[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-openai-122b-ai-infrastructure-push-zh":3,"tags-openai-122b-ai-infrastructure-push-zh":32,"related-lang-openai-122b-ai-infrastructure-push-zh":43,"related-posts-openai-122b-ai-infrastructure-push-zh":47,"series-industry-3c8bbed5-8bb2-479e-b9d7-a85b44ae6add":84},{"id":4,"title":5,"content":6,"summary":7,"source":8,"source_url":9,"author":10,"image_url":11,"keywords":12,"language":20,"translated_content":10,"views":21,"is_premium":22,"created_at":23,"updated_at":23,"cover_image":11,"published_at":24,"rewrite_status":25,"rewrite_error":10,"rewritten_from_id":26,"slug":27,"category":28,"related_article_id":29,"status":30,"google_indexed_at":31,"x_posted_at":10,"tweet_text":10,"title_rewritten_at":10,"title_original":10,"key_takeaways":10,"topic_cluster_id":10,"embedding":10,"is_canonical_seed":22},"3c8bbed5-8bb2-479e-b9d7-a85b44ae6add","OpenAI 1220 億美元拚 AI 基礎設施","\u003Cp>\u003Ca href=\"https:\u002F\u002Fopenai.com\" target=\"_blank\" rel=\"noopener\">OpenAI\u003C\u002Fa> 這次不是只談模型。它把戰線拉到雲端、晶片、資料中心。數字很直接，外媒估算這波投入上看 1220 億美元。說真的，這已經不是一般軟體公司的玩法。\u003C\u002Fp>\u003Cp>你可能會想問，為什麼要燒到這個程度。答案很簡單。AI 現在最缺的，不是 demo，而是算力、電力和機櫃。沒有這三樣，LLM 再強也只能卡在簡報裡。\u003C\u002Fp>\u003Cp>OpenAI 的做法很像在蓋一條自己的供應鏈。它不想只靠單一雲端，也不想只押單一晶片廠。這種分散配置，看起來很麻煩，但在 AI 時代，麻煩常常比斷供便宜。\u003C\u002Fp>\u003Ch2>OpenAI 在蓋什麼\u003C\u002Fh2>\u003Cp>這次最有意思的地方，是它把整個 AI \u003Ca href=\"\u002Fnews\u002Frust-1-94-1-patches-regressions-and-cargo-cves-zh\">st\u003C\u002Fa>ack 都攤開了。從雲端算力，到 GPU 和 ASIC，再到資料中心落地，幾乎每一層都有人接手。這種打法很像超大規模雲端廠商的思路，只是 OpenAI 現在自己也在往那個位置走。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775153037357-jq5f.png\" alt=\"OpenAI 1220 億美元拚 AI 基礎設施\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>OpenAI 提到的合作對象很多。雲端有 \u003Ca href=\"https:\u002F\u002Fwww.microsoft.com\" target=\"_blank\" rel=\"noopener\">Microsoft\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fwww.oracle.com\" target=\"_blank\" rel=\"noopener\">Oracle\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Faws.amazon.com\" target=\"_blank\" rel=\"noopener\">AWS\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fwww.coreweave.com\" target=\"_blank\" rel=\"noopener\">CoreWeave\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fcloud.google.com\" target=\"_blank\" rel=\"noopener\">Google Cloud\u003C\u002Fa>。晶片有 \u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\" target=\"_blank\" rel=\"noopener\">NVIDIA\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fwww.amd.com\" target=\"_blank\" rel=\"noopener\">AMD\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Faws.amazon.com\u002Fec2\u002Finstance-types\u002Ftrn1\u002F\" target=\"_blank\" rel=\"noopener\">AWS Trainium\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fwww.cerebras.net\" target=\"_blank\" rel=\"noopener\">Cerebras\u003C\u002Fa>。自研晶片則找上 \u003Ca href=\"https:\u002F\u002Fwww.broadcom.com\" target=\"_blank\" rel=\"noopener\">Broadcom\u003C\u002Fa>。\u003C\u002Fp>\u003Cp>講白了，這就是備援思維。單一雲端掛掉，還能切到別家。單一晶片供應吃緊，還有其他路。這對訓練模型很重要，對推論服務更重要。因為使用者一多，API 延遲和排隊時間就會直接爆表。\u003C\u002Fp>\u003Cul>\u003Cli>雲端供應商至少 5 家\u003C\u002Fli>\u003Cli>晶片來源至少 5 條路線\u003C\u002Fli>\u003Cli>資料中心合作也同步擴張\u003C\u002Fli>\u003Cli>自研晶片開始進場\u003C\u002Fli>\u003C\u002Ful>\u003Cp>這種配置不是炫耀名單而已。它代表 OpenAI 想把風險拆開。你可以把它想成一個很貴的保險組合。每個環節都要錢，但少掉任何一環，整個 AI 服務就可能卡住。\u003C\u002Fp>\u003Cp>對開發者來說，這也很現實。你在前端看到的是一個 API。背後其實是一整串伺服器、網路、機櫃和電力協調。模型越大，這些東西越不是背景板，而是產品本體的一部分。\u003C\u002Fp>\u003Ch2>為什麼這麼燒錢\u003C\u002Fh2>\u003Cp>AI 的成本結構已經變了。以前做軟體，主要是工程師和伺服器費用。現在做 LLM，還要先搶晶片，再搶電，再搶資料中心空間。這不是比誰寫 c\u003Ca href=\"\u002Fnews\u002Fvs-code-cursor-windsurf-jetbrains-web-ides-2026-zh\">ode\u003C\u002Fa> 快，而是比誰先把供應鏈卡位好。\u003C\u002Fp>\u003Cp>OpenAI 這種玩法，背後其實是稀缺資源競爭。NVIDIA GPU 不是你想買就買得到。高密度機櫃不是你想蓋就蓋得出來。連電網接入都可能拖上好幾個月。這些都會直接影響模型訓練排程。\u003C\u002Fp>\u003Cp>Sam Altman 在 OpenAI 的官方文章裡說過一句話：\u003Ca href=\"https:\u002F\u002Fopenai.com\u002Findex\u002Faccelerating-the-next-phase-ai\u002F\" target=\"_blank\" rel=\"noopener\">“We are in the middle of an unprecedented infrastructure buildout.”\u003C\u002Fa> 我覺得這句話很直白，也很誠實。講白了就是，AI 現在比的是誰能把基礎設施先蓋起來。\u003C\u002Fp>\u003Cblockquote>“We are in the middle of an unprecedented infrastructure buildout.” — Sam Altman\u003C\u002Fblockquote>\u003Cp>你如果看訓練成本，就更懂這件事。前沿模型常常要吃掉成千上萬顆加速器。推論端又是另一個世界。每天幾百萬次請求進來，Token 成本、延遲和併發都會互相拉扯。\u003C\u002Fp>\u003Cp>所以 OpenAI 的策略不是單點加碼，而是整條線一起補。這樣做很貴，但也比較不會被單一供應商掐住脖子。對一家想維持高頻迭代的 AI 公司來說，這種自主性很值錢。\u003C\u002Fp>\u003Ch2>和其他 AI 玩家比起來\u003C\u002Fh2>\u003Cp>OpenAI 不是唯一在搶算力的公司，但它的佈局很廣。很多 AI 團隊只押一家雲端，或只靠一種晶片。OpenAI 則是在多雲、多晶片、多資料中心之間做平衡。這種做法更像大型雲端廠，而不是一般新創。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775153036338-k9ah.png\" alt=\"OpenAI 1220 億美元拚 AI 基礎設施\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>當然，代價也不小。合作夥伴越多，整合越麻煩。合約越多，協調成本越高。不同雲端的網路、儲存、監控工具都不一樣。對工程團隊來說，這意味著更多維運工作，也意味著更多踩雷機會。\u003C\u002Fp>\u003Cp>但如果你把視角拉高，這些麻煩都算小事。真正可怕的是沒有算力。沒有算力，就沒有訓練。沒有訓練，就沒有新模型。沒有新模型，產品就只能吃老本。\u003C\u002Fp>\u003Cul>\u003Cli>\u003Ca href=\"https:\u002F\u002Fwww.microsoft.com\u002Fen-us\u002Fai\" target=\"_blank\" rel=\"noopener\">Microsoft\u003C\u002Fa> 是 OpenAI 最深的雲端合作夥伴之一\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Faws.amazon.com\" target=\"_blank\" rel=\"noopener\">AWS\u003C\u002Fa> 和 \u003Ca href=\"https:\u002F\u002Fcloud.google.com\" target=\"_blank\" rel=\"noopener\">Google Cloud\u003C\u002Fa> 讓 OpenAI 有更多雲端分流空間\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\" target=\"_blank\" rel=\"noopener\">NVIDIA\u003C\u002Fa> 仍是主流 AI 算力核心\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fwww.amd.com\" target=\"_blank\" rel=\"noopener\">AMD\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Faws.amazon.com\u002Fec2\u002Finstance-types\u002Ftrn1\u002F\" target=\"_blank\" rel=\"noopener\">Trainium\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fwww.cerebras.net\" target=\"_blank\" rel=\"noopener\">Cerebras\u003C\u002Fa> 提供替代路線\u003C\u002Fli>\u003C\u002Ful>\u003Cp>我覺得最值得看的，是 \u003Ca href=\"https:\u002F\u002Fwww.broadcom.com\" target=\"_blank\" rel=\"noopener\">Broadcom\u003C\u002Fa> 這條線。自研晶片這件事，已經不是大廠的專利。只要你算得夠多、規模夠大，自己設計 ASIC 就可能比通用 GPU 更划算。這也是 Google、Amazon 一直在做的事。\u003C\u002Fp>\u003Cp>換句話說，AI 競爭正在往基礎設施戰移動。模型能力當然重要，但如果你連穩定供電和足夠機櫃都拿不到，產品就只能停在半路。這點很殘酷，但很真實。\u003C\u002Fp>\u003Ch2>這代表產業開始怎麼變\u003C\u002Fh2>\u003Cp>OpenAI 這種大手筆，會把整個產業的標準往上拉。新創公司以前只要會做模型、會包 API 就能講故事。現在不行了。投資人和\u003Ca href=\"\u002Fnews\u002Fwhat-agentic-workflows-actually-do-enterprise-ai-zh\">企業\u003C\u002Fa>客戶都會追問：你的算力從哪來？成本怎麼控？斷線時怎麼備援？\u003C\u002Fp>\u003Cp>這也會影響產品設計。很多團隊會開始算每個功能吃多少 Token。也會開始看推論延遲、快取命中率、區域部署和模型路由。以前這些是後台細節，現在直接變成商業指標。\u003C\u002Fp>\u003Cp>對台灣團隊來說，這件事也很有感。因為台灣強項本來就在硬體、伺服器、機櫃、散熱和供應鏈。AI 時代一來，這些能力突然變成核心資產。不是只有寫模型的人重要，做機房和電力規劃的人也很關鍵。\u003C\u002Fp>\u003Cp>從更大的脈絡看，AI 產業已經走到一個很現實的階段。比起誰的 demo 比較炫，市場更在意誰能長期供貨。這個轉變很無聊嗎？有點。但商業世界本來就這樣，能穩定交付的人，通常比較能活下來。\u003C\u002Fp>\u003Ch2>接下來該看什麼\u003C\u002Fh2>\u003Cp>我會先盯三件事。第一，OpenAI 的雲端分散是否真的落地。第二，自研晶片進度有沒有明確時間表。第三，資料中心和電力是否跟得上。這三項只要卡一項，整個計畫就會慢下來。\u003C\u002Fp>\u003Cp>如果你是開發者，現在很適合重新檢查自己的 AI 架構。不要只看模型效果。也要看供應商風險、成本曲線、區域延遲和備援方案。AI 不是只有 prompt engineering，還有一大堆基礎設施工程。\u003C\u002Fp>\u003Cp>我的判斷很直接：接下來兩年，AI 公司的差距會越來越像雲端公司的差距。誰有更多算力、更多電力、更多供應鏈彈性，誰就更能撐住產品節奏。OpenAI 這次把這件事講得很清楚。\u003C\u002Fp>\u003Cp>你如果正在做 AI 產品，現在就該問自己一句：你的服務，能不能在單一雲端故障時還活著？\u003C\u002Fp>","OpenAI 把 AI 堆疊拉到雲端、晶片和資料中心。它同時牽手 Microsoft、Oracle、AWS、NVIDIA、AMD、Broadcom，目標是把算力供應鏈做成多路備援。","openai.com","https:\u002F\u002Fopenai.com\u002Findex\u002Faccelerating-the-next-phase-ai\u002F",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775153037357-jq5f.png",[13,14,15,16,17,18,19],"OpenAI","AI 基礎設施","雲端算力","NVIDIA","Broadcom","資料中心","人工智慧","zh",0,false,"2026-04-02T18:03:38.772299+00:00","2026-04-02T18:03:38.628+00:00","done","04cc279f-4a7b-4809-9b1c-da0c1d842cf5","openai-122b-ai-infrastructure-push-zh","industry","849656c6-cf38-484e-bddf-b615da1cd408","published","2026-04-08T09:00:50.043+00:00",[33,35,36,39,41,42],{"name":13,"slug":34},"openai",{"name":19,"slug":19},{"name":37,"slug":38},"Nvidia","nvidia",{"name":17,"slug":40},"broadcom",{"name":18,"slug":18},{"name":15,"slug":15},{"id":29,"slug":44,"title":45,"language":46},"openai-122b-ai-infrastructure-push-en","OpenAI’s $122B push for AI infrastructure","en",[48,54,60,66,72,78],{"id":49,"slug":50,"title":51,"cover_image":52,"image_url":52,"created_at":53,"category":28},"cd078ce9-0a92-485a-b428-2f5523250a19","circles-agent-stack-targets-machine-speed-payments-zh","Circle 推出 Agent Stack，瞄準機器速度支付","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778871663628-uyk5.png","2026-05-15T19:00:44.16849+00:00",{"id":55,"slug":56,"title":57,"cover_image":58,"image_url":58,"created_at":59,"category":28},"96d96399-f674-4269-997a-cddfc34291a0","iren-signs-nvidia-ai-infrastructure-pact-zh","IREN 綁上 Nvidia AI 基建","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778871057561-bukp.png","2026-05-15T18:50:37.57206+00:00",{"id":61,"slug":62,"title":63,"cover_image":64,"image_url":64,"created_at":65,"category":28},"de12a36e-52f9-4bca-8deb-a41cf974ffd9","circle-agent-stack-ai-payments-zh","Circle 推出 Agent Stack 做 AI 付款","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778870462187-t9xv.png","2026-05-15T18:40:30.945394+00:00",{"id":67,"slug":68,"title":69,"cover_image":70,"image_url":70,"created_at":71,"category":28},"e6379f8a-3305-4862-bd15-1192d3247841","why-nebius-ai-pivot-is-more-real-than-hype-zh","為什麼 Nebius 的 AI 轉型比炒作更真實","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778823044520-9mfz.png","2026-05-15T05:30:24.978992+00:00",{"id":73,"slug":74,"title":75,"cover_image":76,"image_url":76,"created_at":77,"category":28},"66c4e357-d84d-43ef-a2e7-120c4609e98e","nvidia-backs-corning-factories-with-billions-zh","Nvidia 出資 Corning 工廠擴產","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778822450270-trdb.png","2026-05-15T05:20:27.701475+00:00",{"id":79,"slug":80,"title":81,"cover_image":82,"image_url":82,"created_at":83,"category":28},"31d8109c-8b0b-46e2-86bc-d274a03269d1","why-anthropic-gates-foundation-ai-public-goods-zh","為什麼 Anthropic 和 Gates Foundation 應該投資 A…","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778796636474-u508.png","2026-05-14T22:10:21.138177+00:00",[85,90,95,100,105,110,115,120,125,130],{"id":86,"slug":87,"title":88,"created_at":89},"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":91,"slug":92,"title":93,"created_at":94},"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":96,"slug":97,"title":98,"created_at":99},"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":101,"slug":102,"title":103,"created_at":104},"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":106,"slug":107,"title":108,"created_at":109},"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":111,"slug":112,"title":113,"created_at":114},"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":116,"slug":117,"title":118,"created_at":119},"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":121,"slug":122,"title":123,"created_at":124},"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":126,"slug":127,"title":128,"created_at":129},"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":131,"slug":132,"title":133,"created_at":134},"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"]