证券研究报告 | 2026年02月10日 | 晨会纪要 | | --- | | 数据日期:2026-02-09 | 上证综指 | 深证成指沪深 | 300 指数 | 中小板综指 | 创业板综指 | 科创 50 | | --- | --- | --- | --- | --- | --- | --- | | 收盘指数（点） | 4123.08 | 14208.43 | 4719.05 | 15487.40 | 4173.83 | 1458.16 | | 涨跌幅度（%） | 1.41 | 2.16 | 1.62 | 1.60 | 2.66 | 2.51 | | 成交金额（亿元） | 9497.26 | 12997.46 | 5227.57 | 4454.52 | 6392.92 | 729.09 | 【常规内容】 宏观与策略 策略周报:杠杆资金和外资流出额增加——2 月第 1 周立体投资策略周报 行业与公司 食品饮料周报（26 年第 6 周）：白酒春节动销渐起，预制菜国标公开征 求意见 海外市场专题:AI Agent 专题：Opus 4.5 开启 AI Agent 拐点，CPU 需 求迎高增 房地产行业快评：如何看 ...

作者：Colin 2月6日，千问App"春节30亿大免单"活动上线首日，9小时内AI订单突破1000万单，服务器一度因"超预期火爆"而卡顿。这场看似是春节红包大战的延续， 实则是阿里精心策划的"AI生活方式全民演习"。当一杯奶茶成为AI Agent普及的导火索，阿里巴巴正在用最接地气的方式，实现从电商巨头到AI超级入口的 蜕变。 在GPLP犀牛财经看来，从活动设置和最终结果上来看，这场春节档的AI大战，千问都毫无疑问的"杀疯了"。 相较元宝用户得通过"分享红包"获得更多抽奖机会，把用户收益和分享行为进行强绑定的设计，让部分用户不得不在微信群中，忍受元宝红包链接泛滥的打 扰，千问的红包玩法更加简单直接，下载就送25元免单卡，兑现门槛低，金额又足够大，甚至有网友给算了一笔账，6口之家，5分钟就能拿下275元的无门 槛免单卡，什么概念呢？如果都用来购买蜜雪冰城的经典饮品柠檬茶，可以免费获得84杯。 同时，相较其他大厂单独一款应用的单兵作战，千问可谓火力全开，1月15日，便接入了淘宝闪购、支付宝、淘宝、飞猪、高德等阿里生态场景，这也意味 着，其他大厂砸下真金白银换来的APP活跃用户，大概率是单纯基于"薅羊毛"心态而来 ...

Investment Rating - The investment rating for the industry is "Outperform" (maintained) [3] Core Insights - The release of the Claude Opus 4.5 model by Anthropic in November 2025 marked a significant turning point for AI agents, leading to a notable increase in CPU demand as it transitioned from a supportive unit to a central scheduling and execution hub [4][6] - The report anticipates a substantial rise in CPU demand driven by the explosion of AI agents, with server CPU configurations evolving from 1:32 to 1:4, and even reaching 1:2 in next-generation products [4][60] - The CPU market is expected to experience a price increase due to rising demand, precious metal material costs, and a scarcity of advanced process capacity, with a 10% price increase already observed as of February 2026 [4][66][69] - The report highlights the competitive landscape of the global CPU market, predicting Intel's market share in server CPUs to be around 55% and AMD's to be approximately 40% by 2026, indicating a clear dominance and head effect [4][72] Summary by Sections 01 Phenomenon of AI Agents - The report discusses the emergence of AI agents as a transformative event in the industry, emphasizing their capabilities in autonomous perception, planning, execution, reflection, and memory [12][18] 02 Evolution of Opus 4.5 Model - Opus 4.5 is described as a significant advancement in AI coding, enabling end-to-end autonomous software engineering capabilities and improving the delivery rate of complex tasks [27][29] - The model's pricing strategy allows for a balance between performance and cost, enhancing its market competitiveness [28][29] 03 Explosion of CPU Demand Under AI Agents - The report outlines how the shift to AI agents necessitates a reevaluation of CPU roles, with CPUs becoming essential for executing complex tasks and managing high concurrency [60][61] - It predicts that the server CPU market will see a significant increase in demand, with a projected market size of $45 billion by 2026 [72]

Group 1 - The stock price of Zhiyuan surged over 40% on February 9, closing at 276.8 HKD per share, marking a 36.22% increase and reaching a market capitalization of 123.4 billion HKD, a record high since its listing [1] - The launch of the global model service platform OpenRouter introduced a mysterious model named "Pony Alpha," which quickly gained attention in the developer community due to its strong coding capabilities and optimization for intelligent agent workflows [1][2] - Predictions from tech bloggers and Silicon Valley entrepreneurs suggest that Pony Alpha may be related to DeepSeek-V4 or Zhiyuan's upcoming GLM-5 model, with a higher likelihood of being developed by a Chinese company due to its name's association with the Year of the Horse [1] Group 2 - Pony Alpha is described as a "cutting-edge foundational model" excelling in programming, intelligent agent workflows, reasoning, and role-playing, with a particular emphasis on its "extremely high tool invocation accuracy" [2] - The emergence of Pony Alpha coincides with the impending explosion of AI Agent applications, which require models to perform multi-round tool invocations, maintain long context memory, and plan complex tasks, leading to exponential increases in token consumption per interaction [2] - Community testing demonstrated that developers used Pony Alpha with Claude Code to run a MineCraft project, generating 170KB of pure JavaScript code in approximately 2 hours, with output quality rated as "beyond expectations" [2]

Investment Rating - The report maintains an "Outperform" rating for the industry [3] Core Insights - The release of the Claude Opus 4.5 model by Anthropic marks a significant turning point for AI agents, leading to a substantial increase in CPU demand as the model transitions from a "super intern" to a "senior architect" [4][12] - The report anticipates a surge in CPU demand driven by the proliferation of AI agents, with server CPU configurations evolving from 1:32 to 1:4, and even reaching 1:2 in advanced products [4][60] - The CPU market is expected to experience a price increase due to rising demand, precious metal price hikes, and limited advanced process capacity, with a 10% price increase already observed as of February 2026 [4][66] Summary by Sections 01 Phenomenon of AI Agents - AI agents are defined as closed-loop intelligent systems capable of autonomous perception, planning, execution, reflection, and memory [12] - The shift from instruction-driven to goal-driven AI is highlighted, enabling end-to-end task closure [12] 02 Evolution of Opus 4.5 Model - Opus 4.5 significantly enhances coding capabilities, acting as a highly autonomous AI engineer, with improved understanding of complex tasks and better performance metrics compared to previous models [29][37] - The model introduces a balanced pricing strategy and a new "effort" parameter for precise cost and performance management [28][29] 03 Explosion of CPU Demand Under AI Agents - The report predicts that the demand for CPUs will increase significantly due to the operational needs of AI agents, which require CPUs to act as central scheduling and execution units [60][61] - The anticipated growth in the server CPU market is projected to reach $45 billion by 2026, with Intel and AMD expected to dominate the market with shares of approximately 55% and 40%, respectively [72]

从端到端语音到超级智能体，FlashLabs以前沿研究回应Agent时代的一次反共识下注。 OpenClaw的爆火，让AI Agent第一次被推向了真实的工程环境。 这一次，Agent不再只是Demo、插件或对话式工具，而是开始尝试进入企业内部，承担持续、复杂、可被验证的工作任务。但几乎与此同时，一个现实问题 也被清晰地暴露出来： 当Agent走向长期运行的真实工作流，它所面临的挑战，远不止是提示词或工具调用，而是部署成本、交互效率，以及底层模型是否 适合"常驻运行"。 这也迫使行业直面一个更底层、却迟早必须回答的问题—— 如果Agent的目标是成为可靠的数字员工，它是否还应该继续建立在上一代模型与交互假设之上？ 在这一阶段，行业事实上已经形成了一种隐含共识：Agent的问题，应当通过更快的产品迭代来解决。 更复杂的Prompt、更精细的流程编排和更丰富的工具调用，成为多数团队默认的前进方向。 但在FlashLabs看来，这种路径回避了一个更根本的问题： 如果底层模型本身并不适合长期运行与实时协作，那么再精巧的产品设计，也只是在放大系统的 结构性上限。 多数团队选择在既有模型能力之上加速产品化，尽快跑通应用与 ...

新浪科技讯 2月9日下午消息，智谱今日股价大涨，收涨近36.22%，报278.8港元。消息面上，2月6日， 全球模型服务平台OpenRouter上线搜索排名第一的匿名模型"Pony Alpha"因其强大的编码能力及针对智 能体工作流的深度优化，引发开发者社区关注，爆火海外社区。 新浪科技讯 2月9日下午消息，智谱今日股价大涨，收涨近36.22%，报278.8港元。消息面上，2月6日， 全球模型服务平台OpenRouter上线搜索排名第一的匿名模型"Pony Alpha"因其强大的编码能力及针对智 能体工作流的深度优化，引发开发者社区关注，爆火海外社区。 据多位硅谷科技企业家及知情人士分析，该模型或为智谱即将发布的新一代模型GLM-5，该模型被描 述为"前沿基础模型"，在编程、智能体工作流、推理及角色扮演方面表现强劲，特别强调其"极高的工 具调用准确率"。这一特性使其在AI Agent（智能体）应用场景中展现出明显优势——开发者可通过 Claude Code等工具调用该模型，实现长达数小时的复杂项目开发。与传统聊天机器人不同，Agentic工 作流需要模型进行多轮工具调用、长上下文记忆与复杂任务规划，这将导致单 ...

克雷西 发自 凹非寺 量子位 | 公众号 QbitAI 2026以来，还有比 OpenClaw （原Clawdbot/Moltbot） 更火的AI和Agent应用吗？ 别的不说，光是GitHub上的星星就有17.7万之多，而且增长速度就像在坐火箭。 能有这样的热度，是因为它早就脱离了聊天机器人的范畴，是公认能真正指挥电脑干活的"数字员工"。 比如这位网友就用三台电脑、15个Agent部署了自己的"数字军团"，他只需要在一个Discord频道里坐镇指挥，就能让它们完成处理邮件、读 PPT、写代码、发推文，甚至撰写每日汇报等一系列工作。 OpenClaw，一行代码真能搞定？ 说实话，为了搭上OpenClaw的这班车，我属实是没少折腾。 不过，想把这位"大神"请进自己的设备，还是得费点周折。 所以很多人都是看着视频里它干活挺爽，但只能眼巴巴地看，真到了自己想用的时候，却发现连大门都进不去。 现在，这个"看得到吃不到"的尴尬局面，终于有解了。 我发现百度智能云搞了个 "作弊级"的方案 ，直接把这个复杂的部署过程给"降维打击"了。 它不需要你懂代码，也不用你折腾环境，直接把这个顶流AI变成了人人都能上手的生产力工具。 ...

但在FlashLabs看来，这种路径回避了一个更根本的问题：如果底层模型本身并不适合长期运行与实时协作，那么再精巧的产品设计，也只是在放大系统 的结构性上限。 多数团队选择在既有模型能力之上加速产品化，尽快跑通应用与商业闭环；而也有少数人选择了一条更慢、风险更高的路径——回到前沿research和模型 层本身，重新审视Agent的基础假设。 FlashLabs，正是后者。 Open Claw的爆火，让AI Agent第一次被推向了真实的工程环境。 这一次，Agent不再只是Demo、插件或对话式工具，而是开始尝试进入企业内部，承担持续、复杂、可被验证的工作任务。但几乎与此同时，一个现实问 题也被清晰地暴露出来：当Agent走向长期运行的真实工作流，它所面临的挑战，远不止是提示词或工具调用，而是部署成本、交互效率，以及底层模型 是否适合"常驻运行"。 这也迫使行业直面一个更底层、却迟早必须回答的问题—— 如果Agent的目标是成为可靠的数字员工，它是否还应该继续建立在上一代模型与交互假设之上？ 在这一阶段，行业事实上已经形成了一种隐含共识：Agent的问题，应当通过更快的产品迭代来解决。 更复杂的Prompt、 ...

据社区实测案例显示，有开发者使用Pony Alpha配合Claude Code运行MineCraft项目，历时约2小时生成170KB纯JavaScript代码，输出质量被评价为"超预 期"。另有测试指出，该模型在SVG生成等细节任务上展现出"Claude Opus 4.5级别的品味"。 Pony Alpha的亮相恰逢AI Agent应用爆发前夜。该模型所主打的"真正可用"的编程与智能体能力，预示着其将在实际应用中产生远超对话类模型的Tokens 消耗量，这为上游AI芯片、存储等半导体产业链带来了明确的需求增长想象空间。 有消息称，包括DeepSeek、智谱在内的多家国内大模型企业都计划于春节前后发布新一代模型。 另外，2月6日，全球模型服务平台OpenRouter悄然上线一款代号为"Pony Alpha"的匿名模型，因其强大的编码能力、超长上下文窗口及针对智能体工作流 的深度优化，迅速引发开发者社区关注。知名X博主karminski-牙医猜测，Pony Alpha是国产大模型，要么是DeepSeek-V4，要么是智谱GLM 新模型。 OpenRouter官方将Pony Alpha描述为"前沿基础模型"，在编程 ...