OpenAI in Talks to Acquire Startup to Revive Sora Video Generation Business
nashnova research
OpenAI is negotiating with a startup co-founded by Sora's former research lead, offering model access in exchange for equity — effectively resurrecting the video-generation model shut down in March. The open question: how much room is left in an AI video market now dominated by Chinese players.
How is this deal structured?
OpenAI would give the startup access to the Sora model in exchange for equity; the exact stake is undisclosed.
The startup has three co-founders: film producer Jeffrey Katzenberg, former Dropbox CFO Sujay Jaswa, and Sora's former research lead Bill Peebles.
This means → Peebles helped design Sora's architecture. Even without full model weights — the weights are essentially the model's learned knowledge; having them means having the model itself — his team can make substantial modifications.
Why was Sora shut down in the first place?
The core problem was cash burn: the Wall Street Journal reported Sora was losing roughly $1 million per day, consuming far more compute than it earned.
User growth couldn't sustain it either. After launching in September 2025, it hit the top of the U.S. App Store with nearly 1 million users, then rapidly fell below 500,000.
This means → Sora's failure wasn't "nobody wanted it." It was "not enough people wanted it to justify the cost." The business model simply didn't work.
What happened to the Disney partnership?
At its peak, Sora had a tentative deal with Disney: $1 billion in investment plus a license to use Disney characters in AI-generated content.
The deal collapsed as soon as OpenAI announced Sora's shutdown — it never formally closed.
Sora also faced disputes over copyright infringement and unauthorized use of public figures' likenesses. Those legal risks remain unresolved.
Why did OpenAI choose to shut it down rather than keep running it?
Competitive pressure was the key backdrop. Anthropic's push into the enterprise and coding markets forced OpenAI to concentrate resources on ChatGPT and other core products.
OpenAI is also preparing for an IPO. Shedding high-cost side projects is a standard pre-listing move.
In plain terms = Sora wasn't worthless. But in a "sprint to IPO + core business under pressure" phase, a side project burning $1 million a day had to go.
How dominant are Chinese companies in this space now?
After Sora's exit, the AI video market didn't stall. ByteDance, MiniMax, and Kuaishou now lead the field.
AI-generated video is widely used in mobile short dramas, short-video platforms, and corporate marketing — use cases more downstream and more practical than Sora's.
This reflects a shift in competition: the race is no longer about "who has the strongest model" but "who can push costs low enough to make money."
Can this deal actually work?
If it closes, OpenAI recovers part of its Sora investment through equity while shifting the high-cost operations to the startup.
But Sora's two core weaknesses — poor user retention and unresolved copyright disputes — haven't gone away. The new company inherits both.
This means → The real question isn't whether the deal gets signed. It's whether the new company can find a differentiated position in a market Chinese firms already dominate — and turn Sora from an expensive technical showcase into a product that earns money.
市场有风险，内容仅供研究参考，不构成投资建议。