighting the competitive challenges ahead [7]. - Lisa Su emphasizes that AMD's vision is not to directly compare itself with Nvidia or Intel but to focus on providing the best solutions across various computing needs [16]. Group 3: AI and Future Prospects - AMD is actively collaborating with major companies like OpenAI, Meta, and Tesla, aiming to establish itself as a strategic partner in the AI sector [6][16]. - The company is training its own AI models not to compete with large model builders but to learn and improve its products [19]. - Lisa Su believes that the future market for AI and computing will exceed $500 billion in the next three to four years, presenting significant opportunities for AMD [16]. Group 4: Geopolitical and Economic Considerations - Lisa Su advocates for bringing semiconductor manufacturing back to the U.S., citing national security and economic benefits, despite acknowledging the complexities involved [12][14]. - The recent U.S. tariffs on chips exported to China pose challenges, but AMD aims to continue its growth trajectory by expanding its user base globally [11][12]. Group 5: Leadership and Vision - Lisa Su is recognized as a prominent female leader in technology, focusing on long-term goals rather than immediate political pressures [5][14]. - She expresses a strong belief in the transformative potential of technology, particularly in healthcare, and aims to leverage AI to improve patient outcomes [26][32].

以下文章来源于睿远FUND ，作者小睿 睿远FUND . 睿远基金官方订阅号，第一时间发布睿远基金动态、分享优质基金及投资内容，做持有人长期利益最大 化的价值投资实践者。 美国半导体巨头英伟达在 6 月初超过微软成为全球市值最高的上市公司之后， 7 月初公司市值突破 4 万亿美元，成为有史以来首家达到这一重要里程碑的企业，当时的股价触及 164.32 美元的历史最高 点，而目前英伟达的股价已经超过了 180 美元。 市场普遍认为，英伟达这波股价的飙升，主要由于投资者对于人工智能变革潜力的坚定信心，并且英伟 达的主要合作伙伴 OpenAI 也在近期发布了最新的 GPT-5 ，英伟达的市值里程碑也凸显了企业正将资 产支出转向 AI 领域的发展方向。 英伟达，最初是游戏芯片制造商，随后转型为加密挖矿芯片制造商，如今则成为人工智能计算芯片制造 巨头，以及该领域无可争议的早期赢家。 那么， 英伟达是如何踏入人工智能计算芯片领域的呢？ 在《黄仁勋：英伟达之芯》一书中，作者为读者们呈现了当时英伟达是如何抓住了这个千载难逢的机 会。 千载难逢的机会 英伟达何以快速转型 在提到英伟达的 AI 之路，需要先介绍一个重要的人物，那 ...

Group 1: Quantum Computing Landscape - Quantum computing is emerging as a significant theme in the AI revolution, with companies like IonQ, Rigetti Computing, D-Wave Quantum, and Quantum Computing Inc. gaining attention [1] - Despite the interest, none of these companies have achieved the critical mass necessary to demonstrate truly disruptive products or services [2] - Nvidia is making strategic moves in the quantum computing space, positioning itself as a key player [2][5] Group 2: Nvidia's Position and Offerings - Nvidia has developed a software platform called CUDA, which supports AI applications and runs on its hardware [3] - The combination of CUDA and GPUs creates a strong ecosystem, allowing Nvidia to control both software and hardware aspects of AI development [4] - Nvidia's extension of CUDA, known as CUDA-Q, is designed for applications in both traditional and quantum computing [5][6] Group 3: Valuation Insights - Investors often misinterpret valuation by focusing solely on stock price, while seasoned investors consider multiple factors [8] - Quantum Computing Inc. has an extraordinarily high price-to-sales (P/S) ratio of nearly 4,800, with a market value of $2.6 billion and only $385,000 in sales [10] - IonQ, Rigetti, and D-Wave also exhibit high P/S multiples, reminiscent of stock market bubbles [11] Group 4: Nvidia's Investment Potential - Nvidia is considered the cheapest quantum computing stock in its peer group based on P/S multiples, despite appearing "expensive" [12] - The long-term upside for smaller speculative quantum computing stocks seems to be priced in, while Nvidia's valuation suggests potential for growth in emerging AI applications [13] - Nvidia's core chip business may provide opportunities in other areas of AI, making it a compelling buy-and-hold investment as quantum computing develops over the next decade [14]

Core Viewpoint - The article discusses the security risks associated with the H20 chip from Nvidia, particularly focusing on the potential for backdoors and the implications of U.S. government policies regarding chip exports to China [3][4][24]. Group 1: Security Risks and Backdoors - The National Internet Information Office of China recently addressed Nvidia regarding security risks related to the H20 chip, suggesting concerns over potential backdoors [3]. - Nvidia's response emphasized that the chip does not contain backdoors, referencing the historical "Clipper chip" incident as a cautionary tale [4][6]. - A recent U.S. legislative proposal led by Congressman Bill Foster aims to mandate U.S. chip companies to incorporate backdoors in export-controlled chips, indicating a shift in policy towards more overt government control [8][10]. Group 2: Technical Aspects of Backdoors - Backdoors can be categorized into hardware and software types, with hardware backdoors being physical circuits left during design or manufacturing, while software backdoors involve implanted instructions in software [11][12]. - The H20 chip's power management module could theoretically implement a remote shutdown feature by embedding a circuit that triggers under specific conditions, such as usage time or environmental factors [14][15]. - The potential for software-based backdoors exists through updates to the CUDA ecosystem, which could allow for tracking and data collection functionalities [16][17]. Group 3: U.S. Government Control Mechanisms - The U.S. government has proposed a "chip governance mechanism" to coordinate chip design and production, ensuring compliance with national security requirements [19][20]. - This mechanism includes features such as license locking, tracking, usage monitoring, and usage restrictions, which could effectively control the deployment of chips like the H20 [20][23]. - Reports indicate that Nvidia's AI chips already possess many of the functionalities required for this governance mechanism, raising concerns about their safety and reliability for foreign markets [21][23]. Group 4: Performance and Environmental Concerns - The H20 chip is deemed not only unsafe but also technologically inferior, with only about 20% of the performance of its standard counterpart, the H100, and a 41% reduction in GPU core count [27]. - Environmental efficiency is also a concern, as the H20's energy efficiency ratio is approximately 0.37 TFLOPS/W, failing to meet the required 0.5 TFLOPS/W for energy-efficient GPUs [28]. - Given its lack of safety, technological advancement, and environmental compliance, the H20 chip is not considered a viable option for consumers [29][30].

Core Viewpoint - Nvidia has stated that its chips do not contain "backdoors," specifically referencing the "Clipper chip" incident from the past [1][3]. Group 1: Historical Context - In 1992, AT&T introduced a hardware device for secure voice transmission, which led to government dissatisfaction and the introduction of the "Clipper chip" containing a "cryptographic backdoor" for government access [3]. - The "Clipper chip" faced widespread resistance and was terminated within three years, leading the U.S. government to adopt a more discreet approach regarding "backdoors" in technology [3]. Group 2: Current Legislative Actions - In May 2023, U.S. Congressman Bill Foster proposed legislation requiring U.S. chip companies to include "backdoors" in export-controlled chips, asserting that the technology is mature and feasible [4][6]. Group 3: Technical Feasibility of Backdoors - There are two types of backdoors: hardware and software. Hardware backdoors involve physical modifications during chip design, while software backdoors involve embedded instructions in software [6]. - The Nvidia H20 chip can theoretically implement a "remote shutdown" feature through its power management module, which could be triggered under specific conditions [6][8]. - Another method for a hardware backdoor involves modifying the H20 chip's firmware to restrict functionality based on certain conditions, such as geographic location [8]. Group 4: Software Ecosystem and Backdoors - The CUDA ecosystem, used by over 4 million developers, could potentially facilitate the activation of backdoors through software updates, allowing for tracking and data collection [9][11]. - The U.S. aims to maintain AI dominance through both hardware and software ecosystems, necessitating other countries to develop independent alternatives [11]. Group 5: On-Chip Governance Mechanism - The U.S. government has proposed an "on-chip governance mechanism" to coordinate chip design and production, which includes features like license locking, tracking, usage monitoring, and usage restrictions [12][14]. - Many leading chip manufacturers, including Nvidia, already possess the necessary functionalities for this governance mechanism, although some may not yet be activated [14]. Group 6: Concerns Regarding Nvidia's H20 Chip - The H20 chip is considered unsafe for China, as it lacks advanced features and has only about 20% of the performance of its standard counterpart, the H100 [17][20]. - The H20 chip's energy efficiency is approximately 0.37 TFLOPS/W, failing to meet the 0.5 TFLOPS/W standard set by China's green development initiatives [18][19]. - Given its lack of safety, advancement, and environmental compliance, the H20 chip is deemed a poor choice for consumers [20][21].

Core Viewpoint - Nvidia has stated that its chips do not contain "backdoors," specifically addressing the "Clipper chip" incident from the past [1][2]. Group 1: Historical Context of Backdoors - In 1992, AT&T launched a hardware device for secure voice transmission, which led to government dissatisfaction and the introduction of the "Clipper chip" containing a "backdoor" for government access [3][4][5]. - The "Clipper chip" faced significant resistance and was terminated within three years, leading the government to adopt a more discreet approach regarding "backdoors" [6]. Group 2: Current Legislative Developments - In May of this year, U.S. Congressman Bill Foster proposed legislation requiring U.S. chip companies to include "backdoors" in export-controlled chips [8]. - Foster, with a background in chip design, asserted that the technology to implement such features is mature and feasible [9]. Group 3: Technical Feasibility of Backdoors - There are two main types of "backdoors": hardware and software [12][20]. - Using Nvidia's H20 chip as an example, a hardware "backdoor" could be implemented through the power management module to enable remote shutdown capabilities [13][14]. - Software "backdoors" can be activated through updates to the CUDA ecosystem, which is widely used by developers globally [16][18]. Group 4: Implications of Backdoor Mechanisms - The "backdoor" mechanisms can facilitate tracking and remote disabling of chips, raising concerns about information security [19][23]. - The U.S. has developed a "chip governance mechanism" to coordinate chip design and production, ensuring control over AI chips [24][29]. Group 5: Concerns Regarding Nvidia's H20 Chip - The H20 chip, which is being exported to China, is considered unsafe due to its limited performance compared to the H100, with only about 20% of the overall computing power and a 41% reduction in GPU core count [33]. - The H20 chip's energy efficiency is also subpar, with a measured efficiency of 0.37 TFLOPS/W, failing to meet the required standards [37]. - Given its lack of advancement, environmental sustainability, and safety, the H20 chip is not seen as a viable option for consumers [40].