行 业 及 产 业 相关研究 券 研 究 报 告 林起贤 A0230519060002 linqx@swsresearch.com 袁伟嘉 A0230519080013 yuanwj@swsresearch.com 任梦妮 A0230521100005 renmn@swsresearch.com 夏嘉励 A0230522090001 xiajl@swsresearch.com 赵航 A0230522100002 zhaohang@swsresearch.com 黄俊儒 A0230525070008 huangjr@swsresearch.com 研究支持 张淇元 A0230124080001 zhangqy@swsresearch.com 联系人 林起贤 A0230519060002 linqx@swsresearch.com 2025 年 12 月 28 日 字节 AI 春晚合作，游戏潮玩视频消 费旺季来临 看好 ——互联网传媒周报 20251222-20251226 本期投资提示： 行 业 研 究 / 行 业 点 评 请务必仔细阅读正文之后的各项信息披露与声明 本研究报告仅通过邮件提供给 中庚基金 使用。 ...

天风证券研报指出，AI应用商业化2025年将快速落地，成本结构有望向研发支出倾斜，软件公司有望 释放业绩。在大模型成熟赋能下，已形成 AI 大模型-AI 应用-商业模式落地的闭环，多模态大模型打开 了 AI 应用的"能力圈"，AI 应用空间进一步扩大。建议关注：企业级AI软件；内容创作工具；营销与客 户服务平台；AI有望赋能教育、金融和医疗等垂直场景实现客户规模增长、付费渗透率提升。 （文章来源：每日经济新闻） 消息方面，近日，南京市科学技术局正式公示《2025年度市级重大科技专项拟立项项目名单》。江苏润 和软件股份有限公司申报的"面向GUI系统的AI测试智能体平台关键技术研究"项目成功入选，标志着公 司在人工智能与软件测试融合创新领域的技术布局与创新实践获得了政府层面的认可与支持。 在AI产业链中，软件行业主要处于中游技术层和下游应用层，扮演着核心技术支撑和应用落地的关键 角色。具体来看，软件行业在中游技术层主要提供AI框架、开发平台和算法模型，这些技术是AI应用 开发的基础。在下游应用层，软件行业通过将AI技术与各行业结合，推动AI应用的落地。 12月26日盘中，A股三大指数集体上涨，盘面上，海南、光伏设 ...

近日，首块L3级自动驾驶专用正式号牌"渝AD0001Z"在重庆诞生，由重庆市公安局交通管理总队正 式授予长安汽车，标志着长安汽车率先开启L3级自动驾驶时代。 二、又一家大模型独角兽叩响港股大门，AI应用商业化进程加速 紧随智谱之后，又一家大模型独角兽叩响港股大门。12月21日晚，MiniMax通过港交所聆讯并披露 招股书，正式冲刺IPO。 点击付费阅读，挖掘行业机会，把握投资风口！ 前言 投资信息如迷雾？市场节奏总踏空？那是你缺少利器！想精准捕捉研报价值？想抢占投资先机？《行研 精选》为你破局！覆盖每个交易日，从事件中发现机会，提炼研报精华，助力捕捉行业风口。 【 精选快读 】 一、首块L3级自动驾驶牌照诞生，产业链迎加速时刻 ...

恒生科技指数由港股上市公司中与科技主题高度相关的、市值最大的30只股票组成，聚焦半导体、机器 人、软件、互联网等科技板块，覆盖AI产业链各环节，权重股包括美团、腾讯控股、阿里巴巴、中芯 国际等龙头企业。从估值来看，指数当前滚动市盈率为22.7倍，位于2020年发布以来30%分位以下，中 长期配置价值显现。 消息面上，AI应用商业化加速发展。QuestMobile数据显示，10月AI相关内容用户渗透率同比增长9.9个 百分点，且深度兴趣用户呈年轻化特征，抖音、微博、快手平台AI插件月活跃用户规模均已超三千万 级。 展望后市，中泰证券指出，美联储降息符合预期，港股短期流动性风险有所下降，而国内重要会议的积 极信号带动市场风险偏好回升。科技板块在流动性风险下行趋势下有望保持相对韧性，政策面强调创新 驱动，为科技成长板块注入强心剂，与"新质生产力"相关的前沿科技领域仍将是资金关注的焦点。 今日早盘，港股反复震荡，科技板块微涨，截至10:45，恒生科技指数上涨0.1%，成份股中，美团-W、 华虹半导体领涨。Wind数据显示，12月以来，恒生科技指数相关ETF合计净流入超80亿元，其中恒生科 技ETF易方达（513010 ...

风险提示：内需政策效果低预期、关税加征幅度继续大幅超预期、市场流动性危机等 1 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：陈果投资策略 摘要 本周国内中央经济工作会议定调以及美联储"鸽派行动鹰派言论"的组合落地，A股与港股走势呈现分 化。A股资金倾向提前布局春季躁动，流向高弹性的科技成长方向，港股市场受盈利预期疲弱、美联储 降息指引偏鹰及美债收益率反弹等压制，表现相对疲软。 虽然联储降息，但美债收益率近期呈现上升迹象，显示流动性环境未必那么宽松有利。美股呈现风格切 换，显示市场担心AI应用商业化可能落后于再通胀，市场风险偏好也并没有那么高亢。除了持续跟踪 AI应用进展外，还需关注美联储新主席人选与经济数据验证。特朗普12日打破了"超级鸽派"哈塞特稳操 胜券的预期，也将具有偏鹰派底色的凯文·沃什推向台前，市场对未来货币宽松程度预期也受到影响； 而下周即将到来的"超级数据周"，在经历了10月数据缺失的盲飞期后，11月CPI和就业数据将成为验证 宽松政策效果及通胀反弹风险的首份关键报告。综合看，美股短期波动或仍将维持高位，对A股风险偏 好可能带来扰动。 日本央行加息在即。日本央 ...

Group 1 - The report highlights a divergence in market performance between A-shares and Hong Kong stocks, with A-shares showing resilience due to positive signals from the Central Economic Work Conference, while Hong Kong stocks are pressured by weak earnings expectations and rising US Treasury yields [7][18][21] - The report notes that despite the Federal Reserve's interest rate cut, US Treasury yields have recently increased, indicating that the liquidity environment may not be as favorable as expected, which could affect market risk appetite [9][14][19] - The upcoming "super data week" is crucial for validating the effectiveness of the Fed's easing policies and assessing inflation risks, with November CPI and employment data being particularly significant [19][20] Group 2 - The report anticipates that the Bank of Japan will likely resume its interest rate hike cycle, which could lead to concerns about capital outflows and increased volatility in the market, although the impact is expected to be manageable due to prior market pricing [20][21] - The report suggests that the upcoming spring market rally is likely, driven by improved micro liquidity and risk appetite, with a focus on technology and cyclical sectors [25][29] - Investment recommendations include maintaining a core position in financial sectors with stable earnings and high dividend yields, while gradually shifting towards sectors with reasonable valuations and clear industry trends, particularly in the AI chain, renewable energy, and innovative pharmaceuticals [25][26][29]

Core Insights - Alibaba reported FY26Q2 revenue of 247.795 billion yuan, a year-on-year increase of 4.77%, with adjusted EBITA of 9.073 billion yuan, down 77.63% year-on-year, and adjusted net profit of 10.352 billion yuan, down 71.65% year-on-year [1] Group 1: Cloud Business Performance - Cloud business revenue exceeded market expectations, growing 34.50% year-on-year to 39.824 billion yuan, with a sequential increase from FY26Q1 [2] - Adjusted EBITA margin for cloud business reached 9.0%, up from 8.8% in FY26Q1, indicating continuous improvement [2] - Alibaba Cloud holds a 35.8% market share in China's AI cloud market, maintaining a strong competitive position [2] Group 2: E-commerce and Retail Performance - CMR revenue for FY26Q2 reached 78.927 billion yuan, a year-on-year increase of 10%, driven by cross-selling and improved merchant advertising willingness [2] - The number of 88VIP members exceeded 56 million, maintaining double-digit year-on-year growth [2] - International e-commerce revenue grew 10% year-on-year to 34.799 billion yuan, with adjusted EBITA turning positive at 1.62 billion yuan due to logistics optimization [3] Group 3: Investment Outlook - Revenue projections for FY2026-FY2028 are estimated at 1.03 trillion, 1.12 trillion, and 1.24 trillion yuan, with adjusted net profits of 116 billion, 148 billion, and 179 billion yuan respectively [3] - The company is expected to benefit from the commercialization of AI applications in its cloud business, enhancing its technology and consumer attributes [3]

Core Viewpoint - The AI application sector is experiencing significant growth, driven by the rising download numbers of Alibaba's Qianwen app, indicating a strong commercial potential for AI applications and related industries [1][2]. Group 1: AI Application Sector Performance - On November 24, the AI application sector saw a collective rise, with multiple Alibaba-related stocks experiencing notable increases, such as BlueFocus up over 15% and 360 reaching its daily limit [1]. - The trading activity in the AI application sector was robust, with Visual China achieving a transaction volume of 5.614 billion yuan and a turnover rate of 32.12% [1]. - Major inflows of capital were noted, with Keda Xunfei seeing a net inflow of 372 million yuan, while Inspur Information and Tianyu Digital Science had net inflows of 202 million yuan and 238 million yuan, respectively [1]. Group 2: Developments from Major Companies - Ant Group launched its multimodal AI assistant "Lingguang," which reached over 1 million downloads within four days, topping the free tools category in the Apple App Store in China [2]. - Tencent's Hunyuan model team announced the open-source release of HunyuanVideo 1.5, a video generation model capable of producing 5-10 second high-definition videos [2]. Group 3: Market Sentiment and Investment Trends - Discussions around an AI bubble are intensifying, with Google CEO Sundar Pichai acknowledging the presence of both rational and irrational factors in AI investments [2]. - Nvidia's CEO Jensen Huang dismissed concerns about an AI bubble, citing strong revenue expectations that support the legitimacy of AI investments [2]. Group 4: Economic Impact and Investment Opportunities - AI-related capital expenditures have surpassed U.S. consumer spending, becoming a key driver of economic growth, with AI stocks contributing significantly to the S&P 500 index returns [3]. - Major tech companies like Amazon, Google, Meta, and Microsoft are projected to invest approximately $400 billion in AI this year, primarily for data center construction, raising questions about the sustainability of returns [3]. - Research indicates that Chinese AI companies are narrowing the gap with global leaders without excessive spending on AI infrastructure, presenting a potential risk diversification opportunity for investors [3]. - Alibaba's strong fundamentals and superior free cash flow margins position it favorably for rational investment, with projected capital expenditures totaling $55.4 billion from FY2026 to FY2028 [3]. Group 5: Optimism for AI Commercialization - Industry experts are optimistic about the commercialization prospects of AI applications, anticipating a significant acceleration in the formation of an AI industry ecosystem [4]. - The upstream AI industry remains robust, with expectations for high growth in computing power and demand for AI chips, storage, and data centers [4]. - Alibaba's new AI products and models are expected to drive increased demand for AI infrastructure, benefiting its partners in data center operations and related fields [4].

Core Insights - The construction of computing power infrastructure in China is accelerating, supported by increasing policy initiatives aimed at establishing a millisecond-level computing network by 2027 [1][2] - The "millisecond computing" initiative is seen as a paradigm shift in computing service models, enhancing efficiency across various sectors including public services, healthcare, and daily internet usage [2] - The initiative also aims to facilitate the commercialization of AI applications by reducing latency, which is critical for the performance of large models [2] Group 1: Policy and Infrastructure Development - The Shanghai Municipal Communication Administration and the Shanghai Municipal Economic and Information Commission have issued a notice to develop a millisecond-level computing network, aiming for comprehensive coverage and efficient connectivity [1] - The initiative reflects a transition from macro planning to micro implementation in computing power construction, as noted by experts [1] Group 2: Demand and Application - The demand for computing power is increasing, necessitating a unified national computing power system to address issues like resource misallocation and scheduling inefficiencies [2] - Millisecond-level computing is expected to significantly enhance public service capabilities, such as real-time traffic management and remote medical procedures [2] Group 3: Challenges in Implementation - The construction of the millisecond computing network faces challenges including scheduling difficulties, fragmented computing power trading markets, and environmental concerns related to energy consumption [3] - The energy consumption for training advanced AI models, such as GPT-3, is substantial, highlighting the need for efficient energy management in computing power infrastructure [4] Group 4: Corporate Initiatives - Major telecommunications companies are actively developing the millisecond computing network, with achievements such as a three-tier latency network and high-speed connections to national hubs [5] - Companies are encouraged to transition from traditional telecommunications roles to computing power service providers, promoting differentiated latency guarantees and enhancing computing power utilization [5]

Summary of Key Points from Conference Call Records Industry Overview - The conference discusses the AI industry, particularly focusing on the commercialization of AI applications and the evolution of AI agents [1][2][3]. Core Insights and Arguments - **Commercialization Milestone**: The AI industry is approaching a critical point of commercialization, similar to the mobile internet explosion three years after the first iPhone launch. The expected peak for AI applications is anticipated around the third anniversary of ChatGPT's release in late 2025 [2][3]. - **Productization Acceleration**: Starting from Q3 2024, the productization of AI is expected to accelerate, with OpenAI's O1 Pro model marking a significant leap in reasoning capabilities, shifting focus from parameter-driven to capability-driven models [1][4]. - **AI Agent Development**: The main development paradigms for AI agents include embedded, co-governance, and agent modes, with the agent mode currently being the most prevalent. This mode emphasizes planning, tool usage, and memory capabilities, which are critical for the underlying capabilities of large models [5][6]. - **Rise of Reasoning Methods**: The emergence of reasoning methods signifies a shift from simple pattern recognition to logical thinking in AI, enhancing autonomous decision-making and process planning capabilities [7][8]. - **Commercial Value of AI Models**: The Copilot model enhances existing products for efficiency, while the AD model simplifies user experience and clarifies functional distribution, impacting the monetization of AI based on the extent of human labor replacement [10][11]. Important but Overlooked Content - **Investment Opportunities**: Future investment opportunities are identified in new reasoning models, AI agent architectures, and systems with short-term and long-term memory capabilities. Companies that actively invest in these areas are likely to lead in the market [6][30]. - **AI Technology Capability Levels**: AI capabilities are categorized into five levels, ranging from simple instruction handling to complex task execution and collaboration with multiple agents [14][15]. - **Market Dynamics**: The AI industry is undergoing three main phases: technological transformation, data flywheel effects, and economies of scale, which are reshaping market structures and value chains [3][16]. - **Global Monetization Progress**: By mid-2025, OpenAI's annual revenue is projected to reach $13 billion, indicating rapid monetization in the AI sector, with significant contributions from various applications [17][18]. - **Labor Market Impact**: The U.S. labor market is experiencing significant changes due to AI, with over 10,000 jobs lost directly due to AI applications in the first seven months of 2025 [19]. Conclusion - The AI industry is on the brink of a significant transformation, with various factors influencing its trajectory, including technological advancements, market dynamics, and investment opportunities. Companies that adapt quickly and effectively to these changes are likely to thrive in the evolving landscape.