中信证券发布研报称，2025年互联网板块在外部扰动反复的环境下走出震荡上行行情，上行动量主要为 AI驱动的价值重估及流动性改善。展望2026年，AI依然是板块估值上行的核心催化，核心关注：1)模型 迭代、应用落地以及业绩兑现带来的叙事强化；2)模型公司上市后带来的估值对标，有望进一步驱动互 联网公司AI业务价值重估。而对于短期AI相关性较低的公司，业绩能见度和估值性价比将是选股核心 要素，建议关注竞争格局相对稳定、业绩韧性较强的垂类龙头、以供给创新带动的高景气赛道以及低估 值下的绝对收益标的。 中信证券主要观点如下： 2025年回顾：AI及流动性宽松带动下的估值抬升。 基础设施：海外算力供给约束缓解及国产算力能力提升后，CAPEX将继续扩张。根据各公司公告以及 彭博一致预期，2025年阿里巴巴/腾讯的资本支出有望达1300/829亿元，占收比分别为12.7%/11.0%(远低 于微软/亚马逊/谷歌/Meta的32.7%/17.5%/25.3%/35.2%)，主要受制于海外高端算力的可得性以及国产芯 片的产能约束。 展望2026年，该行认为互联网公司CAPEX投入将进一步扩张，主要原因系海外算力供给约束的缓解、 ...

近日，在格隆汇举办的"科技赋能·资本破局"分享会上，国联民生研究所副总经理兼海外研究首席分析师孔蓉围绕全球AI发展趋势、关键技术演进与市场 机会等进行了深度分享。 她指出，尽管市场近期对AI是否存在泡沫、资本开支是否可持续等问题存在疑虑与分歧，但基于对海外科技前沿与中国AI生态的持续观察，其对2026年 及之后的AI发展方向保持乐观。 01、明年AI竞争更卷！ AI到底有没有泡沫，明年机会如何？这是过去两个月市场探讨最多的问题。 美股财报季带来的股价回调，让市场对2025—2027年持续加大的资本开支产生疑虑。 孔蓉指出，基于对海外科技前沿与中国AI生态的持续观察，依然看好26年以及后续的AI方向的机会。 今年来，以谷歌Gemini系列为代表的多模态模型取得突破，市场模型持续迭代为市场注入强劲信心。 其中，谷歌凭借其全栈自研能力、长期技术积累与雄厚的资本资源，在长跑中后劲充足。 Meta虽然在2025年经历组织架构与人员调整，市场信心不足，但在资源整合与顶尖AI人才引入后，期待在2026年推出具有竞争力的模型，成为值得关注 的重点。 微软在维持与OpenAI合作的同时，已开始布局自有模型，关注微软后期发力大模 ...

Core Viewpoint - The article discusses the optimistic outlook for AI development beyond 2026, despite current market concerns about potential bubbles and sustainability of capital expenditures [2][6]. Group 1: AI Market Trends - There is ongoing debate in the market regarding whether AI is in a bubble and the sustainability of capital expenditures for 2025-2027 [3][4]. - Major tech companies are expected to shift focus from "infrastructure" to "application realization," with key observations on revenue growth from Google Cloud Platform (GCP), Microsoft Azure, and Amazon AWS [11]. - The release pace of large models is anticipated to accelerate, with major players like OpenAI, xAI, Meta, Microsoft, and Google continuing to launch new models, intensifying industry competition [12][28]. Group 2: Key Players and Innovations - Google has demonstrated strong capabilities with its self-developed technology and resources, maintaining a competitive edge [8]. - Meta is expected to regain market confidence by 2026 after restructuring and integrating top AI talent, aiming to launch competitive models [8]. - Microsoft is focusing on its own models while maintaining collaboration with OpenAI, looking for synergies between its large models and ecosystem [9]. - xAI, despite being a latecomer, is rapidly iterating its models and is considered a significant variable in the market [10]. Group 3: Model Capabilities and Applications - The enhancement of multi-modal capabilities is crucial for transforming content production in advertising and e-commerce, as well as improving user experiences with hardware like AR/VR devices [15][18]. - Breakthroughs in memory and personalization capabilities will allow AI to evolve from general tools to personalized assistants, increasing user engagement and driving token consumption [23][24]. - The overall improvement in model capabilities is fundamental for the commercialization of AI, leading to clearer paths for investment returns [25][26]. Group 4: China's AI Ecosystem - China's AI ecosystem is recognized for its strong competitive advantages, with domestic models gaining international acknowledgment [40]. - Major Chinese tech firms like Alibaba and Tencent are committed to ongoing investments in AI, indicating a long-term strategy [40]. - The country boasts the largest pool of engineers and a rapid product iteration culture, which is expected to replicate the "application innovation" seen in the mobile internet era, creating numerous investment opportunities [40][41]. - Current valuations of Chinese AI companies are considered reasonable compared to their U.S. counterparts, providing a favorable investment margin [41].

具身智能是创业者友好赛道，投资人不想错过。 作者丨Lucas 编辑丨刘恒涛 图源丨Midjourney "你很难想象，大厂对AI人才有多渴望。"英诺基金创始人李竹感慨，"我们之前投过一个计算机博士 的AI Agent项目，但公司成立仅两个月，就被某大厂收购了——他们买下一家公司，实际上就是想要 这个人。" 一边是大厂不惜重金挖具身智能的人才，另一边，是投资人对具身智能人才的贴身紧盯，很多技术精 英纷纷出走大厂，投身具身智能创业。据不完全统计，2023年至今，已有超30位创业者从大厂离职 创业具身智能。 随着AI和具身智能行业的深入发展，这已经成为值得关注的现象：头部大厂不惜重金，抢夺顶尖AI信 息人才；同时，又不断有人从大厂辞职，创办具身智能企业。 这个现象的背后，折射了大厂对这两个领域战略上的态度：对AI，志在必得；对具身智能，观望为 主。 对大厂来说，AI竞争是"生死之战"，具身智能眼前却未必是好生意。所以，在很多投资人眼里，具身 智能是对创业者友好的赛道，相比AI大模型，投资人对具身智能赛道更为青睐。 AI人才争夺战 清华博士年薪几百万 大厂对AI的投入，从招聘和薪资上可以窥见一斑。 "清华计算机系的博 ...

2025 年 AI 大模型行业迎来结构性变革，竞争从单纯的能力竞赛转向可持续性比拼，技术范式、市场格局、应用形态与全球治理四大维度的深刻转变，共同 重塑行业发展轨迹。 技术层面实现多重突破性演进。训练范式从依赖主观反馈的 RLHF 全面转向客观可验证的 RLVR，模型通过自我检验实现推理能力飞跃，成为年度最关键技 术拐点。混合专家（MoE）架构强势回潮，以稀疏激活模式平衡参数规模与计算成本，实现性价比极致追求。多智能体自我博弈与合成数据微调成为常态， 模型摆脱对人类标注的依赖，同时检索增强生成（RAG）成为企业级应用标配，有效解决幻觉与知识时效性问题。此外，模型呈现 "锯齿化" 能力结构，在 数学、编程等形式化智力领域突飞猛进，却在常识推理上仍存短板。 新王登基: 京 谷歌Gemini 3全面 国模型以惊人的成 市场格局呈现集中化与民主化双重张力。谷歌 Gemini 3 凭借自研 TPU v5 芯片与多模态优势，终结 OpenAI 长期领先地位，中国模型以成本效益实现弯道超 车。市场向头部集中，Anthropic 等顶尖初创企业获巨额融资，二三线玩家面临出清，但开源浪潮形成制衡，阿里通义千问、01.ai Yi ...

Group 1 - The core viewpoint emphasizes the focus on the media industry's trend towards prosperity, driven by a strong new product cycle and AI applications [1] - The gaming sector is expected to see performance and valuation upgrades due to a robust new product cycle, with a total of 178 domestic games and 6 imported games approved in November, marking a three-year high [2] - The advertising spending is anticipated to grow as the economy recovers, benefiting the media sector [1] Group 2 - The gaming market revenue in October reached 31.4 billion yuan, a year-on-year increase of 7.8%, primarily driven by strong product cycles from leading companies [2] - The variety show market is performing well, with Tencent's "Now Just Depart Season 3" achieving a market share of 24.96% [2] - The short drama and comic drama segments are gaining popularity, with notable titles leading in viewership [2] Group 3 - AI applications are advancing rapidly, with significant breakthroughs in AI video models, including the public release of the O1 video model capable of generating 3-10 second videos [3] - The release of PixVerse V5.5 marks a milestone as the first domestic AI video model supporting multi-camera and audio-visual synchronization [3] - The rapid development of AI technologies is expected to create new opportunities across various sectors, including advertising and gaming [1][3]

据悉，阿里千问为开源模型，所有人都可以自主下载，无需授权。根据官方披露，截至目前，千问模型 的全球累计下载量已突破7亿次。 Meta引入阿里千问模型，曾是开源霸主 今年以来，由于Llama 4发布后表现拉胯，Meta与OpenAI、谷歌等竞争对手的差距逐渐拉大，在此背景 下，牛油果项目应运而生。 作为Meta下一代旗舰级AI大模型，"牛油果"大模型被视为Meta在AI军备竞赛中的"救命稻草"，目标性能 直指GPT-5，计划于2026年第一季度发布。 当时就有海外网友恶搞做了张梗图：牛油果的芯是鲸鱼（DeepSeek）。 只不过最终牛油果切开不是DeepSeek，而是千问。 12月10日，据报道，Meta在新一代大模型"牛油果"的研发中，引入了阿里巴巴通义千问模型，来对新模 型进行微调优化。 科技每日推送从阿里云独家获悉，Meta事先没有找阿里索要授权，阿里昨晚也是刚知道。 要知道前两年，Meta的Llama模型在全球开源界是绝对的霸主地位，当时国内很多大模型都会被质疑是 套壳Llama，谁能想到几年过去，中国开源模型崛起，Meta反而成了真·套壳的那个。 并且，Meta彻底违背初心，抛弃开源路线。 牛油果大模 ...

Core Insights - Meta's strategic shift from open-source to closed-source AI models, highlighted by the $14.3 billion acquisition and the development of the Avocado model, reflects the pressures of commercial realities and industry competition [2][3] - The transition signifies a critical moment in the commercialization challenges of open-source AI, as Meta's previous open-source efforts yielded over 30 million downloads but generated less than $1 billion in licensing revenue against over $70 billion in annual AI investments [2][3] - The closed-source model is seen as essential for capturing high-value markets, particularly in sectors like finance and healthcare, where data security and compliance are paramount [2][3] Meta's Strategic Shift - Meta's decision to adopt a "technology fusion" approach by integrating technologies from Google, OpenAI, and Alibaba aims to quickly address its shortcomings and meet industry demands [3] - The internal upheaval, including the departure of key personnel and layoffs at the FAIR lab, raises concerns about the compatibility of different model architectures and potential intellectual property disputes [3][4] - This shift marks the beginning of a new phase in the AI industry characterized by a coexistence of open-source and closed-source models, with open-source models still dominating academic research and smaller applications [3][4] Market Implications - Meta's transition to closed-source is expected to accelerate market consolidation, with leading companies likely to build commercial moats through closed-source models, while smaller players may find new opportunities in the open-source space [4] - The integration of Chinese models like Alibaba's Tongyi Qianwen into Meta's technology references indicates the growing global competitiveness of Chinese AI technologies [4] - The release of Avocado in Q1 2026 will be a pivotal moment, with the potential to replicate the success of the Microsoft-OpenAI partnership, creating a "model-hardware-advertising" business loop [4][5] Timeline of Key Events - February 2023: Launch of Llama 1, marking Meta's entry into large models with an open-source approach [5] - July 2023: Llama 2 becomes the most popular open-source large model with over 30 million downloads [5] - June 2025: Meta acquires a stake in Scale AI for $14.3 billion and appoints Alexandr Wang as Chief AI Officer, signaling a shift to closed-source [5][6] - October 2025: Announcement of a $27 billion Hyperion data center plan to support closed-source model capabilities [7] - Q1 2026: Expected launch of Avocado, focusing on complex reasoning and long video analysis, aiming to compete with GPT-5 and Gemini 3 Ultra [9] Strategic Differences in AI Models - U.S. giants primarily focus on closed-source models with clear commercial pathways, while Chinese players adopt a dual approach of open-source and closed-source to balance ecosystem development and monetization [11] - The U.S. strategy emphasizes closed-source to maintain competitive advantages, whereas China's approach leverages open-source to address specific industry needs and accelerate deployment [11][12] - The iteration pace differs, with U.S. companies releasing new versions semi-annually or annually, while Chinese firms adopt a more rapid release cycle driven by community engagement [12][13]

Economic News - In November 2025, the national industrial producer price index decreased by 2.2% year-on-year, while the purchasing price index fell by 2.5% year-on-year. The average for January to November shows a decline of 2.7% for the producer price index and 3.1% for the purchasing price index compared to the same period last year [1] - The consumer price index for November 2025 increased by 0.7% year-on-year, with an average for January to November remaining flat compared to the previous year [1] - The United Nations reported that global trade is expected to exceed $35 trillion for the first time in 2025, with a projected growth of approximately 7% compared to last year [1] - The International Monetary Fund (IMF) forecasts China's economic growth rate to reach 5% in 2025, an increase of 0.2 percentage points from previous estimates [1] - The Ministry of Industry and Information Technology has revised the management measures for public service platforms in industrial technology, focusing on key industries such as new energy, artificial intelligence, and biotechnology [1] Company News - Haiguang Information has terminated its plan to merge with Zhongke Shuguang through a share exchange [4] - China Life Insurance reported total premiums exceeding 700 billion yuan as of November 30, 2025 [4] - SIRUI announced the termination of its plan to acquire shares of Aola and will resume trading on December 10, 2025 [4] - Alibaba's Tongyi Qianwen has launched Qwen Code v0.3.0, enhancing its capabilities for developers [4] - Xiamen Airport is planning to acquire 100% equity of its controlling shareholder's subsidiary, Zhaoxiang Technology, using its own funds [4] Non-Bank Financial Sector - The recent adjustment of risk factors for insurance companies aims to enhance their long-term investment management capabilities and support the real economy [5][6] - Specific adjustments include lowering risk factors for certain stock indices and export credit insurance, which is expected to stabilize the capital market and improve the efficiency of insurance fund utilization [7][8] - The regulatory changes are anticipated to benefit major industry players and promote a balanced and healthy development of the equity market [9][10]

Core Viewpoint - The nubia M153 phone, featuring the "Doubao AI Assistant," has gained significant attention, with its technical preview version selling out at a price of 3499 yuan, indicating strong market interest in AI-integrated mobile devices [4][6]. Group 1: Product Overview - The nubia M153 is marketed primarily to industry professionals for experiencing the Doubao AI Assistant, with a caution that it is not a fully mature product for general consumers [6]. - The phone is an engineering sample developed in collaboration with Doubao, which leads the product definition and experience, while nubia focuses on hardware engineering and technology development [8]. Group 2: Market Dynamics - On second-hand platforms, the nubia M153 is being resold for prices ranging from 3999 to 4599 yuan, with F codes being sold for around 100 yuan [8]. - Doubao is in discussions with multiple smartphone manufacturers for potential collaborations, but there are currently no confirmed partnerships [9]. Group 3: Competitive Landscape - Major smartphone manufacturers like Huawei, Xiaomi, and OPPO have already developed their own AI assistants, indicating a competitive landscape in the AI mobile assistant space [9]. - The integration of AI into mobile operating systems is being explored by various manufacturers, with a focus on creating platform-level AI capabilities [9]. Group 4: Challenges and Considerations - The successful implementation of mobile AI assistants requires overcoming ecological challenges, including obtaining permissions from major internet application providers like Alibaba and Tencent [11]. - The integration of AI and mobile technology necessitates modifications to operating systems, which could conflict with the core interests of smartphone manufacturers [12].