[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-embodied-ai-capital-flows-to-brains-and-bubbles-en":3,"article-related-embodied-ai-capital-flows-to-brains-and-bubbles-en":30,"series-industry-c996e251-6cca-452c-b87d-a88a6c9ada28":72},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":22,"views":26,"created_at":27,"published_at":28,"topic_cluster_id":29},"c996e251-6cca-452c-b87d-a88a6c9ada28","embodied-ai-capital-flows-to-brains-and-bubbles-en","具身智能的钱正在往大脑集中，泡沫也在集中","\u003Cp data-speakable=\"summary\">具身智能资本正在向“大脑派”和世界模型集中。\u003C\u002Fp>\u003Cp>我认为具身智能当前最值得押注的不是本体，而是大脑与数据基础设施。量子位统计显示，2026年上半年国内具身智能赛道融资约438亿元，超过一半流向“大脑派”公司；同期本体派只占12.8%，甚至低于零部件公司。这个分化不是短期噪音，而是资本对行业价值分配的重新定价：硬件决定能不能做，模型决定能不能规模化复制。\u003C\u002Fp>\u003Ch2>第一，资本已经把“上限”押在大脑上了\u003C\u002Fh2>\u003Cp>投资人那句“本体兜住下限，大脑决定上限”，不是口号，是资金流向的真实写照。Pre-A轮平均7亿元，B轮平均22.5亿元，这种量级放在多数行业里已经接近C轮、D轮。它石智航Pre-A轮拿到4.55亿美元，千寻智能在2月至6月连融四轮、累计近50亿元，说明市场不再按传统里程碑定价，而是按“谁更像未来的机器人操作系统”定价。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782207170550-kt87.png\" alt=\"具身智能的钱正在往大脑集中，泡沫也在集中\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>这背后的逻辑很直接。机器人本体当然重要，但它更像制造业问题，考验供应链、成本和量产；大脑则更像AI问题，边际成本低、复用性强、迁移性高。中国在制造端有天然优势，意味着本体会快速进化，但真正能拉开估值差距的，还是谁先把感知、推理和行动串成闭环。资本追逐的是可复制的能力，而不是一台机器的单点性能。\u003C\u002Fp>\u003Ch2>第二，世界模型正在取代VLA成为新叙事中心\u003C\u002Fh2>\u003Cp>今年上半年有融资动态的35家“大脑派”公司里，27家在研发世界模型，占比接近八成，这个数字说明赛道共识已经发生迁移。2024年大家还在比谁的VLA更强、谁的真机数据更多；到了今天，不做世界模型似乎就像没跟上时代。它不只是技术路线变化，更是融资语言变化，因为“世界模型”比“VLA”更能承接想象空间，也更容易让投资人相信公司在做下一代平台。\u003C\u002Fp>\u003Cp>但真正重要的不是名词，而是底层实现。蒋子元的判断很有代表性：很多所谓世界模型，本质上只是把视频生成模型作为backbone，而VLA更多依赖语言模型作为backbone。换句话说，路线之争没有外界想得那么硬，创业公司会不断切换更顺手的架构。就像视频生成从U-Net转向DiT，行业不会为一种范式守身如玉，谁能更快把效果做出来，谁就会成为“正确路线”。\u003C\u002Fp>\u003Ch2>第三，决定胜负的不是路线洁癖，而是数据基础设施\u003C\u002Fh2>\u003Cp>自变量机器人算法负责人甘如饴说得更到位：比起世界模型还是VLA，底层的数据基础设施才是核心竞争力。这套基础设施覆盖数采、训练和评测全流程，目标不是做一次漂亮演示，而是形成可规模化运转的工业级体系。这个判断比任何路线宣言都更接近现实，因为机器人真正难的地方，从来不是写出一个demo，而是让模型在不同场景里持续稳定地变强。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782207171715-e0nk.png\" alt=\"具身智能的钱正在往大脑集中，泡沫也在集中\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>这也是为什么“能不能快速赶上”不取决于今天站在哪条路线，而取决于有没有把数据管道、评测体系和训练闭环先搭好。技术架构会变，今天的世界模型明天可能被更新范式替代，但数据资产不会凭空出现。谁先把真实世界的数据采集、标注、回放、训练和评测做成流水线，谁就拥有跨架构迁移的能力。对创业公司来说，这比押注某个名词更重要。\u003C\u002Fp>\u003Ch2>第四，真正稀缺的是人，不是故事\u003C\u002Fh2>\u003Cp>量子位统计的35家“大脑派”公司中，17家的“一号位”来自高校或科研机构，几乎占到一半。清华和北大是最密集的人才来源，自动驾驶老兵也在大量转场。这说明资本并不是单纯在买概念，而是在买一群能把前沿算法落到真实系统里的人。具身智能还没收敛，路线可以争，但团队判断必须细，因为最后能把系统做出来的，往往是既懂算法又懂工程的人。\u003C\u002Fp>\u003Cp>年轻化同样是个明确信号。00后创始人、年轻PhD、放弃大厂高薪的研究型创业者，正在成为资本偏爱的对象。原因很现实：他们更贴近最新技术演进，路径依赖更少，沟通成本更低，也更敢于在不确定性里重构问题。相比之下，履历再漂亮的老兵，如果站在上一代技术范式上，也未必比年轻团队更有胜算。这个赛道看的是学习速度，不是资历厚度。\u003C\u002Fp>\u003Ch2>第五，泡沫不是问题，错把泡沫当终局才是问题\u003C\u002Fh2>\u003Cp>最强的反对意见很简单：具身智能现在估值太高、融资太快、路线太散，泡沫迟早会破。这个担心不是空穴来风。行业自己也承认，90%以上的公司可能会消失，而且泛化性如果不能兑现，真正可落地的场景会非常有限。与大模型相比，具身智能的商业化确定性更低，今天砸进去的钱，确实有相当一部分可能打水漂。\u003C\u002Fp>\u003Cp>但我不接受“因此现在不该投”的结论。早期产业发展需要泡沫来聚集资本、人才和注意力，这一点在大模型时代已经被验证过。不同的是，具身智能还没收敛到只剩几条清晰赛道，所以最优策略不是回避，而是承认它是一个高波动、高淘汰率的窗口期。问题不在于有没有泡沫，而在于你是否把泡沫当成确定性，把故事当成结果。\u003C\u002Fp>\u003Ch2>What to do with this\u003C\u002Fh2>\u003Cp>如果你是工程师，就别被“世界模型”三个字绑架，先把数据管道、评测集和闭环训练做扎实；如果你是PM，就盯住可复用场景和失败边界，不要把演示当产品；如果你是创始人，就把融资当作抢时间而不是讲故事，尽快建立能跨架构迁移的数据资产和人才密度。这个赛道最后赢的，不是最会喊口号的人，而是最快把不确定性工程化的人。\u003C\u002Fp>","我认为具身智能当前最值得押注的不是本体，而是大脑与数据基础设施。","zhuanlan.zhihu.com","https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F2052091107127906984",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782207170550-kt87.png","industry","en","d1c1656d-c842-49c2-a777-19ce87ab0dc8",[17,18,19,20,21],"具身智能","世界模型","VLA","机器人融资","数据基础设施",[23,24,25],"资本正在向具身智能“大脑派”集中，而不是本体派。","世界模型正在成为新叙事，但数据基础设施才是真正护城河。","这个赛道会经历高泡沫和高淘汰，胜负取决于团队与执行速度。",0,"2026-06-23T09:32:27.665876+00:00","2026-06-23T09:32:27.658+00:00","50ad070c-8891-4ccc-a7ee-038aa8918c86",{"tags":31,"relatedLang":32,"relatedPosts":35},[],{"id":15,"slug":33,"title":6,"language":34},"embodied-ai-capital-flows-to-brains-and-bubbles-zh","zh",[36,42,48,54,60,66],{"id":37,"slug":38,"title":39,"cover_image":40,"image_url":40,"created_at":41,"category":13},"44d464c9-107e-40bf-a83b-e690ce343d8a","microsoft-june-2026-partner-center-changes-en","Microsoft’s June 2026 Partner Center changes matter now","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782216175344-qtpv.png","2026-06-23T12:02:30.58867+00:00",{"id":43,"slug":44,"title":45,"cover_image":46,"image_url":46,"created_at":47,"category":13},"9cdee4fd-2408-479e-b362-bbaa7417aa33","sia-openai-partner-network-enterprise-ai-services-game-en","Sia’s OpenAI partnership is a sign enterprise AI is becoming a servic…","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782208971637-xvmg.png","2026-06-23T10:02:23.543505+00:00",{"id":49,"slug":50,"title":51,"cover_image":52,"image_url":52,"created_at":53,"category":13},"61775efa-14fe-427d-9c1e-5e321959e777","baya-openchip-bet-ai-silicon-data-movement-en","Baya and Openchip are betting the future of AI silicon on data moveme…","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782193668059-ugc5.png","2026-06-23T05:47:24.076812+00:00",{"id":55,"slug":56,"title":57,"cover_image":58,"image_url":58,"created_at":59,"category":13},"af91ca07-713f-4a59-88da-b375a50701b9","citigroup-sees-tokenized-assets-hitting-8-2t-en","Citigroup Sees Tokenized Assets Hitting $8.2T","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782185576141-ygv9.png","2026-06-23T03:32:33.87259+00:00",{"id":61,"slug":62,"title":63,"cover_image":64,"image_url":64,"created_at":65,"category":13},"e4f5babc-93c4-4b55-a8fb-d9c746d4c873","rwa-tokenization-turns-assets-into-on-chain-rails-en","RWA tokenization turns assets into on-chain rails","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782184698158-j904.png","2026-06-23T03:17:50.637664+00:00",{"id":67,"slug":68,"title":69,"cover_image":70,"image_url":70,"created_at":71,"category":13},"50270b29-033a-468f-a188-9163b77e0d0c","blackwell-mlperf-training-6-0-sweep-en","Blackwell’s MLPerf sweep shows why training speeds up","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782181966500-yj09.png","2026-06-23T02:32:26.093865+00:00",[73,78,83,88,93,98,103,108,113,118],{"id":74,"slug":75,"title":76,"created_at":77},"d35a1bd9-e709-412e-a2df-392df1dc572a","ai-impact-2026-developments-market-en","AI's Impact in 2026: Key Developments and Market Shifts","2026-03-25T16:20:33.205823+00:00",{"id":79,"slug":80,"title":81,"created_at":82},"5ed27921-5fd6-492e-8c59-78393bf37710","trumps-ai-legislative-framework-en","Trump's AI Legislative Framework: What's Inside?","2026-03-25T16:22:20.005325+00:00",{"id":84,"slug":85,"title":86,"created_at":87},"e454a642-f03c-4794-b185-5f651aebbaca","nvidia-gtc-2026-key-highlights-innovations-en","NVIDIA GTC 2026: Key Highlights and Innovations","2026-03-25T16:22:47.882615+00:00",{"id":89,"slug":90,"title":91,"created_at":92},"0ebb5b16-774a-4922-945d-5f2ce1df5a6d","claude-usage-diversifies-learning-curves-en","Claude Usage Diversifies, Learning Curves Emerge","2026-03-25T16:25:50.770376+00:00",{"id":94,"slug":95,"title":96,"created_at":97},"69934e86-2fc5-4280-8223-7b917a48ace8","openclaw-ai-commoditization-concerns-en","OpenClaw's Rise Raises Concerns of AI Model Commoditization","2026-03-25T16:26:30.582047+00:00",{"id":99,"slug":100,"title":101,"created_at":102},"b4b2575b-2ac8-46b2-b90e-ab1d7c060797","google-gemini-ai-rollout-2026-en","Google's Gemini AI Rollout Extended to 2026","2026-03-25T16:28:14.808842+00:00",{"id":104,"slug":105,"title":106,"created_at":107},"6e18bc65-42ae-4ad0-b564-67d7f66b979e","meta-llama4-fabricated-results-scandal-en","Meta's Llama 4 Scandal: Fabricated AI Test Results Unveiled","2026-03-25T16:29:15.482836+00:00",{"id":109,"slug":110,"title":111,"created_at":112},"bf888e9d-08be-4f47-996c-7b24b5ab3500","accenture-mistral-ai-deployment-en","Accenture and Mistral AI Team Up for AI Deployment","2026-03-25T16:31:01.894655+00:00",{"id":114,"slug":115,"title":116,"created_at":117},"5382b536-fad2-49c6-ac85-9eb2bae49f35","mistral-ai-high-stakes-2026-en","Mistral AI: Facing High Stakes in 2026","2026-03-25T16:31:39.941974+00:00",{"id":119,"slug":120,"title":121,"created_at":122},"9da3d2d6-b669-4971-ba1d-17fdb3548ed5","cursors-meteoric-rise-pressures-en","Cursor's Meteoric Rise Faces Industry Pressures","2026-03-25T16:32:21.899217+00:00"]