专题：2025《中国企业家》影响力企业家年会 3．在孙悟空的品牌重启中，高光时刻是大闹天宫。但是对一个创业者来说，如果你只共情孙悟空，说 明你还没有成长、成熟。它被压在五行山下五百年，这五百年它肯定反思了。 以下为李光斗现场演讲实录（有删减）： 作为企业家，2026年如何穿越周期？我们要了解几大转变： 认清现实，把握局势；跟对人，走正道，取真经，成正道。 文|《中国企业家》记者 苗诗雨 12月6~7日，由《中国企业家》杂志社主办的2025（第二十三届）影响力企业家年会（原中国企业领袖 年会），在北京召开。在5日进行的前沿大课环节，著名品牌战略专家李光斗作了主题为《商解西游： 商业大变局和品牌重启》的主题演讲。 以下为核心要点： 1.2026年将是中国的AI应用元年，每一个人都应该有自己的数字分身，如果我们没做好自己的数字分 身，可能未来都会被时代所抛弃。 2．孙悟空犯了中层领导常见的错误：第一叫能力陷阱，会做事不等于会管理。如果老板把所有的事都 做完了，你的天花板就是公司的天花板；第二叫效率陷阱，当你在所有事上都做成卷王了，大家都没饭 吃；还有身份错位，你到的是天庭，不是花果山小作坊，这里不是你的私营企业。 ...

Investment Rating - The report suggests focusing on sectors driven by domestic AI development such as servers and IDC, as well as sectors like servers and optical modules driven by overseas AI development [5] Core Insights - Marvell reported revenue of $2.075 billion for the quarter, exceeding market expectations and guidance by $15 million, and announced the acquisition of Celestial AI for approximately $3.25 billion to enhance its position in the CPO field [1][50] - AWS launched the AI training chip Trainium 3 with a computing power of 2.52 PFLOPS FP8 at its annual cloud computing event, indicating strong growth in AI infrastructure [1] - Credo's Q2 FY2026 revenue reached $268 million, a year-over-year increase of 272.1%, driven by growth in its core AEC and IC businesses [1] - DeepSeek introduced two AI models, achieving performance levels close to GPT-5, indicating advancements in domestic AI capabilities [1][47] - ByteDance's Doubao team launched the "Doubao Phone Assistant," with the first device selling out quickly, showcasing the potential for AI applications in consumer electronics [1] Summary by Sections Communication Sector - The communication sector shows a steady upward trend, with significant investments in cloud and IDC businesses compensating for pressures in traditional telecom services [14] Server Sector - The server index decreased by 1.22% this week, but AWS's announcement of a significant expansion in AI/HPC data centers suggests ongoing demand growth [2][7] Optical Modules - The optical module index increased by 4.67%, supported by Marvell's strong quarterly performance and strategic acquisition [2][7] IDC Sector - The IDC index rose by 0.50%, with DeepSeek's new AI models expected to drive demand in data centers [3][8] Core Data Updates - Telecom business revenue reached 1.467 trillion yuan in the first ten months of 2025, a year-over-year increase of 0.9% [4][15] - The export value of optical modules decreased by 27.6% year-over-year in October, attributed to domestic companies establishing overseas factories [31] Market Trends - The communication sector's performance this week ranked second among all industries, with notable gains in specific companies [39][42]

3) DeepSeek Sparse Attention (DSA)DeepSeek’s new V3.2 model introduces DeepSeek Sparse Attention (DSA), which brings complexity down from O(L²) to O(Lk), where k is fixed.How it works:A lightweight Lightning Indexer scores which tokens actually matter for each query.Small number of heads, runs in FP8, computationally cheap.Then a selection mechanism retrieves only the top-k key-value entries.The key insight is that only 2048 tokens get selected per query, regardless of context length.So the expensive attent ...

图片来自AI生成 【数智周报将整合本周最重要的企业级服务、云计算、大数据领域的前沿趋势、重磅政策及行研报 告。】观点华为任正非：大量建设大模型是正确的探索，未来算力一定过剩；AI重在"应用"，着眼于未 来3-5年 华为创始人兼总裁任正非与ICPC（国际大学生程序设计竞赛）主席、教练及获奖选手座谈会纪要公 布。在回答关于计算资源的问题时，任正非直言，"以后是算力过剩的，不是你想象的算力不足"。他认 为，建数千个、数百个大模型都是正确的探索。华为在解释超节点时提到了950、960、970等型号， 但"我们需要多少个'970'，970在哪儿能用？需要多少台，集群怎么连接"，任正非表示，这是一种线性 技术推演，是一定能实现的算力假设。而需求是不是一种线性结构尚未可知，"万一是非线性的呢"。因 此，他说："算力过剩的时代一定会到来的。" 谈及当下火热的AI，任正非的观点很明确：AI重在应用，不重在发明。"发明AI只是一家IT公司，应用 AI是会强大一个国家。"任正非表示，IT公司对人类的贡献只2%，AI在产业上的贡献则会占到98%。仅 以洗煤、炼钢为例，洗煤精度提高0.1%，乘以中国40亿吨的煤炭产量，价值惊人；高炉炼 ...

Core Insights - The report reveals that AI is undergoing a significant revolution, characterized by a shift from traditional models to reasoning models that can think and plan in multiple steps [3][11][12]. Group 1: OpenRouter and Its Importance - OpenRouter is likened to "Meituan" in the AI world, connecting over 500 million developers to more than 300 AI models, making its data highly credible [5][6]. - OpenRouter's daily token processing volume has surpassed 1 trillion, indicating a rapid growth from approximately 100 trillion tokens annually from early 2024 to mid-2025, marking a tenfold increase [8][6]. Group 2: Reasoning Revolution - The report identifies a "reasoning revolution," where AI models evolve from simple response machines to complex reasoning machines capable of multi-step thinking [11][12]. - The launch of OpenAI's o1 reasoning model (codename Strawberry) is a pivotal event, as it incorporates internal reasoning processes that enhance its problem-solving capabilities [18][19]. - Users are increasingly engaging in complex tasks, leading to longer prompts and more dialogue rounds, indicating a shift towards training AI for intricate tasks [20][21][23]. Group 3: Agentic AI - Agentic AI represents a transformation where AI can autonomously plan, execute, and verify tasks, moving from passive response to active engagement [27][30]. - The report highlights that agentic reasoning is the fastest-growing behavior on OpenRouter, indicating a shift in user expectations from simple answers to task completion [34][35]. Group 4: Rise of Open Source Models - Open source models, particularly from Chinese teams like DeepSeek R1 and Kimi K2, are rapidly gaining market share, challenging the dominance of closed-source models [44][47]. - DeepSeek R1 offers significant cost advantages, with a cost of $0.003 per 1K tokens compared to $0.03 for GPT-4, making it attractive for developers [52]. Group 5: Real-World AI Usage - The primary applications driving token usage are creative writing and programming, with AI becoming indispensable for developers [71][72]. - Users are not merely relying on AI for content generation but are engaging in co-creation, indicating a shift in the role of AI from a tool to a creative partner [77][78]. Group 6: Model Personality - Users' choices of AI models are influenced by the "personality" of the models, which affects user retention and engagement [88][95]. - The report suggests that models with unique personalities can outperform those with higher benchmark scores in terms of user loyalty [96][100]. Group 7: Implications for the Chinese AI Industry - The success of Chinese models like DeepSeek R1 and Kimi K2 in the global market indicates that they have competitive capabilities [109]. - The report emphasizes the importance of focusing on reasoning and agentic capabilities as key technological directions for the Chinese AI industry [115].

Group 1: Market Developments - Moole Technology, the first domestic GPU stock, saw its share price surge by 425.46% on its debut, closing at 600.5 CNY per share, with a market capitalization of 282.3 billion CNY, significantly exceeding its issue price of 114.28 CNY per share [1] - DeepSeek launched two new models, DeepSeek-V3.2 and DeepSeek-V3.2-Speciale, claiming global leadership in inference capabilities, with Speciale surpassing Google's Gemini3 Pro in several benchmarks [2] - ByteDance and ZTE announced the release of the "Doubao AI Phone," which features advanced AI capabilities, although initial user feedback indicated some operational issues [3] Group 2: Strategic Moves - OpenAI's CEO Sam Altman declared a "red alert" to prioritize the rapid improvement of ChatGPT, delaying other projects in response to competitive pressures from Google [4] - Baidu's Kunlun chip division is reportedly preparing for an IPO in Hong Kong, aiming to submit its application by Q1 2026 [5][6] - Lenovo introduced its "AI Factory" solution and upgraded its AI server offerings, emphasizing the need for enhanced computational power in AI applications [7] Group 3: Industry Trends - Nvidia's CFO indicated that major model manufacturers are seeking direct partnerships with Nvidia, moving away from reliance on cloud service providers [8] - UBS analysts noted that the likelihood of an AI bubble in China is low, attributing this to limited domestic financing and a cautious approach to capital expenditure [9] - Micron Technology announced its exit from the consumer storage business to focus on providing storage solutions for AI applications [13] Group 4: Technological Innovations - Amazon launched its custom AI chip, Trainium3, which reportedly offers four times the computational speed of its predecessor and can reduce AI model training costs by up to 50% compared to equivalent GPU systems [14] - Nvidia expanded its strategic partnership with Synopsys, investing approximately 2 billion USD to enhance virtual design and testing capabilities in various industries [10]

Group 1: Market Performance and Company Overview - Moer Technology, known as the "first domestic GPU stock," saw its share price increase by 425.46% on its first trading day, closing at 600.5 yuan per share, with a market capitalization of 282.3 billion yuan [2] - The initial public offering (IPO) price was 114.28 yuan per share, indicating a significant rise in value and a potential profit of 240,000 yuan for investors holding one lot [2] - The company focuses on the research, design, and sales of GPUs and related products, targeting AI, cloud and data centers, high-performance rendering, and video acceleration [2] Group 2: Competitive Landscape - Moer Technology's market valuation at the IPO was 53.715 billion yuan, with a projected 2024 diluted static price-to-sales ratio of 122.51 times, higher than the industry average of 111.23 times [2] - The domestic AI chip market, particularly for GPUs, faces intense competition, with Nvidia holding a dominant position globally [2] Group 3: AI Developments and Innovations - DeepSeek launched two new models, DeepSeek-V3.2 and DeepSeek-V3.2-Speciale, which reportedly outperform Google's Gemini3 Pro in inference capabilities [3] - Lenovo introduced the "Lenovo AI Factory" solution and upgraded its heterogeneous computing platform, indicating a shift towards deeper integration of AI in industry applications [8] - Nvidia's CFO highlighted a shift in large model vendors seeking direct collaboration with Nvidia, moving away from reliance on cloud service providers [9] Group 4: Industry Trends and Future Outlook - UBS analysts noted that the likelihood of an AI bubble in China is low, attributing this to limited domestic financing and cautious capital expenditure [10] - Micron Technology announced its exit from the consumer storage business to focus on providing storage products for AI applications, reflecting a strategic pivot towards higher-growth segments [14] - Amazon launched its custom AI chip, Trainium3, which reportedly offers four times the computational speed of its predecessor and can reduce costs by up to 50% compared to equivalent GPU systems [15]

Test-time Scaling 有哪些非共识？流行的 Sequential 和 Parallel 路线有何局限？Test-time Scaling 为何需要「Better Search」？「温度」如何影响 Scaling 效果？Test-time Scaling 有哪些 「Where」需要改进？... 机器之心PRO · 会员通讯 Week 49 --- 本周为您解读 ③ 个值得细品的 AI & Robotics 业内要事 --- 1. 多非共识，Test-time Scaling 能否一直大力出奇迹？ 2. Skills vs MCP，谁才是 「大模型的 HTTP 时刻」？ 一年过去，社区对于 MCP 的定位仍有争议？平均 25 个用户对应 1 个开发者，MCP 目前更多是开发者自娱自乐的产物？「人如其名」，Skills 真是来 kill MCP 的？MCP 能做但 Skills 不能做 的，现在也没什么用？MCP 大规模落地还得看下一个「微信小程序」入口的出现？... 3. 从否定单模 AGI 到回应开源冲击，OpenAI 如何打造「最强平台」？ 曾被视为真理的「单模 AGI」为何在商业现实面前彻底梦 ...

Core Viewpoint - Huang Renxun, CEO of Nvidia, emphasizes that while the U.S. leads in advanced AI models, China's manufacturing strength and open-source contributions position it favorably in the AI competition [1][3][4]. Group 1: AI Competition and Industry Development - Huang Renxun states that China's energy production is double that of the U.S., which significantly impacts industrial development [1]. - He highlights that the U.S. has experienced hollowing out of its manufacturing sector, which is crucial for supporting chip factories and AI data centers [3]. - The majority of the 1.4 million AI models globally are open-source, with China excelling in this area, which is vital for the growth of startups and academic research [3][4]. Group 2: Open Source and Technological Application - Huang uses examples like Linux and PyTorch to illustrate the importance of open-source projects in driving technological advancement [3]. - He notes that the speed of technology application often depends on societal attitudes, suggesting that those who can quickly implement technology will gain a competitive edge [3]. Group 3: Semiconductor Industry Comparison - The compound annual growth rate of the Western semiconductor industry is typically between 20%-30%, while China's semiconductor industry is growing rapidly, indicating its potential to catch up [4]. - Huang points out that nine of the top ten engineering universities are in China, and half of the world's AI talent is Chinese, with 70% of AI patents originating from China [4]. - He warns that if the U.S. does not take action, it may transition from being a technology seller to a buyer in the future [4].

Core Insights - Leading AI developers are increasingly focused on creating models that mimic human reasoning, but these models are significantly more energy-intensive, raising concerns about their impact on power grids [1][4]. Energy Consumption - AI reasoning models consume, on average, 30 times more power to respond to 1,000 prompts compared to alternatives without reasoning capabilities [2]. - A study evaluated 40 open AI models, revealing significant disparities in energy consumption; for instance, DeepSeek's R1 model used 50 watt hours with reasoning off and 7,626 watt hours with reasoning on [3][6]. - Microsoft's Phi 4 reasoning model consumed 9,462 watt hours with reasoning enabled, compared to 18 watt hours with it disabled [8]. Industry Concerns - The rising energy demands of AI have led to scrutiny, with concerns about the strain on power grids and increased energy costs for consumers; wholesale electricity prices near data centers have surged by up to 267% over the past five years [4]. - Tech companies are expanding data centers to support AI, which may complicate their long-term climate objectives [4]. Model Efficiency - The report emphasizes the need for understanding the evolving energy requirements of AI and suggests that not all queries necessitate the use of the most energy-intensive reasoning models [7]. - Google reported that its Gemini AI service's median text prompt used only 0.24 watt-hours, indicating a lower energy consumption than many public estimates [9]. Industry Leadership Perspectives - Tech leaders, including Microsoft CEO Satya Nadella, have acknowledged the need to address AI's energy consumption, emphasizing the importance of using AI for societal benefits and economic growth [10].