尊敬的各位嘉宾、各位网友： 大家好！ 欢迎大家参加摩尔线程智能科技（北京）股份有限公司首次公开发行股票并在科创板上市的网上路演活 动。在此，我谨代表摩尔线程，向长期关心、支持公司发展的广大投资者表示热烈欢迎！向一直以来关 注我国GPU技术突破、智算产业发展的社会各界朋友表示衷心的感谢！很高兴能借助今天上证路演中 心、上海证券报及中国证券网的互动交流平台，与大家真诚沟通、共同探讨摩尔线程的发展与未来。 当前，全球科技竞争格局正在深刻变革，GPU作为支撑通用人工智能、数字孪生、具身智能等前沿产业 的核心算力引擎，其战略地位已升至前所未有的高度。公司自2020年成立以来，始终专注于全功能GPU 的自主研发与设计，是国内高端AI芯片领域极具代表性的企业。公司的目标是成为具备国际竞争力的 GPU领军企业，为融合人工智能和数字孪生的数智世界打造先进的加速计算平台。 公司的发展路径与国家推动高水平科技自立自强的战略方向同频共振。基于自主研发的MUSA统一系统 架构，摩尔线程实现了单芯片同时支持AI计算加速、图形渲染、物理仿真和科学计算、超高清视频编 解码的技术突破，为构建自主可控的高性能算力底座奠定了关键的技术基础。 目前 ...

炒股就看金麒麟分析师研报，权威，专业，及时，全面，助您挖掘潜力主题机会！ （来源：每日经济新闻） 每经记者：可 杨 每经编辑：魏官红 大模型行业正站在一个拐点。过去数年由技术突破点燃的狂热，正在2025年遭遇一场商业化大考。随着 智谱等头部"独角兽"公司相继启动IPO（首次公开募股）流程，市场叙事正从模型能力的"军备竞赛"转 向落地应用的审视。 资本市场开始用更严苛的尺度检验大模型企业，是否具备可持续的商业模式和长期价值。这一冲突，在 开源与闭源的路线之争上尤为激烈。 图 受访者供 在行业普遍认知中，开源意味着共享，商业化则意味着独占。作为中国大模型赛道的关键玩家，智谱却 始终坚持一个反直觉的判断：开源与商业化长远来看并不冲突。这一判断的底层逻辑是什么？在技 术"摸高"与应用落地之间，企业应如何平衡资源？AI（人工智能）的"性价比"竞争是否已经到来？带着 这些问题，11月17日，《每日经济新闻》记者（以下简称NBD）与智谱董事长刘德兵进行了一场深度 对话。在他看来，基础模型的进步对应用落地有巨大的促进作用，但技术"摸高"绝不能做"空的、虚 的"东西，必须能转化成商业优势。 谈模型：大参数模型是技术锚点 NBD ...

Core Viewpoint - The article discusses the recent decline in U.S. stock markets, attributing it to three main pressures: the diminishing expectations of interest rate cuts by the Federal Reserve, rising concerns over an AI investment bubble, and historically high valuations in the stock market [1][2]. Group 1: Reasons for U.S. Market Adjustment - The Federal Reserve's expectations for interest rate cuts have significantly decreased, with a higher probability of pausing cuts in December due to hawkish signals from officials and the impact of government shutdowns on employment data [1][2]. - Concerns about an AI investment bubble are growing, driven by a mismatch between exponential growth in capital expenditure and linear growth in revenue from AI applications. Nvidia's strong earnings report did not alleviate market fears, as its revenue is tied to capital spending rather than actual market demand [2]. - U.S. stock valuations are at historical highs, with the Shiller P/E ratio exceeding levels seen in 2021 and 1929, only surpassed by the peak of the 2000 internet bubble. This suggests that the market has priced in overly optimistic growth expectations, limiting further valuation expansion [2]. Group 2: Outlook for U.S. Markets - The impact of interest rate expectations is likely to be short-term, with a higher probability of a dovish stance from the Federal Reserve in the future. Although December's rate cut is uncertain, rates may drop to around 3% by the end of 2026 [3]. - The core contradiction in the U.S. market lies between high valuations and the uncertain prospects of AI. While AI's potential remains, the timeline for its widespread productivity enhancement is uncertain, leading to justified concerns about an AI bubble [3]. - It is recommended to adjust annual return expectations to align with single-digit profit growth rates and to prepare for potential market corrections of 10%-20%. The current market has already seen a 5% correction, but valuations have not yet returned to reasonable levels, suggesting a continued wait for better entry points [4]. Group 3: Outlook for A-shares and H-shares - A-shares and H-shares experienced a synchronized adjustment due to external market declines and prior pressure releases, influenced by the drop in U.S. markets and changing expectations regarding the Federal Reserve's interest rate decisions [5]. - The core factors affecting A-share performance remain its fundamentals and liquidity. A dovish path from the Federal Reserve is expected to continue, with domestic asset allocation likely favoring equity markets in a low-interest environment [5]. - After the current adjustment phase, A-shares and H-shares are anticipated to continue rising in the following year, supported by improved performance in a recovering inflation environment [5]. Group 4: Sector Insights - High-valuation technology stocks are sensitive to liquidity changes and may face adjustment pressures, while dividend stocks and technology sectors exhibit a seesaw effect, with dividend stocks currently showing advantages [6]. - Consumer stocks have been less affected by liquidity expectations due to their adjusted valuations, presenting opportunities for left-side positioning despite limited fundamental improvement [6]. - The Hang Seng Technology Index has seen a 20% adjustment, with historical bull market corrections typically ranging from 20%-30%, indicating potential for increased focus once adjustments are complete [6].

Core Viewpoint - The rapid advancement of generative AI and large models contrasts with the slow commercial adoption, as evidenced by a recent decline in the percentage of U.S. companies using paid AI products [2][3]. Group 1: AI Project Challenges - Approximately 90% of vertical enterprises do not truly understand AI, leading to ineffective implementation without tailored models [3]. - The telecom industry has historically absorbed new technologies, and AI is seen as the next evolution, with significant advancements expected by 2025 [3]. Group 2: Future of AI and Agent Technology - The AI industry is at a crossroads, with a shift from foundational model development to large-scale application deployment, raising questions about the future of basic model research [6]. - There is a consensus that future AGI will rely on world models that integrate multiple modalities, although specific applications may require tailored models for efficiency [6][7]. - The development of specialized models for various industries is viewed as a practical approach to achieving commercial viability before moving towards universal models [7]. Group 3: Agent Technology Implementation - By 2025, agent technology is expected to become a core trend, with practical applications emerging across various industries, including healthcare and education [8]. - Current implementations of agent technology have demonstrated effectiveness, with plans for broader deployment in 2026 [8]. - Challenges remain in integrating agents into existing workflows, primarily due to limitations in multi-modal capabilities of large models [8][9]. Group 4: Computational Power and Industry Growth - The AI industry faces ongoing challenges related to computational power, with domestic GPU companies accelerating their development to address these needs [9]. - As computational issues are resolved, significant advancements in multi-modal models and agent technology are anticipated [9][10]. Group 5: Consumer Acceptance and Market Trends - Consumer acceptance of AI products is increasing, with a shift towards deploying AI capabilities from cloud to edge devices [9][10]. - The mobile AI sector is expected to see rapid growth, with small models achieving high accuracy in practical applications [11]. Group 6: Humanoid Robots and Industry Development - Humanoid robots are still in the exploratory phase, with significant technical challenges remaining before widespread commercial deployment [12][13]. - The manufacturing of humanoid robots involves complex components, with a focus on developing autonomous control capabilities as a critical bottleneck [13]. - The path to commercial viability for humanoid robots is expected to begin in industrial settings before expanding to consumer applications [14][15].

当人工智能的浪潮从虚拟世界涌向物理现实，汽车产业正站在这场变革的最前沿。自动驾驶领域技术路线纷繁复杂，"端到端"、"世界模型"、"VLA"等术语 层出不穷，业界对于何为"最优解"莫衷一是。 在由南方财经全媒体集团指导、21世纪经济报道主办的2025新汽车年度盛典上，南方财经全媒体集团编委会成员、集团客户端编委会执行总编辑、21世纪经 济报道编委会成员袁丁主持题为《汽车有"魂"，AI如何重塑用车体验？》的圆桌论坛，邀请到地平线副总裁吕鹏、千里智驾首席科学家秦海龙和场景实验室 创始人吴声三位业界专家，旨在穿透技术路线的迷雾，深入探讨AI如何超越工具属性，为汽车注入可进化、可感知的"灵魂"，从而彻底重塑人类的用车体 验。 场景实验室创始人吴声旗帜鲜明地指出，AI驾驶已超越"工具"范畴，正成为一种真实的生活方式，是通向AGI(通用人工智能)的最佳实践。 袁丁：当AI加速渗透物理世界，自动驾驶领域正处在充满分歧的路口。端到端、世界模型、VLA路线，哪条路线是自动驾驶的"最优解"？我们还要等待多久 才能迎来物理AI的曙光？吴声老师曾提出"AI场景革命元年"的概念，在汽车行业，AI场景应如何定义？目前是处于元年，还是已进入 ...

Core Insights - Google DeepMind has launched SIMA 2, a general AI agent capable of autonomous gaming, reasoning, and continuous learning in virtual 3D environments, marking a significant step towards general artificial intelligence [1][4] - SIMA 2 represents a major advancement from its predecessor, SIMA, evolving from a passive instruction follower to an interactive gaming companion that can plan and reason in complex environments [4][7] Development and Capabilities - SIMA 2 integrates advanced capabilities from the Gemini model, allowing it to understand user intentions, plan actions, and execute them in real-time, enhancing its interaction with users [4][11] - The new architecture enables SIMA 2 to perform multi-step reasoning, transforming the process from language to action into a more complex chain of language to intention to planning to action [11][16] - SIMA 2 demonstrates improved generalization and reliability, successfully executing complex instructions in unfamiliar scenarios, such as new games [16][22] Learning and Adaptation - SIMA 2 exhibits self-improvement capabilities, learning through trial and error and feedback from the Gemini model, allowing it to tackle increasingly complex tasks without additional human-generated data [25][28] - The agent's ability to transfer learning concepts across different games signifies a leap towards human-like cognitive generalization [22][29] Future Implications - SIMA 2's performance across various gaming environments serves as a critical testing ground for general intelligence, enabling the agent to master skills and engage in complex reasoning [29][30] - The research highlights the potential for SIMA 2 to contribute to robotics, as the skills learned are foundational for future physical AI assistants [30][31]

Core Viewpoint - Google DeepMind has launched SIMA 2, a general AI agent capable of autonomous gaming, reasoning, and continuous learning in virtual 3D environments, marking a significant step towards general artificial intelligence [2][3][6]. Group 1: SIMA 2 Overview - SIMA 2 represents a major leap from its predecessor, SIMA, evolving from a passive instruction follower to an interactive gaming companion that can autonomously plan and reason in complex environments [6][10]. - The integration of the Gemini model enhances SIMA 2's capabilities, allowing it to understand user intentions, formulate plans, and execute actions through a multi-step cognitive chain [15][20]. Group 2: Performance and Capabilities - SIMA 2 can understand and execute complex instructions with higher success rates, even in unfamiliar scenarios, showcasing its ability to generalize across different tasks and environments [24][30]. - The agent demonstrates self-improvement capabilities, learning through trial and error and utilizing feedback from the Gemini model to enhance its skills without additional human-generated data [35][39]. Group 3: Future Implications - SIMA 2's ability to operate across various gaming environments serves as a critical testing ground for general intelligence, enabling the agent to master skills and engage in complex reasoning [41][43]. - The research highlights the potential for SIMA 2 to contribute to robotics and physical AI applications, as it learns essential skills for future AI assistants in the physical world [43].

Core Insights - The future of AI may hinge on understanding the evolutionary codes of the human brain, as highlighted by Yann LeCun's departure from Meta to focus on "World Models" [1] - Fei-Fei Li emphasizes that the advancement of AI should pivot from merely expanding model parameters to embedding "Spatial Intelligence," a fundamental cognitive ability that humans possess from infancy [1][3] - The launch of Marble by World Labs, which utilizes multimodal world models to create persistent 3D digital twin spaces, marks a significant step towards achieving spatial intelligence in AI [1] Group 1: AI Development Perspectives - Yann LeCun's vision diverges from Meta's focus on large language models (LLMs), arguing that LLMs cannot replicate human reasoning capabilities [3] - LLMs are constrained by data quality and scale, leading to cognitive limitations that hinder their ability to model the physical world and perform dynamic causal reasoning [3][4] - The reliance on text data restricts AI's ability to break free from "symbolic cages," necessitating a shift towards a structured understanding of the world for true AI evolution [4] Group 2: World Models vs. Large Language Models - World models are seen as a solution to the fundamental limitations of LLMs, focusing on high-dimensional perceptual data to model the physical world directly [4][5] - The key characteristics of world models include internal representation and prediction, physical cognition, and counterfactual reasoning capabilities [11] - A complete world model consists of state representation, dynamic models, and decision-making models, enabling AI to simulate and plan actions in a virtual environment [12][13] Group 3: Industry Trends and Innovations - Recent advancements in world models have been made by major tech companies, with Google DeepMind's Genie series and Meta's Code World Model leading the charge [16] - The concept of "physical AI" is gaining traction, with Nvidia's CEO asserting that the next growth phase will stem from these new models, which will revolutionize robotics [16] - The application of world models is already influencing various sectors, including autonomous driving and robotics, as companies like Tesla integrate these models for real-world learning and validation [17] Group 4: Challenges and Future Directions - The development of world models faces technical challenges, including the need for extensive multimodal data and the lack of standardized training datasets [20] - Cognitive challenges arise from the complexity of decision-making processes within world models, raising concerns about transparency and alignment with human values [20][21] - Despite the challenges, the global competition in the world model space is intensifying, with the potential to redefine industries and enhance human-AI collaboration [21][22]

[Table_Title] 研究报告 Research Report 20 Nov 2025 中国电子 China (Overseas) Technology Google Gemini 3 ：双版本发布&多模态更新 Google Gemini 3: Dual-Version Launch & Multimodal Updates 姚书桥 Barney Yao barney.sq.yao@htisec.com [Table_yemei1] 热点速评 Flash Analysis [Table_summary] (Please see APPENDIX 1 for English summary) 事件 Gemini 3 于 11 月 18 日正式发布，首日即实现亿级用户覆盖与多项行业基准测试登顶。作为谷歌迄今最强 AI 模型， 其核心更新包括发布当天同步集成至谷歌搜索 AI Mode、Gemini App 及企业级平台，直接触达 20 亿 AI Overviews 用户 与 6.5 亿 Gemini App 月活用户，创下行业最快分发纪录。该模型推出 Pro 标准版与 Deep Think 增强推理版双版本， ...

Core Insights - China's artificial intelligence (AI) technology is rapidly advancing, closing the gap with the United States, as evidenced by Tsinghua University's leading position in global AI research and patent filings [1][2][4] Group 1: Research and Development - Tsinghua University has published the highest number of AI papers among global universities, with 4,986 AI-related patents granted from 2005 to the end of 2024, including over 900 new patents in the last year [1][4] - Despite China's advancements, the U.S. still holds the most influential patents and superior AI models, with 40 notable AI models developed by U.S. institutions compared to 15 from China [1][2] Group 2: Talent and Innovation - The proportion of top global AI researchers from China increased from 10% to 26% between 2019 and 2022, while the U.S. share decreased from 35% to 28% [2] - Tsinghua University is fostering a collaborative environment for AI innovation, with several startups founded by its graduates, such as DeepSeek, which has developed a competitive large language model [5][6] Group 3: Educational Initiatives - Tsinghua University is integrating AI technology across various disciplines, providing subsidies for students to access new AI computing platforms for research [6][7] - The university's Brain and Intelligence Laboratory is producing innovative AI models, such as the Hierarchical Reasoning Model (HRM), which outperforms larger models from U.S. companies in specific tasks [5][6]