阿里巴巴集团董事兼CEO吴泳铭发表主旨演讲。他表示，AGI是具备人类公共认知能力的智能系统，现 在在行业看来已经成为现实的世界。但吴泳铭指出，AGI并非是发展的终点，只是全新的起点，AI肯定 不会止步于AGI， 它将迈向超越人类智能能够自我迭代进化的超级人工智能行业，称为ASI超级人工智 能。吴泳铭称，资本开支会在3800亿基础上，增加更多的投入。 【免责声明】本文仅代表作者本人观点，与和讯网无关。和讯网站对文中陈述、观点判断保持中立，不 对所包含内容的准确性、可靠性或完整性提供任何明示或暗示的保证。请读者仅作参考，并请自行承担 全部责任。邮箱：news_center@staff.hexun.com 阿里巴巴短线拉升，现涨超4%，报价165.8港元，夜盘涨近3%。 | 09988 阿里巴巴-W | | | | --- | --- | --- | | 165.800↑+6.400 +4.02% | | 交易中 09/24 10:40 | | SEES BEBER | | ● 快捷交易 | | 最高价 165.900 | 开盘价 158.500 | 成交量 5635.01万 | | 最低价 157.600 | 昨 ...

新模型、Agent（智能体）应用等AI软件产品及AI服务器等基础设施硬件或是看点。阿里巴巴CEO吴泳铭表示，大模型是 下一代操作系统。AI Cloud是下一代计算机。未来全世界或只会有5-6个超级云计算平台。 作者 | 龙 玥 编辑 | 硬 AI 9月24日，阿里巴巴2025云栖大会在杭州正式开幕。 本次大会将于24日至26日举行三天，以"云智一体·碳硅共生"为主题，设置 三大主论坛"云栖前瞻""云栖技 术""无法计算的价值"，共同探讨AI、云计算与产业应用的最新趋势。 硬·AI 新模型、Agent（智能体）应用等AI软件产品及AI服务器等基础设施硬件或是看点。上证报记者获悉，阿 里云将展示超大规模集群、分布式训练、推理加速等能力，并首次展出单柜支持144个计算节点的高密度 AI服务器和新一代高性能网络架构HPN8.0。 阿里巴巴CEO吴泳铭发表主旨演讲称， 资本开支会在3800亿基础上，增加更多的投入。通义千问的目标 是打造AI时代的安卓系统。大模型是下一代操作系统。AI Cloud是下一代计算机。未来全世界或只会有5- 6个超级云计算平台。 01 吴泳铭： ASI才是开始 大模型将是下一代的操作系统 大会 ...

Core Insights - Alibaba is actively advancing its AI infrastructure with a planned investment of 380 billion RMB, aiming to enhance its cloud computing capabilities significantly by 2032 [1] - The company is focusing on the development of Artificial General Intelligence (AGI) and Super Artificial Intelligence (ASI), with a roadmap that includes three stages: emergence of intelligence, autonomous action, and self-iteration [1] - Alibaba's recent product launches include six new AI models and APIs, showcasing its commitment to rapid innovation in the AI space [4][6][7] Investment and Financial Performance - Alibaba's cloud intelligence group reported a 26% year-on-year revenue growth in Q2 2025, with AI-related product revenue achieving triple-digit growth for the eighth consecutive quarter [9] - The company maintains its target of investing 380 billion RMB in cloud and AI hardware infrastructure over the next three years, adapting to supply chain uncertainties through partnerships [9] - The market share of Alibaba's Tongyi models in the enterprise market is 17.7%, making it the most chosen large model by Chinese enterprises [8] Product Development and Innovation - The newly launched products include a security model (Guard), a travel planner (Qwen Travel Planner), a live translation API, an upgraded programming model API (Qwen3-Coder), a visual-language model (Qwen3-VL), and a flagship model (Qwen3-Max) [5][6][7] - The Qwen3-Max model, with a parameter count of 1 trillion, significantly enhances capabilities in understanding complex instructions and tool usage, positioning it among the top global models [7] - Alibaba's rapid iteration of models is aimed at maintaining a competitive edge in the fast-evolving AI landscape, with a focus on commercial applications and performance metrics [8]

消息面上，9月24日，阿里巴巴集团董事兼首席执行官，阿里云智能集团董事长兼首席执行官吴泳铭在 2025云栖大会上表示，实现AGI已是确定性事件，但这仅是起点，终极目标是发展出能自我迭代、全面 超越人类的超级人工智能（ASI）。吴泳铭称，目前阿里正积极推进3800亿的AI基础设施建设，并计划 追加更大的投入。为了迎接ASI时代的到来，对比2022年这个GenAI的元年，到2032年阿里云全球数据 中心的能耗规模将提升10倍。这意味着阿里云算力投入将指数级提升。 阿里巴巴-W(09988)拉升逾4%，高见166港元再度刷新近四年来新高，本月股价已累涨超四成。截至发 稿，涨4.04%，报166港元，成交额107.94亿港元。 ...

Core Viewpoint - Alibaba is positioning itself as a leader in AI and cloud computing, emphasizing the development of large models and AI infrastructure as critical to its future growth and competitiveness in the global market [1][9]. Group 1: AI Infrastructure and Development - Alibaba is investing 380 billion yuan in AI infrastructure, with plans for further investment to support the anticipated growth in AI capabilities [1][8]. - The company predicts that by 2032, the energy consumption of its global data centers will increase tenfold compared to 2022, indicating a significant escalation in computational power [1]. - The CEO of Alibaba Cloud stated that achieving Artificial General Intelligence (AGI) is a certainty, with the ultimate goal being the development of Super Artificial Intelligence (ASI) to tackle major scientific challenges [1]. Group 2: New Product Launches - Alibaba recently launched six new products, including an open-source security model, a travel planning AI, a real-time translation API, an upgraded programming model API, a visual-language model, and a flagship model with enhanced capabilities [4][5][6]. - The new products aim to enhance user experience and operational efficiency across various applications, particularly in e-commerce and programming [5][6]. Group 3: Market Position and Performance - Alibaba's AI models have gained significant traction, with a reported 17.7% market share in the enterprise-level large model consumption in China, making it the most chosen model by enterprises [7]. - The company's cloud intelligence group reported a 26% year-on-year revenue growth, with AI-related product revenue achieving triple-digit growth for eight consecutive quarters [8]. - Analysts suggest that the "cloud + AI" strategy will serve as a second growth curve for Alibaba, potentially enhancing its valuation and competitive edge in the market [9].

大会上，吴泳铭提出了"超级人工智能（ASI）"这一概念。在他看来，ASI时代开源模型创造的价值和 能够渗透的环境将会远远大于闭源模型。"我们坚定的选择开源，就是为了全力支持开发者生态，与全 球所有开发者一起探索应用的无限可能。"吴泳铭表示。（文猛） 责任编辑：江钰涵 新浪科技讯 9月24日上午消息，今日举办的2025云栖大会上，阿里巴巴集团董事兼首席执行官、阿里云 智能集团董事长兼首席执行官吴泳铭在分享中指出，通义千问选择开放开源的路线，这背后的目标是希 望打造"AI时代的Android系统"。 ...

Core Viewpoint - SoftBank's Vision Fund is set to lay off nearly 20% of its workforce, marking the largest layoffs in venture capital history, as it shifts its investment focus towards artificial intelligence [1][2]. Group 1: Layoff Details - The Vision Fund, established in 2017, has a current scale of $157.3 billion, including its first and second funds and a Latin American fund [2]. - The fund peaked at 474 employees in March 2020 but has since reduced its workforce to over 300 employees [2]. - The layoffs occur despite strong performance, with the fund reporting a net revenue of 1.82 trillion yen for Q1 FY2025-2026, a 7% year-on-year increase, and a net profit of 421.82 billion yen, surpassing market expectations [2]. Group 2: Strategic Shift - The layoffs are part of a broader strategy to focus on bold investments in AI and breakthrough technologies, as stated by a Vision Fund spokesperson [3]. - The fund will continue to invest, with remaining employees directed towards ambitious projects like the $500 billion Stargate initiative in collaboration with OpenAI and Oracle [3]. Group 3: AI Investment Focus - The shift towards AI marks a departure from previous diversified investment strategies, aligning with SoftBank's recent capital-intensive investments in AI infrastructure and models [4]. - SoftBank has invested $9.7 billion in OpenAI and is pursuing a capital-intensive infrastructure strategy centered around its chip designer Arm, acquiring several AI chip companies [4]. - As of March 2023, SoftBank's investments in Nvidia and TSMC have increased to approximately $3 billion and $330 million, respectively [4]. Group 4: Financial Position - SoftBank's CEO Masayoshi Son emphasizes the importance of achieving Artificial Super Intelligence (ASI) and has outlined four key areas of focus: AI chips, AI robots, AI data centers, and energy [5]. - The company has a strong cash reserve of 4 trillion yen (approximately $27 billion), bolstered by the sale of T-Mobile shares, providing a solid foundation for aggressive investments in AI [5][6]. - The CFO stated that SoftBank's balance sheet supports its aggressive investments in AI, including a $30 billion investment in OpenAI and participation in the $500 billion Stargate project [6].

Core Insights - The article discusses the advancements of an AI system named Physics Supernova, which has shown remarkable performance in solving complex physics problems, indicating a significant breakthrough in AI's ability to tackle scientific reasoning tasks [1][3][4]. Group 1: AI System Performance - Physics Supernova achieved a score of 23.5 out of 30 in the 2025 International Physics Olympiad, ranking 14th among 406 participants and surpassing the average score of human gold medalists [1][3]. - The system utilizes a combination of image analysis and answer verification tools, enhancing its reasoning and problem-solving capabilities in complex scientific contexts [3][4]. Group 2: Technological Framework - The AI system is built on the smolagents framework and employs the CodeAgent architecture, allowing for flexible self-planning and dynamic tool invocation based on problem-solving progress [4][6]. - Two specialized tools, ImageAnalyzer and AnswerReviewer, are integrated into the system to improve its performance in interpreting experimental results and self-correcting errors [6][9]. Group 3: Future Directions - The research team aims to expand the capabilities of AI systems to perform experimental tasks, moving beyond theoretical problems to include practical applications in physics [10][11]. - Future developments may focus on creating methods for verifying physical expressions and enhancing the answer review system with broader knowledge tools [11][12].

Group 1 - The 12th TiD Quality Competitiveness Conference was held in Beijing, focusing on "New Quality Productivity and High-Quality Development of Software" with over 200 experts and industry elites discussing new paths and opportunities for China's software industry [1] - The director of the Haidian District Science and Technology and Economic Information Bureau emphasized the need for traditional software vendors to embrace new technologies and software innovation, including clear AI strategies and optimizing user experiences [1] - The conference highlighted the systemic reconstruction of quality definitions and evaluation methods in the context of AI products and services, as stated by a member of the National Manufacturing Power Construction Strategic Advisory Committee [1][2] Group 2 - The conference featured nearly 30 specialized sessions discussing disruptive practices such as AI-native development and large model-driven testing, aiming to build a safe, reliable, and efficient software industry ecosystem [3] - Emerging industry forums were set up to discuss topics like precision medicine and health management practices in the era of large AI models, indicating a focus on future industry content layout [3]

Core Insights - Alibaba Cloud has launched the new Qwen3 model, which is the first "hybrid reasoning model" in China, integrating "fast thinking" and "slow thinking" into a single model, significantly reducing deployment costs and enhancing performance compared to previous models [1][4] Group 1: Model Performance and Architecture - Qwen3 features a total parameter count of 235 billion, with only 22 billion activated, and utilizes a mixture of experts (MoE) architecture [2][3] - The model has achieved a performance leverage of over 10 times with its 30B parameter MoE model, requiring only 3 billion to match the performance of the previous Qwen2.5-32B model [3] - Qwen3 has outperformed global top models like DeepSeek-R1 and OpenAI-o1 in various benchmarks, securing its position as the strongest open-source model globally [1][2] Group 2: Cost Efficiency and Deployment - The deployment cost for Qwen3 has significantly decreased, requiring only 4 H20 units for full deployment, with memory usage being one-third of that of DeepSeek-R1 [1][3] - All Qwen3 models are hybrid reasoning models, allowing users to set a "thinking budget" for performance and cost optimization in AI applications [3][4] Group 3: Future Developments and Goals - Future enhancements for Qwen3 will focus on expanding data scale, increasing model size, extending context length, and broadening modality range, while leveraging environmental feedback for long-term reasoning [4] - The Qwen3 team views this launch as a significant milestone towards achieving general artificial intelligence (AGI) and superintelligent AI (ASI) [4]