OpenAI 親手關掉 Sora：每天燒 1500 萬美元、迪士尼 10 億合作破局，AI 影片的商業現實 | OpenAI Kills Sora: $15M/Day Burn, Disney $1B Deal Gone — The Brutal Business Reality of AI Video
By Kit 小克 | AI Tool Observer | 2026-03-28
🇹🇼 OpenAI 親手關掉 Sora：每天燒 1500 萬美元、迪士尼 10 億合作破局，AI 影片的商業現實
短短六個月：從轟動發表到悄悄關閉
2025 年 9 月，OpenAI 正式推出 Sora 獨立應用程式。那時候的新聞標題驚呼：AI 影片時代來了。六個月後的 2026 年 3 月 24 日，OpenAI 宣布關閉 Sora，理由是「資源重新分配」。
這個速度快到讓人傻眼。
數字說話：每天燒 1500 萬，一生賺 210 萬
這才是最讓人冷靜的部分：
- Sora 推論成本：高峰期每天約 1,500 萬美元
- Sora 應用程式終身總收入：210 萬美元
- 下載量高峰（2025 年 11 月）：330 萬次
- 下載量低谷（2026 年 2 月）：113 萬次，掉了 66%
算起來，每賺 1 塊錢，要先燒掉 7 塊。這不是 startup 困境，這是根本性的商業模式問題：推論成本根本沒辦法被消費者願意付的價格所覆蓋。
迪士尼 10 億美元投資泡湯
更戲劇化的是迪士尼的撤資。原本計畫：迪士尼以 10 億美元入股 OpenAI，換取授權超過 200 個迪士尼、Marvel、Pixar、星際大戰角色來生成 AI 影片，預計 2026 年初推出。
結果：沒有一分錢到位，合作就結束了。OpenAI 確認「雙方未有任何金錢往來」。從頭條到零，不到一年。
OpenAI 的下一步：機器人
Sora 研究團隊沒有被裁員，而是轉移到「世界模擬」研究，目標是為機器人訓練提供動態環境模型。OpenAI 官方說法：
「Sora 研究團隊將繼續專注於世界模擬研究，以推進能幫助人們解決真實物理任務的機器人技術。」
這是一個重要信號：OpenAI 正在從「AI 娛樂工具」悄悄轉向「AI 物理世界基礎設施」，把籌碼押在機器人這個更大的市場。
對 AI 消費品的啟示
Sora 的失敗不代表 AI 影片技術沒有未來，Runway、Kling、Google Veo 都還活著。但它揭示了幾個現實：
- 推論成本還沒降到消費者可承受的程度——「酷炫」不等於「願意付錢」
- 免費試用帶來使用者，不帶來收入——下載數暴跌 66% 就是最好的證明
- 娛樂場景 AI 競爭激烈，但沒有任何一家做到「印鈔機」水準
- B2B 可能才是 AI 產品的活路，機器人、企業工作流比消費 App 更耐打
OpenAI 在正確的時間做了正確的決定：把算力從虧損的消費品，押注到更大的產業賭注。這不是失敗，這是策略轉向。
好不好用，試了才知道——但如果試用成本讓你破產，那答案已經很清楚了。
🇺🇸 OpenAI Kills Sora: $15M/Day Burn, Disney $1B Deal Gone — The Brutal Business Reality of AI Video
Six Months: From Viral Launch to Quiet Shutdown
OpenAI launched Sora as a standalone app in September 2025 to massive fanfare. The narrative was clear: the AI video era had arrived. On March 24, 2026 — just six months later — OpenAI announced Sora was shutting down. The official reason: "resource reallocation."
That is a polite way of saying the math never worked.
The Numbers That Explain Everything
Here is what the business actually looked like under the hood:
- Peak daily inference cost: approximately $15 million
- Lifetime app revenue: $2.1 million total
- Peak downloads (November 2025): 3.33 million
- Downloads by February 2026: 1.13 million — a 66% collapse in three months
For every dollar Sora earned, OpenAI spent roughly seven. That is not a startup struggling through a rough patch — that is a product with a fundamental gap between what it costs to run and what consumers are willing to pay. Impressive demos do not pay the inference bill.
Disney's $1 Billion Deal That Evaporated
The Disney angle is what makes this story genuinely dramatic. The original plan: Disney would invest $1 billion in OpenAI and license more than 200 characters from Disney, Marvel, Pixar, and Star Wars for AI-generated video content. The rollout was expected in early 2026.
None of it happened. OpenAI confirmed no money changed hands before the partnership dissolved. A $1 billion headline deal — gone to zero in under a year, before a single frame of Mickey Mouse was ever generated.
Where the Sora Team Is Actually Going
The research team is not being laid off. They are being redirected to "world simulation" — building dynamic environment models to train robots for physical tasks in the real world. OpenAI's official statement:
"The Sora research team continues to focus on world simulation research to advance robotics that will help people solve real-world, physical tasks."
This is a significant strategic signal. OpenAI is quietly repositioning from AI entertainment company to infrastructure provider for the physical world. The technology behind Sora has real value — just not for consumer video subscriptions.
What This Means for Anyone Building AI Products
Sora's shutdown is not proof that AI video is dead — Runway, Kling, and Google Veo are all still operating. What it does reveal are some uncomfortable truths:
- Inference costs have not dropped enough for mass consumer AI to be profitable at scale
- Free trials generate downloads, not revenue — user curiosity does not convert to retention
- Entertainment AI is a contested, high-cost market where even the best-funded players struggle
- The most durable AI products may be B2B, not B2C — enterprise workflows and robotics beat consumer apps on unit economics
OpenAI made the strategically correct call: redirect scarce compute from a money-losing consumer product toward a larger, longer-term bet on robotics. The lesson for everyone building AI products right now is clear — a viral launch and millions of downloads mean nothing if you cannot close the gap between cost and willingness to pay.
好不好用，試了才知道 — but if the cost of trying bankrupts you first, you already have your answer.
Sources / 資料來源
- OpenAI is shutting down its Sora video app just months after launch — CNN Business
- OpenAI will shutter Sora video app — Axios
- OpenAI Shuts Down Sora and Ends Planned $1 Billion Disney Partnership — gHacks
AI 工具觀察站 — 每日精選 AI Agent 與工具趨勢
AI Tool Observer — Daily curated AI Agent & tool trends
留言
張貼留言