编译 | Tina Gemini 在数据上第一次"反超"了 ChatGPT。 根据《金融时报》披露的最新统计，截至 2025 年底，用户在桌面端和移动网页端单次使用 Gemini 的平均停留时长已经达到约 7.2 分钟，首次超过 ChatGPT 的约 6 分钟，也略高于 Anthropic Claude 大约 6 分钟的水平。 更长的停留时间，意味着这已经不是"下个 App 玩两下"的新鲜感，而是用户真的愿意在 Gemini 里待 更久、反复用它解决问题。 与此同时，在 App 下载量 上，虽然 ChatGPT 依然以约 8700 万的月度下载量领先，但 Gemini 的追 赶速度非常惊人——从 2025 年年中每月约 1500 万下载，一路飙升到年底的约 6600 万。 这说明，谷歌把 Gemini 深度嵌入自家生态——从 Android 设备到 Google Workspace——的策略， 正在用户获取上逐步见效。 停留时长尤其值得关注，因为它反映的是真实使用，而不是"下载即走"。用户主动选择待得更久，说 明 Gemini 的回答、功能或整体体验确实发生了可感知的提升。 对谷歌来说，这次"逆袭"格外解气： ...

Core Insights - Gemini has surpassed ChatGPT in average user session duration, reaching approximately 7.2 minutes compared to ChatGPT's 6 minutes, indicating a deeper user engagement with the platform [1][8] - Despite ChatGPT leading in monthly downloads at around 87 million, Gemini has shown remarkable growth, increasing from approximately 15 million monthly downloads in mid-2025 to about 66 million by the end of the year [4] - Google's strategy of deeply integrating Gemini into its ecosystem, from Android devices to Google Workspace, is beginning to yield positive results in user acquisition [5] User Engagement - The increase in session duration reflects genuine user interest and satisfaction with Gemini's responses and overall experience, marking a significant improvement from its earlier iterations [8] - The launch of Gemini 3 coincides with this uptick in user engagement, suggesting that the new model has contributed to the enhanced user experience [8][9] Competitive Landscape - Gemini 3 has outperformed OpenAI's current advanced models in various benchmark tests, leveraging Google's extensive computational resources and infrastructure [9][10] - While OpenAI maintains an advantage in algorithmic capabilities, Google's strength lies in its comprehensive "full-stack" approach, integrating models, TPU, data centers, and infrastructure [10] Organizational Culture - The resurgence of early Google culture, characterized by high-density talent collaboration and active idea exchange, is seen as a critical factor in Gemini's competitive edge [11][12] - The integration of teams and resources, particularly between Google Brain and Google DeepMind, has facilitated a more agile and innovative environment [14][32] Future Outlook - The company is focused on long-term investments in AI and infrastructure, with plans for ambitious projects like building data centers in space, reflecting a commitment to future technological advancements [32][33] - The rise of "Vibe Coding" is democratizing software creation, allowing more individuals to engage in programming and creative tasks, which could lead to significant economic value [35][37]

Core Insights - Google has made a significant comeback in the AI sector with the release of its new model, Gemini 3, which outperforms larger models and reestablishes its dominance in AI-generated imagery, causing concern for competitors like OpenAI [1][2] - The narrative surrounding Google has shifted from being perceived as outdated and bureaucratic to being recognized as a "waking giant" that is potentially redefining the industry's technological direction [2][3] - The transformation of Google is attributed to its consistent investment in an "AI-first" strategy since 2016, which includes developing its own AI chips (TPU) and training large models, despite setbacks like the Bard failure [5][7] Group 1: Google's AI Strategy and Developments - The release of Gemini 3 and its variants, such as Nano Banana Pro, showcases Google's advancements in AI capabilities and its strategic focus on a unified architecture for its models [1][12] - The merger of Google Brain and DeepMind into a single team has streamlined its AI research efforts, allowing for a more cohesive approach to product development and innovation [8][10] - Google's extensive resources, including its global search infrastructure and vast amounts of training data from platforms like YouTube and Google Photos, have positioned it uniquely in the AI landscape [7][15] Group 2: Competitive Landscape and Implications - The competitive landscape is evolving, with companies like Alibaba and Baidu also making strides in AI, but Google's comprehensive ecosystem and historical investments provide it with a significant advantage [16][22] - The success of Google's AI applications, such as NotebookLM and Nano Banana, indicates a shift towards native AI solutions that enhance user experience and knowledge management [15][21] - The ongoing competition in the AI sector highlights that success will depend not just on speed but on the ability to integrate models, computing power, and applications into a cohesive system [22]

Investment Rating - Maintain "Buy" rating for Alibaba (BABA) [4][15] Core Insights - Alibaba's revenue for FY2Q26 was RMB 247.8 billion, representing a 5% year-over-year growth, with a like-for-like growth of 15% when excluding disposed businesses [9] - The company is shifting its strategic focus from platform economy to a broader consumption ecosystem, enhancing traffic synergies through full-site promotions and instant retail [10] - Cloud business revenue grew by 34% year-over-year, with AI-related product revenue experiencing triple-digit growth for nine consecutive quarters [11] - International Digital Commerce group turned profitable with a revenue increase of 10% year-over-year, achieving an adjusted EBITA of RMB 162 million [13] - The company actively repurchased shares, spending USD 253 million to buy back 17 million common shares [14] Financial Data and Profit Forecast - Revenue projections for Alibaba are as follows: - FY24: RMB 941,168 million - FY25: RMB 996,347 million - FY26E: RMB 1,038,609 million - FY27E: RMB 1,143,436 million - FY28E: RMB 1,250,635 million - Non-GAAP net profit forecasts: - FY26E: RMB 101,943 million - FY27E: RMB 145,452 million - FY28E: RMB 183,640 million [5][18]

Core Insights - Baidu's third-quarter financial report reveals a significant growth in AI business revenue, marking a transition from investment to commercialization phase [1][3] - The AI revenue is approaching 10 billion, with notable contributions from AI cloud, AI applications, and AI native online marketing services [1][4] AI Revenue Breakdown - AI cloud revenue grew by 33% year-on-year, with high-performance computing subscription revenue increasing by 128% [4] - AI applications generated 2.6 billion, while AI native marketing services revenue surged by 262% to 2.8 billion [4] - Baidu's AI cloud holds a 24.6% market share, maintaining the top position in China's AI public cloud market for six consecutive years [4] Strategic Developments - Baidu's AI strategy is supported by a comprehensive ecosystem and deep technological expertise, positioning the company for accelerated commercial value release [3][9] - The company has launched several AI applications, including GenFlow 3.0 and Baidu Famo, which have attracted significant user engagement [5][9] Market Position and Competitiveness - Baidu is recognized as one of the few companies providing a full-stack AI service, which enhances its ability to deliver efficient and cost-effective solutions [7][9] - The company has made significant advancements in AI chip technology, with its Kunlun chip being selected for major projects, indicating strong market validation [7][8] Stock Performance and Valuation - Baidu's stock has seen substantial increases, with Hong Kong shares rising by 37% and U.S. shares by 35% since the beginning of 2025 [10] - Analysts believe that Baidu's valuation has room for growth compared to global peers like Google, especially in the rapidly expanding Robotaxi market [10][11] Future Outlook - The potential for Baidu's AI digital employees and intelligent agents to reshape traditional advertising models is significant, with new revenue streams emerging [11][12] - Investment firms are optimistic about Baidu's long-term growth potential, raising target prices and maintaining "buy" ratings [14][16]

10月底，谷歌发布了一份史上最强财报，也是唯一一家财报发布后股价飙涨的美国大厂。回答分析师提问时，谷歌CEO桑达尔·皮查伊特别强调，谷歌业绩 增长的驱动力在于，他们提供了全栈式AI服务，这也是谷歌区别于其他美国大厂AI能力的关键因素。 和谷歌一样拥有全栈式AI能力的公司，在国内也有一家，就是百度。11月13日百度世界2025大会上，百度展示了多个AI应用产品，无论是百度内部原生的 AI应用文库、秒哒，还是服务外部企业的慧播星数字人、百度伐谋等，都有令人惊艳的表现。 百度这些AI产品的成绩都建立在全栈式AI能力基础上。有了全栈式AI，可以整合从底层算力到顶层应用的各个环节，为企业提供一体化的AI解决方案，它 能显著降低技术门槛，并提升AI应用的效率和可靠性。 同在百度世界2025大会上，百度开放了慧播星数字人、伐谋等面向外部企业的AI能力，把这些能力开放给全球各行各业。"今天任何一家企业，只有用好AI 才能立于不败，任何一个个体只有用好AI才能掌控未来。"百度创始人李彦宏指出，当AI能力被内化为一种原生能力时，智能就不再是成本，而是生产力。 在这个维度，所有企业的业务都值得用AI再做一遍。 全栈AI能力加持 得 ...

Core Viewpoint - Alibaba is evolving into a full-stack AI technology infrastructure platform, positioning itself alongside Google as one of the two leading global AI companies capable of integrating hardware and applications into a complete ecosystem [2][21]. Group 1: Capital Investment and Financial Performance - Alibaba reported a record capital expenditure (Capex) of 38.6 billion yuan for Q1 of fiscal year 2026, with over 100 billion yuan invested in AI infrastructure and product development over the past four quarters [3][16]. - The revenue from Alibaba Cloud grew by 26%, and AI-related product revenue has seen triple-digit year-on-year growth for eight consecutive quarters, with AI revenue now accounting for over 20% of external commercial revenue [3][4]. Group 2: AI Model Development and Ecosystem - Alibaba has rapidly released and open-sourced several significant AI models, including Qwen3, which has become the world's strongest open-source inference model, and Qwen-Image, which topped the Hugging Face model rankings [3][4]. - The number of derivative models from Tongyi Qianwen has surpassed 140,000, making it the largest AI open-source model globally, with over 400 million downloads [4][16]. Group 3: Strategic Alignment with Google - Both Alibaba and Google are pursuing a capital-intensive, full-stack AI strategy, indicating a shift in the global AI competition paradigm from algorithmic races to comprehensive system integration and ecosystem control [11][21]. - Google has significantly increased its capital expenditure guidance for fiscal year 2025 from $75 billion to $85 billion to enhance its AI capabilities, reflecting a shift from its traditionally cautious investment approach [6][11]. Group 4: Infrastructure and Model Centralization - Both companies view controlling physical infrastructure as foundational to their full-stack strategy, with Alibaba planning to invest 380 billion yuan over the next three years to build a globally competitive cloud computing network [16][17]. - Alibaba's Tongyi Qianwen series serves as the core of its developer ecosystem, while Google's Gemini has attracted over 9 million developers, showcasing their respective strategies in building robust AI ecosystems [17][18]. Group 5: Application Layer and Market Impact - Alibaba leverages its extensive consumer applications to test and distribute AI technologies, enhancing user engagement and operational efficiency across platforms like Taobao and DingTalk [19][20]. - Google's AI features have significantly increased user engagement in its search and cloud services, demonstrating the effectiveness of integrating AI into existing applications [20][21]. Conclusion - The competition in the AI landscape is shifting towards building comprehensive ecosystems rather than merely focusing on individual algorithms or products, with Alibaba positioned advantageously due to its integrated business model and full-stack AI strategy [21][22].

Core Insights - Alibaba and Google are both pursuing a full-stack AI strategy, integrating hardware, models, and applications to create a comprehensive ecosystem in the competitive global AI landscape of 2025 [1][4][15] - Alibaba reported a record capital expenditure of 38.6 billion yuan for Q1 of fiscal year 2026, with over 100 billion yuan invested in AI infrastructure and product development over the past four quarters [1][15] - Google's capital expenditure guidance for fiscal year 2025 has been raised from $75 billion to $85 billion, reflecting a significant commitment to AI development [3][7] Group 1: Investment and Financial Performance - Alibaba's Q1 AI and cloud performance showed a revenue growth acceleration of 26%, with AI-related product revenue achieving triple-digit year-on-year growth for eight consecutive quarters [1] - The proportion of AI revenue in Alibaba's external commercialization revenue has exceeded 20% for the first time [1] - Google's token call volume doubled from 480 trillion to 980 trillion in one month due to enhanced model capabilities and low-cost applications [3] Group 2: AI Model Development - Alibaba has released and open-sourced several significant models, including Qwen3, which is recognized as the strongest open-source inference model globally [1][2] - The number of derivative models from Alibaba's Tongyi Qianwen has surpassed 140,000, making it the largest AI open-source model globally with over 400 million downloads [2][9] - Google's Gemini model has attracted over 100 million developers, showcasing its integration into existing workflows and achieving a processing capacity of over 980 trillion tokens monthly [10] Group 3: Full-Stack AI Strategy - The full-stack AI strategy is characterized by three layers: hardware/infrastructure, foundational models, and application/eco-system [5][6] - Both companies are investing heavily in their respective hardware and infrastructure, with Alibaba planning to invest 380 billion yuan over the next three years to build a competitive global cloud computing network [8] - The application layer is crucial, as both companies leverage their dominant consumer applications to test, distribute, and gather data for AI technologies [12][14] Group 4: Competitive Landscape - The shift from algorithm competition to a comprehensive battle over integrated systems, capital strength, and ecosystem control is evident in the AI industry [4][15] - The ultimate value of the full-stack model lies in its application layer, where both companies create a feedback loop that enhances user experience and generates valuable data for model training [12][13] - The competition will not be determined by the smartest single model but by the ability to build a robust, synergistic AI-driven business ecosystem [15]