[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-openai-jalapeno-threatens-nvidia-realistically-zh":3,"article-related-openai-jalapeno-threatens-nvidia-realistically-zh":31,"series-model-release-76ca309d-9b8f-4595-a732-8cdb801b25e1":76},{"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":23,"views":27,"created_at":28,"published_at":29,"topic_cluster_id":30},"76ca309d-9b8f-4595-a732-8cdb801b25e1","openai-jalapeno-threatens-nvidia-realistically-zh","OpenAI自研芯片不是秀肌肉，而是英伟达的真实威胁","\u003Cp data-speakable=\"summary\">\u003Ca href=\"\u002Ftag\u002Fopenai\">OpenAI\u003C\u002Fa>的首颗自研推理芯片Jalapeño不是公关展示，而是英伟达定价权开始松动的信号。\u003C\u002Fp>\u003Cp>我认为，OpenAI这颗名为 Jalapeño 的自研芯片，不是一次秀肌肉，而是英伟达最该警惕的现实威胁。它从零到流片只用了九个月，目标也很明确：专门为大模型推理打造一颗“Intelligence Processor”，把最贵、最耗电、最依赖供应链的一段算力链条，直接收回到自己手里。\u003C\u002Fp>\u003Ch2>第一個論點：威脅首先來自議價權，而不是峰值性能\u003C\u002Fh2>\u003Cp>真正让英伟达难受的，不是 OpenAI 有没有做出一颗能跑模型的芯片，而是 OpenAI 开始拥有替代选项。只要推理工作负载的一部分从 \u003Ca href=\"\u002Ftag\u002Fgpu\">GPU\u003C\u002Fa> 迁移到自研 ASIC，英伟达就不再是唯一答案。对一个每年要烧掉巨量推理成本的公司来说，哪怕只把一小部分流量切出去，都足以在采购谈判里改变桌上的筹码。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782793064557-8hyd.png\" alt=\"OpenAI自研芯片不是秀肌肉，而是英伟达的真实威胁\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>这类变化在云厂商身上已经演过一遍。\u003Ca href=\"\u002Ftag\u002Faws\">AWS\u003C\u002Fa> 有 Graviton，\u003Ca href=\"\u002Ftag\u002Fgoogle\">Google\u003C\u002Fa> 有 TPU，微软也在推进自研加速器，逻辑都一样：先从最稳定、最可预测的负载下手，再慢慢扩大适用范围。OpenAI 现在做的不是“取代 GPU”，而是把 GPU 从绝对必要变\u003Ca href=\"\u002Fnews\u002Fdeno-29-desktop-apps-runtime-bet-zh\">成可\u003C\u002Fa>替代，这一步本身就足够危险。\u003C\u002Fp>\u003Ch2>第二個論點：推理經濟學比訓練競賽更能改寫市場\u003C\u002Fh2>\u003Cp>Jalapeño 的重点是推理，不是训练，这一点非常关键。训练前沿模型需要极端灵活的并行能力和成熟的软件生态，GPU 仍然占优；但推理更看重单位成本、吞吐、延迟和功耗。大模型真正的商业化压力，往往不是训练一次要多少钱，而是上线后每一次回答、每一次检索、每一次工具调用都在持续消耗算力。\u003C\u002Fp>\u003Cp>如果 OpenAI 能把推理成本压下来，它得到的不是技术新闻，而是产品利润。面向海量用户的聊天、摘要、检索增强和代理调用，都是高频推理场景。谁能把每千次请求的成本降下去，谁就能把更低价格、更高毛利和更激进的产品策略同时拿到手。英伟达卖的是通用算力，OpenAI 要的是把算力变成可控的产品成本。\u003C\u002Fp>\u003Ch2>第三個論點：供應鏈控制權本身就是戰略資產\u003C\u002Fh2>\u003Cp>九个月从白纸到流片，这个速度说明 OpenAI 已经不满足于“买现成的最强芯片”，而是在搭建自己的基础设施主权。对前沿模型公司来说，最大风险从来不只是性能落后，而是供货、配额、交付周期和地缘政治\u003Ca href=\"\u002Fnews\u002Fanthropic-export-ban-shift-changes-ai-access-zh\">限制\u003C\u002Fa>。自研芯片的意义之一，就是把关键能力从外部供应商的排期表里拿回来。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782793064123-8cmd.png\" alt=\"OpenAI自研芯片不是秀肌肉，而是英伟达的真实威胁\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>这件事在规模上会越来越重要。AI 公司一旦进入大规模服务阶段，芯片不是一次性采购，而是持续扩容、持续替换、持续优化的资产。自研 ASIC 哪怕只覆盖一部分推理集群，也能让公司在产能紧张、出口限制和价格波动时保留缓冲。英伟达最怕的不是一个客户少买几块卡，而是大客户开始把未来算力规划写进自己的芯片路线图。\u003C\u002Fp>\u003Ch2>反方可能怎麼說\u003C\u002Fh2>\u003Cp>反对者会说，OpenAI 这件事被夸大了。ASIC 确实能在特定任务上更便宜、更高效，但它也更窄、更难迭代。大模型系统变化太快，今天流行的是某种注意力结构，明天就可能换成别的推理范式。GPU 的优势就在于通用性和软件生态，\u003Ca href=\"\u002Ftag\u002Fcuda\">CUDA\u003C\u002Fa> 和成熟开发工具链不是一颗新芯片三个月、九个月就能补齐的。\u003C\u002Fp>\u003Cp>还有一个现实问题：做芯片不等于做成芯片生意。流片只是开始，真正难的是良率、封装、供电、散热、驱动、编译器和大规模部署。很多自研硬件项目都在“能跑”和“能规模化赚钱”之间折戟。站在这个角度看，Jalapeño 更像是一枚试探性的棋子，而不是立即改写行业格局的终局武器。\u003C\u002Fp>\u003Cp>但这个反驳只成立一半。因为 OpenAI 并不需要用 Jalapeño 全面打败 GPU，它只需要在推理这个最赚钱、最重复、最稳定的环节里拿到结构性优势。只要它证明自研芯片能降低成本、稳定供给并支撑真实业务，英伟达的护城河就会从“不可替代”变成“仍然强，\u003Ca href=\"\u002Fnews\u002Fsora-smash-ultimate-final-dlc-pick-balanced-zh\">但不\u003C\u002Fa>再绝对”。这已经足够让市场重新定价。\u003C\u002Fp>\u003Ch2>你能做什麼\u003C\u002Fh2>\u003Cp>如果你是工程师，别把这件事理解成“芯片新闻”，而要理解成系统设计信号：未来的大模型栈会越来越垂直整合，模型、编译器、推理引擎和硬件会一起优化。如果你是 PM，优先盯住推理成本、延迟和单位请求毛利，因为真正决定产品能否扩张的，不是模型参数，而是每次调用的经济账。如果你是创始人，这条新闻的启示更直接：当你的核心业务足够大时，供应商不会永远是供应商，你必须尽早把关键依赖变成自己的能力。\u003C\u002Fp>","OpenAI的首颗自研推理芯片Jalapeño不是公关展示，而是英伟达定价权开始松动的信号，因為推理成本、供應鏈與議價權都在被重新分配。","zhuanlan.zhihu.com","https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F2053473940031460150",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782793064557-8hyd.png","model-release","zh","81c51b29-6f78-43bb-a264-e6b208644d4f",[17,18,19,20,21,22],"OpenAI","Jalapeño","英伟达","自研芯片","推理成本","议价权",[24,25,26],"OpenAI 自研芯片的核心威胁在于推理负载替代，而不是训练性能对决。","一旦推理成本下降，OpenAI 就能同时获得更高毛利、更低价格和更强议价权。","自研芯片的战略价值还在供应链控制，能降低交付、配额与地缘风险。",0,"2026-06-30T04:17:21.527935+00:00","2026-06-30T04:17:21.5+00:00","0ccb5d2e-69f1-4354-a3e0-cb370221cd95",{"tags":32,"relatedLang":35,"relatedPosts":39},[33],{"name":17,"slug":34},"openai",{"id":15,"slug":36,"title":37,"language":38},"openai-jalapeno-threatens-nvidia-realistically-en","OpenAI自研芯片不是秀肌肉，而是对英伟达的真实威胁","en",[40,46,52,58,64,70],{"id":41,"slug":42,"title":43,"cover_image":44,"image_url":44,"created_at":45,"category":13},"4c695cb3-da7c-494f-b36f-d184efae3cf0","anthropic-claude-tag-slack-ai-coworker-zh","Claude Tag 把 Slack 變成 AI 同事","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782802983792-5qek.png","2026-06-30T07:02:32.884446+00:00",{"id":47,"slug":48,"title":49,"cover_image":50,"image_url":50,"created_at":51,"category":13},"6ec6c861-8bc8-4d40-a0bd-b46389e31030","k3s-v1-34-9-kubernetes-1-34-9-release-zh","K3s v1.34.9 更新重點整理","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782781393060-r0zj.png","2026-06-30T01:02:51.491478+00:00",{"id":53,"slug":54,"title":55,"cover_image":56,"image_url":56,"created_at":57,"category":13},"cfb68e08-fe4e-49f6-b449-e566faf56311","kimi-2-7-price-coding-benchmark-zh","Kimi 2.7 讓價格成為真正的寫碼基準","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782746270225-tcs9.png","2026-06-29T15:17:24.321277+00:00",{"id":59,"slug":60,"title":61,"cover_image":62,"image_url":62,"created_at":63,"category":13},"ca1e6960-10e7-4fa7-949f-c5991c99fc7e","kimi-k26-open-source-coding-agentic-ai-benchmarks-zh","Kimi K2.6 登頂程式與代理式 AI 基準","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782739078329-qvne.png","2026-06-29T13:17:26.530857+00:00",{"id":65,"slug":66,"title":67,"cover_image":68,"image_url":68,"created_at":69,"category":13},"edf8e66b-c717-4cc1-b15a-96839bb7bbcf","llama-legends-380-season-3-heroes-raids-zh","Llama Legends 3.8.0 推出 Season 3 英雄與突襲","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782711179415-qurv.png","2026-06-29T05:32:32.733919+00:00",{"id":71,"slug":72,"title":73,"cover_image":74,"image_url":74,"created_at":75,"category":13},"88d353ca-468b-4774-922d-ef0cbc2edd68","omlx-045-dev1-glm52-minimax-m3-speedups-zh","oMLX 0.4.5.dev1 讓長上下文更快","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782709372375-25nm.png","2026-06-29T05:02:28.341041+00:00",[77,82,87,92,97,102,107,112,117,122],{"id":78,"slug":79,"title":80,"created_at":81},"58b64033-7eb6-49b9-9aab-01cf8ae1b2f2","nvidia-rubin-six-chips-one-ai-supercomputer-zh","NVIDIA Rubin 把六顆晶片塞進 AI 機櫃","2026-03-26T07:18:45.861277+00:00",{"id":83,"slug":84,"title":85,"created_at":86},"0dcc2c61-c2a6-480d-adb8-dd225fc68914","march-2026-ai-model-news-what-mattered-zh","2026 年 3 月 AI 模型新聞重點","2026-03-26T07:32:08.386348+00:00",{"id":88,"slug":89,"title":90,"created_at":91},"214ab08b-5ce5-4b5c-8b72-47619d8675dd","why-small-models-are-winning-on-device-ai-zh","小模型為何吃下裝置端 AI","2026-03-26T07:36:30.488966+00:00",{"id":93,"slug":94,"title":95,"created_at":96},"785624b2-0355-4b82-adc3-de5e45eecd88","midjourney-v8-faster-images-higher-costs-zh","Midjourney V8 變快了，也變貴了","2026-03-26T07:52:03.562971+00:00",{"id":98,"slug":99,"title":100,"created_at":101},"cda76b92-d209-4134-86c1-a60f5bc7b128","xiaomi-mimo-trio-agents-robots-voice-zh","小米 MiMo 三模型瞄準代理、機器人與語音","2026-03-28T03:05:08.779489+00:00",{"id":103,"slug":104,"title":105,"created_at":106},"9e1044b4-946d-47fe-9e2a-c2ee032e1164","xiaomi-mimo-v2-pro-1t-moe-agents-zh","小米 MiMo-V2-Pro 登場：1T MoE 模型","2026-03-28T03:06:19.002353+00:00",{"id":108,"slug":109,"title":110,"created_at":111},"c4b6186f-bd84-4598-997e-c6e31d543c0d","cursor-composer-2-agentic-coding-model-zh","Cursor Composer 2 走向代理式寫碼","2026-03-28T03:13:06.422716+00:00",{"id":113,"slug":114,"title":115,"created_at":116},"e112e76f-ec3b-408f-810e-e93ae21a888a","apple-siri-gemini-distilled-models-zh","Apple Siri 牽手 Gemini 的真相","2026-03-29T04:52:57.886544+00:00",{"id":118,"slug":119,"title":120,"created_at":121},"c679b51f-194a-463b-87fc-7695256ff752","mimo-v2-pro-vs-omni-vs-flash-2026-zh","MiMo V2 Pro、Omni、Flash 怎麼選","2026-04-02T01:18:43.576128+00:00",{"id":123,"slug":124,"title":125,"created_at":126},"3b988fd7-6749-4f01-ba25-c0ad7486dc31","z-ai-glm-5v-turbo-design2code-claude-zh","GLM-5V-Turbo 在 Design2Code 贏了…","2026-04-02T04:03:36.31741+00:00"]