[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-anthropic-hiring-custom-chip-design-team-zh":3,"article-related-anthropic-hiring-custom-chip-design-team-zh":32,"series-industry-ff5c9d98-358f-49e4-9e5d-95fe8d7b05c7":79},{"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":25,"views":29,"created_at":30,"published_at":31,"topic_cluster_id":11},"ff5c9d98-358f-49e4-9e5d-95fe8d7b05c7","anthropic-hiring-custom-chip-design-team-zh","Anthropic 也要自己做晶片","\u003Cp data-speakable=\"summary\">\u003Ca href=\"\u002Ftag\u002Fanthropic\">Anthropic\u003C\u002Fa> 正在招募晶片設計團隊，想替 \u003Ca href=\"\u002Ftag\u002Fclaude\">Claude\u003C\u002Fa> 做客製化硬體。\u003C\u002Fp>\u003Cp>這件事很直白。AI 模型玩到後面，算力成本才是硬仗。\u003Ca href=\"https:\u002F\u002Fwww.anthropic.com\u002F\" target=\"_blank\" rel=\"noopener\">Anthropic\u003C\u002Fa> 已經把手伸進 \u003Ca href=\"https:\u002F\u002Ftechcrunch.com\u002F\" target=\"_blank\" rel=\"noopener\">TechCrunch\u003C\u002Fa> 提到的自研晶片領域，想把硬體和模型一起設計。\u003C\u002Fp>\u003Cp>這代表 Claude 不只要跑得快，還要跑得省。對一間靠模型吃飯的公司來說，這種動作很務實，也很貴。它同時碰到 \u003Ca href=\"https:\u002F\u002Faws.amazon.com\u002F\" target=\"_blank\" rel=\"noopener\">AWS\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fcloud.google.com\u002F\" target=\"_blank\" rel=\"noopener\">Google Cloud\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002F\" target=\"_blank\" rel=\"noopener\">Nvidia\u003C\u002Fa> 和 \u003Ca href=\"https:\u002F\u002Fwww.amd.com\u002F\" target=\"_blank\" rel=\"noopener\">AMD\u003C\u002Fa> 的地盤。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>項目\u003C\u002Fth>\u003Cth>內容\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>公司\u003C\u002Ftd>\u003Ctd>Anthropic\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>新團隊\u003C\u002Ftd>\u003Ctd>Custom silicon team\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>目標\u003C\u002Ftd>\u003Ctd>共同設計硬體與模型\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>公開確認\u003C\u002Ftd>\u003Ctd>2026 年 8 月 5 日\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>Anthropic 想把算力握回來\u003C\u002Fh2>\u003Cp>AI 公司最怕的，不是模型寫不出來，是每次推理都在燒錢。只要依賴外部晶片供應，價格、供貨、排程都要看別人臉色。\u003Ca href=\"https:\u002F\u002Fwww.anthropic.com\u002Fnews\" target=\"_blank\" rel=\"noopener\">Anthropic 的新聞頁\u003C\u002Fa> 近年幾乎都在講模型和產品，現在開始談硬體，方向很清楚。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786125773175-ek96.png\" alt=\"Anthropic 也要自己做晶片\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>自研晶片的價值，在於可以針對 Claude 的工作負載做調校。像記憶體頻寬、延遲、功耗，這些細節都會影響推理成本。模型越多人用，這些差異就越痛。\u003C\u002Fp>\u003Cp>如果每天要處理幾百萬次請求，省下 5% 到 10% 的推理成本，都可能變成真金白銀。對 Anthropic 這種還在擴張中的公司，這種節省會直接影響產品節奏。\u003C\u002Fp>\u003Cul>\u003Cli>Anthropic 已確認在招募自研晶片團隊。\u003C\u002Fli>\u003Cli>目標是共設計硬體與模型。\u003C\u002Fli>\u003Cli>目前仍依賴 AWS、Google、Nvidia、AMD。\u003C\u002Fli>\u003Cli>核心訴求是讓 Claude 更快，也更省。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>這波是整個 AI 硬體競賽的一部分\u003C\u002Fh2>\u003Cp>Anthropic 不是第一個往這裡走的。\u003Ca href=\"https:\u002F\u002Fopenai.com\u002F\" target=\"_blank\" rel=\"noopener\">OpenAI\u003C\u002Fa> 已經和 \u003Ca href=\"https:\u002F\u002Fwww.broadcom.com\u002F\" target=\"_blank\" rel=\"noopener\">Broadcom\u003C\u002Fa> \u003Ca href=\"\u002Fnews\u002Fmillennium-anthropic-ai-risk-analyst-zh\">合作\u003C\u002Fa>做自家晶片計畫，\u003Ca href=\"\u002Fnews\u002Fmodel-y-l-us-launch-buyer-details-zh\">重點\u003C\u002Fa>也放在 \u003Ca href=\"\u002Ftag\u002Finference\">inference\u003C\u002Fa>。\u003Ca href=\"https:\u002F\u002Fdeepmind.google\u002F\" target=\"_blank\" rel=\"noopener\">Google DeepMind\u003C\u002Fa> 長期靠 TPU 撐住模型服務，\u003Ca href=\"https:\u002F\u002Fabout.fb.com\u002F\" target=\"_blank\" rel=\"noopener\">Meta\u003C\u002Fa> 也在推 MTIA。\u003C\u002Fp>\u003Cp>這些動作放在一起看，訊號很一致。大型 AI 公司不想再把算力當成通用資源，而是想把硬體做成自己的工作流程。這樣一來，模型、伺服器、成本結構都能綁得更緊。\u003C\u002Fp>\u003Cblockquote>“We are building custom silicon to support our future AI systems,” said Dario Amodei, Anthropic’s co-founder and CEO, in a 2024 blog post about the company’s infrastructure plans.\u003C\u002Fblockquote>\u003Cp>而且這條路沒有想像中短。\u003Ca href=\"https:\u002F\u002Fwww.theinformation.com\u002F\" target=\"_blank\" rel=\"noopener\">The Information\u003C\u002Fa> 先前提到，Anthropic 也在找 \u003Ca href=\"https:\u002F\u002Fwww.samsung.com\u002F\" target=\"_blank\" rel=\"noopener\">Samsung\u003C\u002Fa> 看能不能合作製造晶片。這很合理，因為設計只是前半段，真正難的是把設計變成能量產的矽片。\u003C\u002Fp>\u003Ch2>招募資訊透露了 Anthropic 還在起步\u003C\u002Fh2>\u003Cp>目前看到的是招人，不是成品。這通常代表公司還在組架構、補人才、定\u003Ca href=\"\u002Fnews\u002F2027-tesla-model-y-l-exterior-photos-specs-zh\">規格\u003C\u002Fa>，離 tape-out 還有一段路。晶片專案很少有快的，尤其是 AI 公司第一次自己碰硬體。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786125773730-xm23.png\" alt=\"Anthropic 也要自己做晶片\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>要把晶片做出來，還得處理驗證、軟體堆疊、製程協調、供應鏈配合。每一步都燒錢，也都吃經驗。這也是為什麼很多 AI 團隊最後還是繼續買 \u003Ca href=\"\u002Ftag\u002Fnvidia\">Nvidia\u003C\u002Fa> GPU。\u003C\u002Fp>\u003Cp>如果你看這件事的實際意義，它比較像是 Anthropic 在先卡位。先把團隊建起來，先摸清楚 Claude 的需求，再決定要做多深。這種節奏比直接喊要自製整套伺服器，現實很多。\u003C\u002Fp>\u003Cul>\u003Cli>自研晶片團隊通常先做架構與驗證。\u003C\u002Fli>\u003Cli>之後才會進入 tape-out 與製造協調。\u003C\u002Fli>\u003Cli>Nvidia GPU 目前仍是 AI 服務主力。\u003C\u002Fli>\u003Cli>Google TPU 顯示客製化硬體能緊扣模型需求。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>對 Claude 使用者會有什麼差別\u003C\u002Fh2>\u003Cp>如果 Anthropic 做成了，Claude 可能會更便宜、更快，也更穩。這對企業客戶很重要，因為他們在乎 SLA，也在乎每個 token 的成本。對開發者來說，API 延遲下降，整個產品體驗會順很多。\u003C\u002Fp>\u003Cp>更實際的好處，是 Anthropic 能少受外部供應波動影響。當 GPU 缺貨或價格變動時，自己有一部分硬體選項，談判空間就會大一些。這種優勢很少出現在行銷文案裡，但財務報表會很誠實。\u003C\u002Fp>\u003Cp>我覺得 Anthropic 目前更像是在追求推理成本控制，而不是立刻變成晶片公司。它的重點應該是讓 Claude 的服務效率更好，先把可控性拉回來，再看要不要往更完整的硬體路線走。\u003C\u002Fp>\u003Ch2>AI 公司為何都開始碰硬體\u003C\u002Fh2>\u003Cp>原因其實很簡單。模型越大，推理越貴。訓練完只是開始，真正長期花錢的是上線之後的服務。當產品規模變大，硬體就不再只是採購項目，而是產品的一部分。\u003C\u002Fp>\u003Cp>這也是為什麼你會看到 \u003Ca href=\"\u002Ftag\u002Fopenai\">OpenAI\u003C\u002Fa>、Google、Meta 都在往自家晶片靠。誰能把模型和硬體綁得更緊，誰就更容易把成本壓下來。這不是抽象的技術理想，是很現實的商業算術。\u003C\u002Fp>\u003Cp>對台灣開發者來說，這件事也有意思。未來你選 LLM API，不只要看模型能力，還要看背後跑在哪種硬體上。延遲、價格、配額、穩定性，最後都會影響你做出來的產品。\u003C\u002Fp>\u003Ch2>接下來要看什麼\u003C\u002Fh2>\u003Cp>接下來最值得盯的，不是 Anthropic 會不會立刻發表晶片，而是它會先補哪些職缺。架構、後端、驗證、EDA、製造協作，這些關鍵字會告訴你它想做多深。\u003C\u002Fp>\u003Cp>如果後續真的出現合作製造夥伴，這件事就會從招募新聞變成硬體策略。到那時候，Claude 的競爭力就不只是模型本身，還包括它背後的算力效率。這會直接影響 API 價格和產品速度。\u003C\u002Fp>\u003Cp>我會繼續看兩件事：Anthropic 會不會公開更多硬體職缺，以及它會不會把自研晶片放進未來的基礎設施敘事。這兩個訊號一出來，方向就很清楚了。\u003C\u002Fp>","Anthropic 正在招募自研晶片團隊，想為 Claude 共設計客製化硬體，降低對外部 GPU 與雲端算力的依賴。","techcrunch.com","https:\u002F\u002Ftechcrunch.com\u002F2026\u002F08\u002F05\u002Fanthropic-is-hiring-an-ai-chip-design-team\u002F",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786125773175-ek96.png","industry","zh","2d0b3834-f02d-4b09-932d-eb0fda9f0c44",[17,18,19,20,21,22,23,24],"Anthropic","Claude","custom silicon","AI 晶片","推理成本","Nvidia","AWS","Google Cloud",[26,27,28],"Anthropic 正在招募自研晶片團隊，目標是替 Claude 共設計硬體與模型。","這代表 AI 公司的競爭已經從模型能力，延伸到硬體與推理成本。","短期看，Anthropic 更像是在先卡位，重點是控制成本與供應風險。",1,"2026-08-07T18:02:28.707393+00:00","2026-08-07T18:02:28.681+00:00",{"tags":33,"relatedLang":38,"relatedPosts":42},[34,36],{"name":17,"slug":35},"anthropic",{"name":18,"slug":37},"claude",{"id":15,"slug":39,"title":40,"language":41},"anthropic-hiring-custom-chip-design-team-en","Anthropic is hiring a custom chip design team","en",[43,49,55,61,67,73],{"id":44,"slug":45,"title":46,"cover_image":47,"image_url":47,"created_at":48,"category":13},"2bf5f0b0-b001-4301-9e6a-038713200705","webassembly-jvm-shift-java-portable-zh","5 個 WebAssembly 讓 JVM 更可攜的變化","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786129376756-a938.png","2026-08-07T19:02:29.468071+00:00",{"id":50,"slug":51,"title":52,"cover_image":53,"image_url":53,"created_at":54,"category":13},"96f0a314-ebfb-4213-992e-a9acd79dc301","model-y-l-us-launch-buyer-details-zh","Model Y L 美國首發，6 個買家重點","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786113169727-5kfd.png","2026-08-07T14:32:23.067484+00:00",{"id":56,"slug":57,"title":58,"cover_image":59,"image_url":59,"created_at":60,"category":13},"7f14f2f1-3346-47ed-96d2-e2aa698fd050","2027-tesla-model-y-l-exterior-photos-specs-zh","2027 Tesla Model Y L 外觀與規格解析","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786111389416-38tb.png","2026-08-07T14:02:37.558357+00:00",{"id":62,"slug":63,"title":64,"cover_image":65,"image_url":65,"created_at":66,"category":13},"3366150a-47a8-46a2-942b-b362626ad09a","millennium-anthropic-ai-risk-analyst-zh","Millennium 與 Anthropic 的風險分析合作，顯示 AI 正進…","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786104170460-28cy.png","2026-08-07T12:02:21.941565+00:00",{"id":68,"slug":69,"title":70,"cover_image":71,"image_url":71,"created_at":72,"category":13},"64f6010e-1a3b-4703-8d28-c0d6d6cb4baf","cuda-moat-tested-by-ai-coding-agents-zh","CUDA 的護城河，正被 AI 編碼代理測試","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786066369413-f836.png","2026-08-07T01:32:21.476706+00:00",{"id":74,"slug":75,"title":76,"cover_image":77,"image_url":77,"created_at":78,"category":13},"f0b42c04-fe18-4926-9698-5cf00ef69d59","rust-2026-updates-combat-base-play-zh","5 個改變 Rust 玩法的 2026 更新","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1786062763613-5s4p.png","2026-08-07T00:32:20.352886+00:00",[80,85,90,95,100,105,110,115,120,125],{"id":81,"slug":82,"title":83,"created_at":84},"ee073da7-28b3-4752-a319-5a501459fb87","ai-in-2026-what-actually-matters-now-zh","2026 AI 真正重要的事","2026-03-26T07:09:12.008134+00:00",{"id":86,"slug":87,"title":88,"created_at":89},"83bd1795-8548-44c9-9a7e-de50a0923f71","trump-ai-framework-power-speech-state-preemption-zh","川普 AI 框架瞄準電力、言論與州權","2026-03-26T07:12:18.695466+00:00",{"id":91,"slug":92,"title":93,"created_at":94},"ea6be18b-c903-4e54-97b7-5f7447a612e0","nvidia-gtc-2026-big-ai-announcements-zh","NVIDIA GTC 2026 重點拆解","2026-03-26T07:14:26.62638+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"4bcec76f-4c36-4daa-909f-54cd702f7c93","claude-users-spreading-out-and-getting-better-zh","Claude 用戶更分散，也更會用","2026-03-26T07:22:52.325888+00:00",{"id":101,"slug":102,"title":103,"created_at":104},"bd903b15-2473-4178-9789-b7557816e535","openclaw-raises-hard-question-for-ai-models-zh","OpenClaw 逼問 AI 模型價值","2026-03-26T07:24:54.707486+00:00",{"id":106,"slug":107,"title":108,"created_at":109},"eeac6b9e-ad9d-4831-8eec-8bba3f9bca6a","gap-google-gemini-checkout-fashion-search-zh","Gap 把結帳搬進 Gemini","2026-03-26T07:28:23.937768+00:00",{"id":111,"slug":112,"title":113,"created_at":114},"0740e53f-605d-4d57-8601-c10beb126f3c","google-pushes-gemini-transition-to-march-2026-zh","Google 把 Gemini 轉換延到 2026 年 3…","2026-03-26T07:30:12.825269+00:00",{"id":116,"slug":117,"title":118,"created_at":119},"e660d801-2421-4529-8fa9-86b82b066990","metas-llama-4-benchmark-scandal-gets-worse-zh","Meta Llama 4 分數風波又擴大","2026-03-26T07:34:21.156421+00:00",{"id":121,"slug":122,"title":123,"created_at":124},"183f9e7c-e143-40bb-a6d5-67ba84a3a8bc","accenture-mistral-ai-sovereign-enterprise-deal-zh","Accenture 攜手 Mistral AI 賣主權 AI","2026-03-26T07:38:14.818906+00:00",{"id":126,"slug":127,"title":128,"created_at":129},"191d9b1b-768a-478c-978c-dd7431a38149","mistral-ai-faces-its-hardest-year-yet-zh","Mistral AI 迎來最硬的一年","2026-03-26T07:40:23.716374+00:00"]