Workflow
OpenAI
icon
Search documents
给AI装上手和脚,这账能算平吗?
3 6 Ke· 2026-02-27 09:11
大模型市场,直接来了一个「新物种大分叉」。 全球最大AI模型API聚合平台OpenRouter给出数据: 9号到15号这一周,中国大模型的调用量冲到了 4.12万亿Token,第一次超过美国模型的2.94万亿Token。 调用量前五的模型里,中国直接占了四个:MiniMax的M2.5、月之暗面的KimiK2.5、智谱GLM-5、 DeepSeek V3.2。 市场已经彻底分成两拨人:BAT正在「向内收缩」,把模型塞进搜索、电商、办公这些老业务里;新势 力Kimi、智谱、MiniMax则「向外长手」,靠成百上千个Agent死磕开发者生态。 这是商业底层逻辑的彻底分家,大家都在算一笔名为投资回报率的生死账。 01 先得搞清楚一件事:这4.12万亿Token的调用量,到底是谁在用? 要光看新闻标题,肯定以为中国老百姓突然爱上跟AI唠嗑了;OpenRouter的用户构成很有意 思,上面美国开发者占了47.17%,中国开发者只有6.01%。 这4.12万亿Token,主要是全球那帮程序员「用脚投票」投出来的,跟咱们在抖音刷到的那些AI聊天没 啥关系。 所以,Token调用量比DAU实在多了,日活用户能被红包砸出来,能被 ...
黄仁勋:英伟达已接近与OpenAI达成协议 千亿美元投资非承诺但合作深化
Sou Hu Cai Jing· 2026-02-27 08:59
英伟达CEO黄仁勋在2月25日财报电话会上确认,公司已接近与OpenAI达成合作协议,但澄清此前传闻 的1000亿美元投资"从来不是一项承诺"。该协议涉及英伟达向OpenAI提供先进AI芯片及技术支持,以 支撑其大模型训练与推理需求,同时英伟达或将参与OpenAI新一轮融资。黄仁勋形容OpenAI为"一代人 中仅此一家"的企业,强调双方自2016年合作至今,OpenAI所有工作均基于英伟达技术平台。 值得关注的是,OpenAI近期也与AMD达成战略合作,计划部署6吉瓦AMD Instinct GPU,显示其在芯片 供应上正寻求多元化。分析认为,英伟达与OpenAI的协议若最终敲定,将进一步巩固前者在AI算力领 域的霸权地位,同时助力OpenAI推进"星际之门"等大规模AI基础设施建设。 此次合作标志着OpenAI采购模式的重要转变——从通过微软等云服务商间接采购,转向与英伟达直接 合作。与此同时,英伟达正加速AI生态布局,本季度已宣布向Anthropic投资最高100亿美元,并与Meta 签署多年期、价值数十亿美元的芯片供应协议,后者将部署数百万颗Blackwell和Rubin GPU。黄仁勋表 示,公司投资 ...
Anthropic指控中国AI“抄袭”,背后有何资本算计?
Sou Hu Cai Jing· 2026-02-27 08:32
图源:网络 需要说明的是,"蒸馏"是全球部分AI公司训练自家大模型的常用手段,而Anthropic自己,恰恰是"蒸馏"技术的使用者,甚至存在更激进的数据抓取行为。 值得一提的是,Anthropic如今却成了人工智能领域对中国最不友好的公司之一。2025年9月,Anthropic在官方文件中明确封禁对中资企业的服务。其实指 责中国企业进行数据蒸馏,已经成了美国企业惯用套路。就在2026年2月12日,OpenAI向美国国会提交内部备忘录,明确指控DeepSeek通过复杂的混淆手 段,绕过其安全防护对GPT系列模型实施蒸馏行为。 Anthropic这次直接对中国企业"高调指控",一个重大背景是进入2026年2月,美国AI概念股出现多轮明显下跌,资本市场对人工智能的未来产生了"颠覆传 统商业模式"的担忧。CNN评论也指出,不少所谓大模型,其实只是精心包装过的搜索引擎。 马斯克嘲讽、资本变脸,中美AI大战升级,中国凭什么破局? 开工第一天,AI圈就上演"大戏"!美国AI独角兽Anthropic突然发难,指控DeepSeek、月之暗面(Kimi)等3家中国AI企业,靠"蒸馏攻击"抄袭技术。 "蒸馏攻击"是什么?先按下不表 ...
AI“电荒”催命,科技巨头被迫自供电!板块批量涨停
Ge Long Hui· 2026-02-27 08:23
机构:高景气格局有望持续 | 代码 | 名称 | 现价 | 张唱� 涨跌 | | --- | --- | --- | --- | | 300317 珈伟新能 | | 5.36 | +0.89 +19.91% | | 003032 | 南网能源 | 8.22 | +0.75 +10.04% | | 600452 涪陵电力 | | 13.28 | +1.21 +10.02% | | 002015 | 协鑫能科 | 14.16 | +1.29 +10.02% | | 000899 | 赣能股份 | 14.61 | +1.33 +10.02% | | 6660000 | 闽东电力 | 14.43 | +1.31 +9.98% | | 001896 | 豫能控股 | 13.34 | +1.21 +9.98% | | 600744 华银电力 | | 7.94 | +0.72 +9.97% | | 600821 | 金开新能 | 7.39 | +0.67 +9.97% | | 600644 乐山电力 | | 13.12 | +1.04 +8.61% | | 600157 永泰能源 | | 1.84 | +0.13 +7.6 ...
AI集群互连散热专题报告:散热需求向互连系统延伸,连接器散热成为重要补充
Dongguan Securities· 2026-02-27 08:04
2026 年 2 月 27 日 陈伟光 S0340520060001 电话:0769-22119430 邮箱: chenweiguang@dgzq.com.cn SAC 执业证书编号: S0340521020001 电话:0769-22110619 邮箱: luoweibin@dgzq.com.cn S0340524070002 电话:0769-22119302 邮箱: chenzhanqian@dgzq.com.cn 超配(维持) 散热需求向互连系统延伸,连接器散热成为重要补充 深 度 研 AI 集群互连散热专题报告 证 券 研 究 报 告 资料来源:iFind,东莞证券研究所 相关报告 投资要点: 本报告的风险等级为中高风险。 本报告的信息均来自已公开信息,关于信息的准确性与完整性,建议投资者谨慎判断,据此入市,风险自担。 请务必阅读末页声明。 通信行业 SAC 执业证书编号: 罗炜斌 AI集群功耗上扬,集群散热需求增长。AI算力需求呈指数级爆发直接推 动了AI集群功耗上扬,从单芯片到机柜级别的功耗密度的激增已经超越 了传统数据中心的设计极限。以英伟达产品为例,2026年是其AI硬件产 品将从H100/H ...
海外价值获验证,国内市场开启高增长周期
Dongguan Securities· 2026-02-27 08:04
AI 编程行业深度报告 2026 年 2 月 27 日 卢芷心 S0340524100001 电话:0769-22119297 邮箱: luzhixin@dgzq.com.cn S0340521020001 电话:0769-22110619 邮箱: luoweibin@dgzq.com.cn S0340520060001 电话:0769-22119430 邮箱: chenweiguang@dgzq.com.cn 超配(维持) 海外价值获验证,国内市场开启高增长周期 深 度 资料来源:iFind,东莞证券研究所 投资要点: 本报告的风险等级为中高风险。 本报告的信息均来自已公开信息,关于信息的准确性与完整性,建议投资者谨慎判断,据此入市,风险自担。 请务必阅读末页声明。 SAC 执业证书编号: 罗炜斌 SAC 执业证书编号: AI Coding向"自主型Agent"跃迁,未来市场空间广阔。AI Coding已 成为众多AI应用领域中落地速度最快、发展潜力最大的细分赛道之一, 其产品或功能演进高度依赖底层大模型能力的发展。当前,国际前沿编 程大模型发展由海外厂商引领,而国产大模型正展现出强劲的追赶态 势。在发展趋 ...
马斯克要把数据中心送太空!黄仁勋冷笑:散热成本比火箭还高!
Sou Hu Cai Jing· 2026-02-27 07:55
数据不会说谎。2025年全球AI芯片市场规模突破1200亿美元,台积电、三星的3nm产线火力全开,单块 GPU的算力三年翻了10倍。可另一边,全球电力增速常年维持在3%左右,欧洲甚至因能源危机砍了数 据中心的供电配额。马斯克的焦虑不是空穴来风:当单个AI集群的功耗突破100兆瓦(相当于一座小型 核电站),当数据中心耗电量占全球10%且还在飙升,地面上的"电"确实快不够用了。 "记住我的话,36个月内太空会成为部署人工智能的最便宜去处。"马斯克的语气带着惯有的笃定,像极 了当年宣布"殖民火星"时的决绝。但这次,他的论据扎得很痛:"芯片产出几乎呈指数级增长,但电力 产出是平的。那么你要如何让这些芯片通电?靠魔法电源?魔法电力精灵吗?" 当马斯克用"魔法电力精灵"调侃地面电力困境时,黄仁勋正拿着散热板图纸算尺寸。一个在太空画饼, 一个在现实算账——这场关于"太空数据中心"的隔空对话,撕开了AI时代最尖锐的矛盾:人类对算力的 贪婪,正在把地球逼到电力极限。 二、黄仁勋的"冰冷现实":没空气的太空,散热板得造到"看得见" 就在马斯克畅想太空算力乌托邦时,黄仁勋在财报电话会议上泼了盆冰水:"在太空中,能源与散热是 两回事 ...
拒绝成为“战争机器”!逾百名谷歌员工联名上书:要求在美军合同中划定红线
Hua Er Jie Jian Wen· 2026-02-27 07:48
Core Viewpoint - The intense negotiation between the Pentagon and AI startup Anthropic regarding military technology boundaries is causing significant reactions in Silicon Valley, with employees from major tech companies voicing their concerns about ethical implications and potential misuse of AI technology [1][2]. Group 1: Employee Reactions - Over 100 Google AI employees submitted a letter to management demanding clear boundaries in collaboration with the military, specifically opposing the use of their technology for mass surveillance or autonomous weapon systems [1]. - Nearly 50 OpenAI employees and 175 Google employees also published an open letter criticizing the Pentagon's strategy to divide tech companies and urging them to unite against unethical practices [1]. Group 2: Pentagon Pressure and Responses - The Pentagon has exerted significant pressure on Anthropic to allow the military to use its Claude model for "all legitimate purposes," which Anthropic's CEO Dario Amodei has firmly rejected, citing moral objections [2]. - Google employees expressed their desire to prevent any transactions that would cross ethical boundaries, indicating a strong internal push against military collaborations [2]. Group 3: Google Executives' Stance - Jeff Dean, a prominent Google engineer, publicly supported Anthropic's position, emphasizing that mass surveillance violates constitutional rights and can lead to misuse for political or discriminatory purposes [3]. - Google has a complex history with employee activism, having previously faced protests that led to the cancellation of military contracts, highlighting ongoing internal ethical scrutiny [3]. Group 4: Military Strategy and AI Risks - In response to Anthropic's stance, the Pentagon is seeking alternative solutions, having already reached an agreement with xAI to use its Grok model for military applications [4]. - The Pentagon's negotiations with Google are ongoing, and there are threats to invoke the Defense Production Act to compel Anthropic's compliance, indicating a high-stakes environment for AI in military applications [4]. - Concerns about the potential risks of AI in military contexts are underscored by simulations showing that top AI models could opt for nuclear weapon use under pressure, raising alarms about the implications of AI in decision-making [5][6].
广发证券:SRAM提升AI推理速度 相关架构进入主流大厂视野
Zhi Tong Cai Jing· 2026-02-27 07:35
广发证券发布研报称,在大模型应用中,相比依赖外置HBM,SRAM可显著降低权重与激活数据的访 延迟与抖动,从而改善Time-to-First-Token与尾时延表现。目前,Groq与Cerebras都相继推出基于 SRAMAI芯片。SRAM架构进入主流视野,根据Groq官网以及市场媒体报道,英伟达此前斥资200亿美 元获得Groq的知识产权的非独家授权;OpenAI与Cerebras签署100亿美元合同,部署多达750兆瓦的定制 AI芯片。 广发证券主要观点如下: SRAM是片上高带宽存储层 存储分级为SRAM、HBM、DRAM和SSD,其中SRAM(静态随机存取存储器)集成在CPU、GPU计算核 心附近的片上存储,具备纳秒级访问时延与高度确定性的带宽特性,带宽高但容量小、成本高。 SRAM可提升AI推理速度 根据Cerebras官网,其晶圆级引擎3(WSE-3)芯片集成44GB SRAM,片上存储带宽达21PB/s,在OpenAI GPTOSS120B推理任务中实现>3000tokens/s的输出速度,较主流GPU云推理快约15×。此外,2026年2 月,OpenAI推出首个运行在Cerebras Syst ...
东吴证券:端云协同驱动AI入口重塑 端侧模型牵引硬件重构
智通财经网· 2026-02-27 07:07
Core Insights - The evaluation system for cloud-based large models is shifting from purely capability metrics to the actual completion of tasks, with a focus on code capabilities and multi-agent systems by leading overseas companies since 2026 [1] - The dual capability stack of "fast interaction + long reasoning" is expected to become a significant evolution direction for general-purpose agents in the near future [2] - The collaboration between edge models and cloud models is emphasized, with edge models handling high-frequency, lightweight tasks locally, while heavier reasoning tasks are processed in the cloud [3] Cloud Models - The expansion of capability boundaries and cost restructuring are occurring simultaneously in cloud models, with a focus on task completion [1] - Leading companies are intensively laying out code capabilities and multi-agent systems to enhance performance [2] Code Models - The reasoning demands in the era of intelligent agents are evolving along two optimization directions: long-chain complex reasoning and real-time interaction [2] - Low-latency agents like OpenAI's Codex-Spark prioritize interactive AI experiences, while agents like Claude4.6 focus on improving success rates in complex tasks through increased context length [2] Edge Models - The evolution of edge models is characterized by efficiency optimization and capability compression under a collaborative framework with cloud models [3] - Multi-modal capabilities are becoming a key competitive point for edge models, with a focus on achieving zero-latency interactions [3] Hardware Reconstruction - The industry is expected to focus on high-frequency demand scenarios in 2024, with a shift towards multi-modal creative capabilities by 2025 [4] - Key components for edge models are undergoing upgrades in memory and power consumption to enhance user experience [4] Future Outlook - Next-generation flagship SoC platforms like Qualcomm's Snapdragon 8 Elite Gen 6 are anticipated to provide enhanced hardware support for the complexity and multi-modality of edge AI functions [5]