[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-sora-shutdown-ai-vendor-risk-zh":3,"tags-sora-shutdown-ai-vendor-risk-zh":33,"related-lang-sora-shutdown-ai-vendor-risk-zh":46,"related-posts-sora-shutdown-ai-vendor-risk-zh":50,"series-industry-7b92e9d9-c122-4e67-8e40-5f32817817fd":87},{"id":4,"title":5,"content":6,"summary":7,"source":8,"source_url":9,"author":10,"image_url":11,"keywords":12,"language":21,"translated_content":10,"views":22,"is_premium":23,"created_at":24,"updated_at":24,"cover_image":11,"published_at":25,"rewrite_status":26,"rewrite_error":10,"rewritten_from_id":27,"slug":28,"category":29,"related_article_id":30,"status":31,"google_indexed_at":32,"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":23},"7b92e9d9-c122-4e67-8e40-5f32817817fd","Sora 停運背後的 AI 供應商風險","\u003Cp>\u003Ca href=\"https:\u002F\u002Fopenai.com\u002Fsora\" target=\"_blank\" rel=\"noopener\">Sora\u003C\u002Fa> 曾衝到約 100 萬用戶。後來又掉到 50 萬以下。更狠的是，它每天大概燒掉 100 萬美元算力。講白了，這種帳一算下去，產品再酷也會出事。\u003C\u002Fp>\u003Cp>這件事很適合拿來看 \u003Ca href=\"\u002Fnews\u002Fai-pc-build-budget-config-guide-zh\">AI\u003C\u002Fa> 產品的死穴。模型可以很會秀。用戶也可能一開始很買單。但如果每次生成都很貴，流量又留不住，商業模式就會直接裂開。\u003C\u002Fp>\u003Cp>我覺得這不是單一產品失敗。這是整個 AI 供應鏈都會碰到的問題。你買的是模型能力，但付的是\u003Ca href=\"\u002Fnews\u002Falibaba-risc-v-ai-cpu-server-chips-zh\">伺服器\u003C\u002Fa>和 GPU 的帳單。\u003C\u002Fp>\u003Ch2>先看 Sora 為何撐不住\u003C\u002Fh2>\u003Cp>先講最現實的數字。100 萬用戶聽起來很多。可是一旦掉到 50 萬以下，經濟模型就會變得很難看。尤其是影片生成這種工作，算力成本本來就高。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775198567172-1cmi.png\" alt=\"Sora 停運背後的 AI 供應商風險\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>影片不是聊天。聊天回一段字，成本還能壓。影片生成要跑更大的模型，還要吃更長的推理時間。每多一分鐘，都是真金白銀在燒。\u003C\u002Fp>\u003Cp>問題不只在成本高。問題還在用戶不一定會常回來。很多人是衝著新鮮感來試一次。拍幾支影片、分享一下、然後就放著不用。這種產品很容易爆紅，但很難養成日常使用。\u003C\u002Fp>\u003Cul>\u003Cli>峰值用戶：約 100 萬\u003C\u002Fli>\u003Cli>後期用戶：低於 50 萬\u003C\u002Fli>\u003Cli>每日算力燒錢：約 100 萬美元\u003C\u002Fli>\u003Cli>主要工作負載：影片生成，AI 裡最貴的類型之一\u003C\u002Fli>\u003C\u002Ful>\u003Cp>這裡的重點很簡單。只要使用頻率不夠高，單次成本又太重，營收就很難追上支出。就算你有很漂亮的 demo，也不代表這個產品能活。\u003C\u002Fp>\u003Cp>而且這種問題通常不是慢慢壞掉。它會突然爆。前面看起來像成長，後面一算帳，才發現每個用戶都在拖累毛利。\u003C\u002Fp>\u003Ch2>AI 供應商為什麼常被打臉\u003C\u002Fh2>\u003Cp>這種事不只發生在 S\u003Ca href=\"\u002Fnews\u002Fcloudflare-emdash-wordpress-successor-zh\">or\u003C\u002Fa>a。很多 AI 供應商都會遇到同一個坑。你賣的是 API、訂閱，或是某種使用額度。但真正付錢給雲端和晶片廠的，是你自己。\u003C\u002Fp>\u003Cp>這個落差很要命。因為收入通常是固定的，成本卻跟著實際使用量跑。只要用戶一多，或每次請求一重，帳單就會往上飛。\u003C\u002Fp>\u003Cp>再來是產品型態的差別。聊天工具可以靠高頻使用撐營收。影片工具卻常常是低頻、重度、偶發。這代表留存率和變現方式都更難做。\u003C\u002Fp>\u003Cp>像 \u003Ca href=\"https:\u002F\u002Fwww.anthropic.com\" target=\"_blank\" rel=\"noopener\">Anthropic\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fwww.openai.com\" target=\"_blank\" rel=\"noopener\">OpenAI\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fgemini.google\u002F\" target=\"_blank\" rel=\"noopener\">Google Gemini\u003C\u002Fa> 這類產品，都得一直算推理成本。只是影片比文字更兇。文字像小吃店。影片像開火鍋店。不是不能做，是成本結構完全不同。\u003C\u002Fp>\u003Cblockquote>“Every cloud has a silver lining.” — Anais Nin\u003C\u002Fblockquote>\u003Cp>這句話很適合拿來看 AI 產品。表面上是成長。背後可能是成本炸裂。你看到的是漂亮畫面，財務看到的是 GPU 帳單。\u003C\u002Fp>\u003Cp>所以供應商真正該管的，不是只有模型準不準。還要管每次輸出到底花多少錢。這才是能不能活下去的核心。\u003C\u002Fp>\u003Ch2>跟便宜 AI 產品比，差很多\u003C\u002Fh2>\u003Cp>如果拿 Sora 跟文字型 AI 比，差距很直觀。文字回覆通常很便宜。就算是長文，也比影片生成省太多。這也是為什麼很多公司先做聊天，再碰影片。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775198567434-7bzn.png\" alt=\"Sora 停運背後的 AI 供應商風險\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>圖片生成也比影片好養。雖然圖片比文字貴，但至少還能控制在可接受範圍。影片就不一樣了。它的輸出更大，運算更重，等待時間也更長。\u003C\u002Fp>\u003Cp>這會直接影響商業模式。文字工具比較容易塞進訂閱方案。影片工具如果沒有很強的付費理由，很容易變成一次性玩具。你可能很想玩，但不一定會一直付錢。\u003C\u002Fp>\u003Cul>\u003Cli>文字生成：單次成本低，適合訂閱制\u003C\u002Fli>\u003Cli>圖片生成：比文字貴，但仍可控\u003C\u002Fli>\u003Cli>影片生成：算力最高，最難大規模收費\u003C\u002Fli>\u003Cli>留存表現：日常工作工具通常高於創意新奇工具\u003C\u002Fli>\u003C\u002Ful>\u003Cp>再看開發者角度，差別更明顯。你如果用 \u003Ca href=\"https:\u002F\u002Fplatform.openai.com\" target=\"_blank\" rel=\"noopener\">OpenAI API\u003C\u002Fa>，一定要先算每個任務的成本。不要只看 prototype。prototype 看起來都很香，真的上線才知道帳單有多兇。\u003C\u002Fp>\u003Cp>這也解釋了為什麼很多團隊會先做聊天、摘要、搜尋這些低成本功能。因為這些功能比較能撐住大量請求。影片或高解析生成，通常只能當高價功能。\u003C\u002Fp>\u003Ch2>對開發者來說，風險不只在模型\u003C\u002Fh2>\u003Cp>很多人以為 AI 風險只在模型準不準。其實更麻煩的是供應商風險。你一旦把產品綁在單一平台上，就等於把價格、政策、停機、配額，全交給對方決定。\u003C\u002Fp>\u003Cp>這件事在台灣團隊特別要小心。很多新創很愛先接一個 API 就開幹。速度是快啦。可是等到價格調整，或模型下架，整個產品線就會被卡住。\u003C\u002Fp>\u003Cp>所以比較實際的做法，是先設成本上限。再來做使用量控管。最後留一條 fallback 路線。不要把所有功能都押在同一顆模型上。\u003C\u002Fp>\u003Cp>像 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftransformers\" target=\"_blank\" rel=\"noopener\">Hugging Face Transformers\u003C\u002Fa> 和 \u003Ca href=\"https:\u002F\u002Fpytorch.org\" target=\"_blank\" rel=\"noopener\">PyTorch\u003C\u002Fa> 這類工具，就是很多團隊的備援選項。不是因為它們最便宜，而是因為你能自己掌握更多主控權。\u003C\u002Fp>\u003Cp>我覺得最該做的不是追最強模型，而是看單位經濟。每個 active user 花多少？每次輸出花多少？用戶三天後還回來嗎？這些數字，比 demo 好看太多了。\u003C\u002Fp>\u003Ch2>這波其實在提醒整個產業\u003C\u002Fh2>\u003Cp>AI 產業現在很像早期雲端時代。大家都在拼速度，拼功能，拼曝光。可是最後活下來的，通常不是最會喊口號的，而是最會算帳的。\u003C\u002Fp>\u003Cp>影片生成尤其如此。因為它同時吃模型、吃 GPU、吃頻寬、吃等待時間。只要有一個環節卡住，體驗就會掉。只要體驗掉，留存就會掉。留存一掉，成本就更難攤平。\u003C\u002Fp>\u003Cp>所以我會把 Sora 的停運，看成一個很直接的警告。AI 產品不是只看效果。還要看成本曲線。尤其是當你把產品做給一般消費者時，付費意願通常沒有你想像中高。\u003C\u002Fp>\u003Cp>這也是為什麼很多團隊現在會把 AI 功能拆小。先從文字、搜尋、助理開始。等成本模型穩了，再碰影片或多模態。這不是保守。這是活下去的方法。\u003C\u002Fp>\u003Ch2>結尾：先算帳，再談酷不酷\u003C\u002Fh2>\u003Cp>如果你正在做 AI 產品，我的建議很直接。先算每個功能的成本。再算留存。最後才看成長。順序顛倒，通常會很慘。\u003C\u002Fp>\u003Cp>接下來 12 個月，真正能活下來的 AI 服務，應該都會有一件事很像：它們不一定最炫，但一定知道自己每次呼叫 API 到底燒多少錢。你如果現在還沒算這筆帳，真的該開始了。\u003C\u002Fp>","Sora 曾衝到約 100 萬用戶，後來跌破 50 萬，還每天燒掉約 100 萬美元算力。這篇拆解它為何撐不住，也談 AI 供應商風險、成本結構和開發者該怎麼避雷。","www.forbes.com","https:\u002F\u002Fwww.forbes.com\u002Fsites\u002Fjohnsviokla\u002F2026\u002F04\u002F02\u002Fwhen-ai-vendors-fail-lessons-from-the-sora-shutdown\u002F",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775198567172-1cmi.png",[13,14,15,16,17,18,19,20],"Sora","AI 供應商風險","單位經濟","影片生成","OpenAI","算力成本","AI API","台灣開發者","zh",2,false,"2026-04-03T06:42:29.115955+00:00","2026-04-03T06:42:29.088+00:00","done","97e26ccc-a5b2-448d-a473-cc40b3ae5484","sora-shutdown-ai-vendor-risk-zh","industry","677cc705-258e-4459-ba27-e44a987d3325","published","2026-04-07T07:41:10.547+00:00",[34,36,37,39,40,42,44,45],{"name":19,"slug":35},"ai-api",{"name":15,"slug":15},{"name":17,"slug":38},"openai",{"name":18,"slug":18},{"name":13,"slug":41},"sora",{"name":14,"slug":43},"ai-供應商風險",{"name":20,"slug":20},{"name":16,"slug":16},{"id":30,"slug":47,"title":48,"language":49},"sora-shutdown-ai-vendor-risk-en","Inside Sora's Shutdown and AI Vendor Risk","en",[51,57,63,69,75,81],{"id":52,"slug":53,"title":54,"cover_image":55,"image_url":55,"created_at":56,"category":29},"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":58,"slug":59,"title":60,"cover_image":61,"image_url":61,"created_at":62,"category":29},"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":64,"slug":65,"title":66,"cover_image":67,"image_url":67,"created_at":68,"category":29},"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",{"id":70,"slug":71,"title":72,"cover_image":73,"image_url":73,"created_at":74,"category":29},"17cafb6e-9f2c-43c4-9ba3-ef211d2780b1","why-observability-is-critical-cloud-native-systems-zh","為什麼可觀測性是雲原生系統的生存條件","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778794245143-tfqn.png","2026-05-14T21:30:25.97324+00:00",{"id":76,"slug":77,"title":78,"cover_image":79,"image_url":79,"created_at":80,"category":29},"2fb441af-d3c6-4af8-a356-a40b25a67c00","data-centers-pushing-homeowners-to-solar-zh","資料中心推升房主裝太陽能","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778793651300-gi06.png","2026-05-14T21:20:40.899115+00:00",{"id":82,"slug":83,"title":84,"cover_image":85,"image_url":85,"created_at":86,"category":29},"387bddd8-e5fc-4aa9-8d1b-43a34b0ece43","how-to-choose-gpu-for-yihuan-zh","怎麼選《异环》GPU","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778786461303-39mx.png","2026-05-14T19:20:29.220124+00:00",[88,93,98,103,108,113,118,123,128,133],{"id":89,"slug":90,"title":91,"created_at":92},"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":94,"slug":95,"title":96,"created_at":97},"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":99,"slug":100,"title":101,"created_at":102},"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":104,"slug":105,"title":106,"created_at":107},"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":109,"slug":110,"title":111,"created_at":112},"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":114,"slug":115,"title":116,"created_at":117},"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":119,"slug":120,"title":121,"created_at":122},"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":124,"slug":125,"title":126,"created_at":127},"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":129,"slug":130,"title":131,"created_at":132},"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":134,"slug":135,"title":136,"created_at":137},"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"]