[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fyjU3A6R1HU40h959qjBmyBo9L1gm7IClQTQ1AE5fJ5o":3},{"code":4,"msg":5,"data":6},200,"操作成功",{"id":7,"title":8,"content":9,"digest":10,"source":10,"coverPath":11,"thumbsCoverPath":12,"isTop":13,"isShow":14,"baseClick":13,"clickCount":15,"createTime":16,"typeId":17,"isNewest":18,"newsInfoTypeRespVo":19,"voiceUrl":22,"voiceSize":23,"taskId":24,"releaseTime":25,"titleEn":26,"contentEn":27,"voiceUrlEn":28,"taskIdEn":29,"voiceSizeEn":30},1563,"Meta买下Manus，买的是AI变现模板","\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002Fbee7a4e650974290ac4964ccec819c04.png\" width=\"631\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong class=\"ql-lineHeight-1-75\" style=\"font-size: 18px; color: rgb(255, 153, 0);\">临近2025年岁末，AI圈又沸腾了起来。\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">12月30日，Meta宣布收购AI应用Manus背后的公司蝴蝶效应。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">有消息称，此番收购金额高达数十亿美元。作为Manus的早期投资者，真格基金官方发文表示，这是Meta自成立以来第三大的并购，仅次于WhatsApp和Scale AI。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">多位接近交易的人士透露，从双方正式接触到达成协议，谈判周期仅十余天。“快到还怀疑过这是不是一个假的offer。”真格基金合伙人刘元表示。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">收购完成后，蝴蝶效应将保持独立运作，创始人肖弘出任Meta副总裁。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">蝴蝶效应成立于2022年，三年时间估值从1400万美元增长超过200倍。我们了解到，在Meta收购前，Manus正以约200亿美元估值进行新一轮融资。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">三年，估值增长200倍\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">蝴蝶效应的融资节奏极快。 \u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">2023年2月完成种子轮融资，投后估值1400万美元，投资人是真格基金。当时，蝴蝶效应第一款产品Monica刚上线不久，这是一款AI浏览器插件。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">半年后，2023年8月，蝴蝶效应完成天使轮，估值升至5000万美元，真格基金继续跟投。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">2024年11月A轮，蝴蝶效应估值达到了8500万美元。这一轮由红杉中国、腾讯、王慧文、真格基金共同投资。此时，Monica海外用户已突破1000万，是中国AI行业少有的盈利产品。 \u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">2025年4月B轮，蝴蝶效应估值近5亿美元，领投方有了硅谷顶级VC Benchmark Capital。这家机构投资过Uber、Twitter、Snapchat的早期。此时，距离Manus正式发布仅过去一个月。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">接下来，就是今天这则引发轰动的、被Meta收购的消息。按保守估计30亿美元计算，真格基金种子轮投资Manus两年获得超过200倍回报。A轮投资者回报约35倍，B轮投资者半年多获得5-6倍回报。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">支撑估值短时间内飞升的，是Manus的商业化速度。2025年12月，Manus宣布年度经常性收入（ARR）突破1亿美元。从3月正式发布到12月，用时270天。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">作为对比，AI工作管理及效率工具Slack从0到1亿美元ARR用了2.5年。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">值得关注的是，Manus的增长策略是零市场预算。肖弘判断，AI的Token成本会越来越便宜，但互联网用户获取成本会越来越贵。3月发布时，Manus在社交媒体自然传播，内测邀请码被炒到上千元，官网因流量过大崩溃。在GAIA基准测试中，Manus超越了OpenAI的同类产品。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">产品逻辑的差异在于，Chatbot给答案，Agent给结果。“做一个研究并生成精美的PPT，全程无需干预，只需几分钟，”肖弘在一次采访中表示，Agent直接交付结果，而不是让用户拿着答案再花两小时变成结果。 \u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">技术架构上，Manus采用“大模型+云端虚拟机”。肖弘认为，AI的终极“外壳”应该是电脑，给AI配一台专属虚拟机，它理论上能完成人类在电脑上能做的所有工作。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">“我记得第一次感到震撼，是看到Manus执行git clone命令，将开源项目下载到自己的‘电脑’里来解决问题，这像极了人类‘使用工具’的行为。”肖弘称。 \u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">这个架构能处理海量长尾任务，无论是安装特定软件还是运行自己编写的代码，挑战在于速度和资源消耗，但长期看这些问题会解决。 \u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">为何拒绝字节跳动，转投Meta？\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">理解肖弘为什么接受Meta，要先理解他为什么拒绝字节跳动。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">市场传言，2024年初，Monica刚开始高速增长时，字节跳动高层与肖弘在香港单独会面，出价3000万美元收购蝴蝶效应。肖弘拒绝了。 \u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">当时，Monica已实现盈利，现金流健康。更重要的是，肖弘看到了更大机会。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">2024年，Cursor、Devin等产品验证了Agent可行性。肖弘意识到，技术能力一直在进化，但产品形态常常滞后。去年底，Agent这种能进行复杂规划和自主执行的新能力出现，但缺乏好的产品化形态。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">拒绝字节后，蝴蝶效应做了一个更大胆的决定。2024年初立项AI浏览器，团队投入7个月，从底层技术开始自己编译Chrome内核。 \u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">这看起来是顺理成章的演进，但2024年中，肖弘叫停了这个项目。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">肖弘表示：“AI浏览器本质是抢用户的鼠标，会产生冲突，更重要的是，浏览器天花板有限，无法处理需要专业软件、复杂环境的长尾任务。”\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">砍掉投入7个月的项目后，转向全新方向，今年3月，Manus正式发布。肖弘认为，“永远从技术能解决用户的什么根本问题出发。真正的难度不在分析，而在于有无勇气坚持清晰的答案。”\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">从拒绝3000万美元到接受数十亿美元，中间不到两年。团队成员已快速实现财务自由，肖弘进入全球科技权力核心；在战略层面，扎克伯格承诺保持独立运作。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">肖弘用独立上市的可能性，换取站在巨人肩膀上快速覆盖全球的确定性。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Meta为什么必须买Manus\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">这桩投资案能迅速达成，很大程度上，不是投资委员会看财务模型的决策，而是CEO亲身体验后的判断。 \u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">马克·扎克伯格（Mark Zuckerberg）及Meta多位核心高管都是Manus的长期用户。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Meta在AI上投入数百亿美元，向顶尖研究者开出上亿美元年薪，开源Llama曾经一度赢得开发者好评。但一个核心问题始终未解：如何把模型能力转化为持续收入。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">反观硅谷其他科技巨头，OpenAI旗下的ChatGPT付费用户数（Plus\u002FPro）订阅用户已突破3700万，企业版在快速增长。Google将Gemini整合进Workspace，押注企业市场；Anthropic的Claude在编程、研究场景建立了优势。微软通过Copilot将AI渗透企业级应用。 \u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">相比之下，Meta的Llama开源战略虽然一度很成功，但变现路径不清晰，尤其是今年Llama 4的仓促发布，损害了信誉，引发了社区对Meta的信任危机，直接透露出了扎克伯格在激烈竞争下的焦虑与失策。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">与此同时，Agent被视为科技大厂下一个激烈战场。今年7月，扎克伯格在公开信中写道：“超级智能将带来个人赋能的新时代，它赋予人们更强大的行动力，让我们能按自己的意愿去推动建设世界。”\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Manus提供了一个现成答案。它已验证订阅制商业模式（ARR破1亿美元），表明Agent产品也拥有与Chatbot一样的商业变现前景，有的时候，用户付费的需求不仅仅是为了获得更好的答案，也需要“把事办完”。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">留给Meta的时间窗口已经在收紧。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">行业普遍认为，2023年是Chatbot时代，2024年Cursor、Devin验证了Agent可行性，2025年Agent进入商业化爆发期。OpenAI准备推出更强Agent能力，Google收购或投资多家AI应用公司，微软通过Copilot建立护城河。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">在全球范围内，同时满足1亿美元ARR、技术壁垒、零市场预算增长这些条件的标的极少。如果Meta自研，可能需要2-3年跑通商业化路径，而且成功概率未知。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">收购不仅是买产品，更是买团队。肖弘、季逸超（高中时开发猛犸浏览器的天才程序员）、张涛（资深产品经理）这个组合已证明执行力。 \u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Meta超级智能实验室负责人汪滔（Alexandr Wang）转发收购消息时说，Manus团队在探索当今大模型能力过剩问题上处于世界领先水平。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">该实验室正在新加坡扩展团队，Manus原有约100名成员已加入其当地组织。 \u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">AI军备赛，新战线启动\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">过去两年，AI竞争聚焦在模型上。OpenAI的GPT-4、Google的Gemini、Anthropic的Claude、Meta的Llama，比的是参数量、Benchmark跑分、技术论文。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">但在这条路上，烧钱容易，赚钱难。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">正如汪韬所言，Manus团队在探索大模型能力过剩问题上处于世界领先水平。能力过剩，也就意味着，模型已经足够强，瓶颈不在技术，而在于如何把能力转化为用户愿意付费的产品。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">所以，收购Manus，Meta买的不是一个产品，而是一套已验证的AI变现模板。Manus的订阅模式、定价策略、用户获取方式、产品迭代节奏，都可以复制到Meta其他AI产品线。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">过去，科技巨头自信“自研至上”，现在转向“收购补齐”，面对AI行业激烈的竞争，科技巨头战略也在快速切换。在AI时代，似乎并购比自研更高效，时间更短、确定性更高、能直接获得验证过的商业模式和团队。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">同时，Manus也代表了一种路径——中国团队、全球产品、被美国巨头收购。不需要在中国市场做大，而是可以直接出海，在全球舞台竞争。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">其背后有中国市场的特殊属性，AI应用面临更激烈竞争，用户付费意愿相对较低。而在全球，市场天花板更高，用户付费意愿有一定基础，全球顶级VC和科技巨头也愿意给出更高估值。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">这也许将促使成为科技巨头进一步转变思路，模型层自研保持技术主导权，应用层收购快速获取商业化能力。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">正如十年间，连续五次押注这支年轻团队的刘元说的那样，“属于中国这一代年轻创业者的时代已经到来”，AI打破了传统科技创业的资源壁垒。过去做搜索、做社交网络，需要庞大服务器、复杂技术架构、海量资金投入，现在做AI应用，调用大模型API，小团队就能快速验证产品。 \u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">从3000万美元到数十亿美元，从拒绝字节到接受Meta，从Monica到Manus，肖弘与蝴蝶效应故事，是AI竞争进入新阶段的一个的缩影。Manus用270天证明Agent产品可以快速实现订阅收入，Meta用数十亿美元证明应用层的价值不亚于模型层。看样子，AI巨头的战争，又开辟了一个新的战线。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(136, 136, 136);\">【新闻来源】新浪网 \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fcj.sina.com.cn\u002Farticles\u002Fview\u002F2853016445\u002Faa0d937d0200131y4\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(136, 136, 136);\"> https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1RTDQy?ocid=BingHp01&cvid=6936317f054647a2afcd53fafcde084a&ei\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(136, 136, 136);\">（本网转发此文章，旨在为读者提供更多的信息资讯，所涉内容不构成投资、消费建议。文章事实如有疑问，请与有关方核实，文章观点非本网观点，仅供读者参考。）\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>","","https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F01\u002Fbb87cd5a3813488d913b88cc65af91c7\u002FAI领域.jpg","https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F01\u002Fthumbs\u002Fbb87cd5a3813488d913b88cc65af91c7\u002FAI领域.jpg",0,1,39,"2026-01-05 10:39",2,false,{"id":17,"name":20,"enName":21},"芯位视野","Xinwei Vision","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3A89a4fe52-76f3-4e81-9c1f-59dba63d248c%3A0.wav?Expires=1768846454&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=FCIdCDiwGsLLE2d8AloBPeFz68Q%3D",18700890,"89a4fe52-76f3-4e81-9c1f-59dba63d248c","2026-01-05 10:36","Meta bought Manus, it's buying an AI monetization template","\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002Fbee7a4e650974290ac4964ccec819c04.png\" width=\"631\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong class=\"ql-lineHeight-1-75\" style=\"font-size: 18px; color: rgb(255, 153, 0);\">As the end of 2025 approaches, the AI circle is once again in an uproar.\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">On December 30, Meta announced the acquisition of the company behind the AI application Manus, Butterfly Effect.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">It is reported that the acquisition amount reached several billion dollars. As an early investor in Manus, Zhenge Fund officially stated that this is the third largest acquisition by Meta since its establishment, second only to WhatsApp and Scale AI.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Several people close to the deal revealed that the negotiation period from the formal contact between the two sides to the agreement was only a few days. \"It was fast enough that I even doubted whether it was a fake offer,\" said Liu Yuan, a partner at Zhenge Fund.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">After the acquisition, Butterfly Effect will continue to operate independently, with Xiao Hong, the founder, serving as Vice President of Meta.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Butterfly Effect was founded in 2022, and its valuation has grown more than 200 times in three years, from 14 million USD. We learned that before Meta's acquisition, Manus was raising funds at a valuation of about 2 billion USD.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Three years, valuation increased 200 times\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">The financing rhythm of Butterfly Effect is extremely fast.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">In February 2023, Butterfly Effect completed the seed round, with a post-investment valuation of 14 million USD, and the investor was Zhenge Fund. At that time, Butterfly Effect's first product Monica had just been launched, which is an AI browser extension.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Half a year later, in August 2023, Butterfly Effect completed the angel round, with the valuation rising to 50 million USD, and Zhenge Fund continued to invest.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">In November 2024, the A round, Butterfly Effect's valuation reached 85 million USD. This round was jointly invested by Sequoia China, Tencent, Wang Huiwen, and Zhenge Fund. At that time, Monica's overseas users had exceeded 10 million, and it was one of the few profitable products in the Chinese AI industry.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">In April 2025, the B round, Butterfly Effect's valuation approached 500 million USD, led by Silicon Valley's top VC Benchmark Capital. This institution has invested in Uber, Twitter, and Snapchat in their early stages. At that time, it had been only one month since Manus was officially launched.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Then came today's sensational news of being acquired by Meta. According to a conservative estimate of 3 billion USD, Zhenge Fund received over 200 times return on its seed round investment in Manus after two years. The return for A round investors was about 35 times, and the B round investors gained 5-6 times returns in more than half a year.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">The rapid increase in valuation is supported by Manus' commercialization speed. In December 2025, Manus announced that its annual recurring revenue (ARR) had exceeded 100 million USD. From March when it officially launched to December, it took 270 days.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">By comparison, the AI work management and productivity tool Slack took 2.5 years to reach 100 million USD ARR.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Notably, Manus' growth strategy is zero marketing budget. Xiao Hong judged that the cost of AI tokens would become cheaper, but the cost of acquiring internet users would become more expensive. When Manus was launched in March, it spread naturally on social media, and the internal test invitation codes were sold for thousands of yuan, and the official website crashed due to excessive traffic. In the GAIA benchmark test, Manus surpassed the similar products of OpenAI.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">The difference in product logic is that Chatbot gives answers, while Agent gives results. \"Do a research and generate a beautiful PPT, without any intervention, it takes just a few minutes,\" said Xiao Hong in an interview. Agent directly delivers results, rather than letting users take two hours to turn the answer into results.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Technically, Manus uses \"large model + cloud virtual machine\". Xiao Hong believes that the ultimate \"shell\" of AI should be a computer. Provide an exclusive virtual machine for AI, which theoretically can complete all tasks humans can do on a computer.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">\"I remember the first time I felt shocked was when I saw Manus execute the git clone command, downloading open source projects into its own 'computer' to solve problems, which is very similar to human 'using tools' behavior,\" said Xiao Hong.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">This architecture can handle massive long-tail tasks, whether installing specific software or running code written by oneself. The challenge lies in speed and resource consumption, but these issues will be solved in the long run.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Why reject ByteDance and join Meta?\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">To understand why Xiao Hong accepted Meta, we need to first understand why he rejected ByteDance.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Market rumors say that in early 2024, when Monica started to grow rapidly, ByteDance executives met with Xiao Hong alone in Hong Kong, offering 30 million USD to acquire Butterfly Effect. Xiao Hong refused.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">At that time, Monica had already achieved profitability, with healthy cash flow. More importantly, Xiao Hong saw a bigger opportunity.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">In 2024, products like Cursor and Devin validated the feasibility of Agents. Xiao Hong realized that technological capabilities have been evolving, but product forms often lag behind. At the end of last year, new capabilities such as complex planning and autonomous execution emerged, but there was a lack of good productization forms.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">After rejecting ByteDance, Butterfly Effect made a bolder decision. In early 2024, they initiated an AI browser project, and the team spent seven months working from the bottom up to compile the Chrome kernel themselves.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">This seemed like a natural evolution, but in mid-2024, Xiao Hong halted the project.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Xiao Hong said, \"An AI browser essentially competes for the user's mouse, causing conflicts. More importantly, the browser has limited ceiling and cannot handle long-tail tasks requiring professional software and complex environments.\"\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">After cutting off the project that had taken seven months, they shifted direction, and in March of this year, Manus was officially launched. Xiao Hong believes, \"Always start from the fundamental problems that technology can solve for users. The real difficulty is not analysis, but having the courage to stick to a clear answer.\"\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">From refusing 30 million USD to accepting several billion USD, it was less than two years. Team members have quickly achieved financial freedom, and Xiao Hong entered the core of global tech power.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">On the strategic level, Zuckerberg promised to maintain independent operations.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Xiao Hong exchanged the possibility of an independent listing for the certainty of rapidly covering the globe by standing on the shoulders of giants.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Why must Meta buy Manus\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">The quick completion of this investment largely wasn't a decision based on the financial models reviewed by the investment committee, but rather a judgment after the CEO experienced it personally.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Mark Zuckerberg and many other key executives at Meta have been long-term users of Manus.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Meta has invested hundreds of billions of dollars in AI, offering top researchers salaries of tens of millions of dollars, and open-sourcing Llama once won praise from developers. However, a core issue has always remained unsolved: how to convert model capabilities into continuous revenue.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">In contrast, other tech giants in Silicon Valley, the paid user count (Plus\u002FPro) of ChatGPT under OpenAI has exceeded 37 million, and the enterprise version is growing rapidly. Google integrated Gemini into Workspace, betting on the enterprise market; Anthropic's Claude has established advantages in programming and research scenarios. Microsoft has penetrated enterprise-level applications through Copilot.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Compared to this, although Meta's open-source strategy for Llama was once successful, its monetization path was unclear, especially the hasty release of Llama 4 this year damaged credibility and triggered a trust crisis within the community, directly revealing Zuckerberg's anxiety and missteps in intense competition.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Meanwhile, Agents are seen as the next fierce battlefield for tech giants. In July this year, Zuckerberg wrote in an open letter, \"Super intelligence will bring a new era of personal empowerment, empowering people with stronger action capabilities, allowing us to build the world according to our own desires.\"\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Manus provides a ready-made answer. It has verified the subscription-based business model (ARR exceeding 100 million USD), indicating that Agent products also have the same commercial potential as Chatbots. Sometimes, users pay not just for better answers, but to get things done.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">The time window left for Meta is tightening.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Industry consensus is that 2023 was the Chatbot era, 2024 saw Cursor and Devin validate the feasibility of Agents, and 2025 marks the commercialization boom of Agents. OpenAI is preparing to launch a stronger Agent capability, Google is acquiring or investing in multiple AI application companies, and Microsoft is building a moat through Copilot.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Globally, there are few targets that meet these conditions simultaneously: 100 million USD ARR, technical barriers, and growth with zero marketing budget. If Meta develops it in-house, it may take 2-3 years to verify the commercialization path, and the success probability is unknown.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Acquisition is not just buying a product, but also a team. The combination of Xiao Hong, Ji Yichao (a genius programmer who developed the Mammoth browser in high school), and Zhang Tao (a senior product manager) has proven their execution ability.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Alexandr Wang, head of Meta's Super Intelligence Lab, when forwarding the acquisition news, said that the Manus team is leading the world in exploring the issue of excess model capabilities.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">The lab is expanding its team in Singapore, and the original 100 members of Manus have joined its local organization.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">AI arms race, new front line starts\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Over the past two years, AI competition focused on models. OpenAI's GPT-4, Google's Gemini, Anthropic's Claude, and Meta's Llama compete on parameters, benchmark scores, and technical papers.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">But on this path, spending money is easy, making money is hard.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">As Wang Tao said, the Manus team is leading the world in exploring the problem of excess model capabilities. Excess capacity means that the model is strong enough, and the bottleneck is not technology, but how to convert the capability into a product that users are willing to pay for.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Therefore, acquiring Manus, Meta is not buying a product, but a verified AI monetization template. Manus's subscription model, pricing strategy, user acquisition methods, and product iteration rhythm can all be replicated across Meta's other AI product lines.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">In the past, tech giants were confident in \"self-research supremacy\", but now they are shifting towards \"acquisition to fill gaps\". Facing the fierce competition in the AI industry, tech giants' strategies are also rapidly changing. In the AI era, it seems that acquisitions are more efficient than self-research, with shorter timeframes, higher certainty, and direct access to verified business models and teams.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">At the same time, Manus also represents a path - a Chinese team, a global product, acquired by an American giant. There's no need to grow big in the Chinese market, but instead, directly go global and compete on the world stage.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Behind it is the special nature of the Chinese market, where AI applications face more intense competition, and users' willingness to pay is relatively low. In contrast, the global market has a higher ceiling, and users' willingness to pay has a certain foundation. Global top VCs and tech giants are also willing to give higher valuations.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">This may prompt tech giants to further change their thinking: self-developing the model layer to maintain technological leadership, and acquiring the application layer to quickly gain commercialization capabilities.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">As Liu Yuan, who has continuously bet on this young team for five times, said, \"The era for this generation of young entrepreneurs in China has already arrived.\" AI has broken through the traditional resource barriers of tech startups. In the past, doing search or social networks required large servers, complex technical architectures, and massive capital investment. Now, with AI applications, small teams can quickly validate products by calling large model APIs.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">From 30 million USD to several billion USD, from rejecting ByteDance to accepting Meta, from Monica to Manus, Xiao Hong and Butterfly Effect's story is a microcosm of a new phase in AI competition. Manus proved in 270 days that Agent products can quickly achieve subscription revenue, and Meta proved with several billion USD that the value of the application layer is no less than that of the model layer. It seems that the war among AI giants has opened a new front.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(136, 136, 136);\">【News Source】 Sina.com \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fcj.sina.com.cn\u002Farticles\u002Fview\u002F2853016445\u002Faa0d937d0200131y4\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(136, 136, 136);\"> https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1RTDQy?ocid=BingHp01&cvid=6936317f054647a2afcd53fafcde084a&ei\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(136, 136, 136);\">（This article is reprinted by the website to provide readers with more information and news. The content mentioned does not constitute investment or consumer advice. If there are any questions about the facts of the article, please verify with relevant parties. The views expressed in the article are not the views of the website and are for reference only.）\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3A3cacb05b-8823-4503-9b3b-2f73da47bf30%3A0.wav?Expires=1774838434&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=naolv6FlhG2miQWmj07HhARC%2BOY%3D","3cacb05b-8823-4503-9b3b-2f73da47bf30",18242930]