OpenAI just killed Sora. What does this tell us about building AI products right now?
The news dropped yesterday: OpenAI is shutting down Sora, their AI video app, six months after launch. The Disney $1B deal is off, and the API is going away, too.
The arc is fascinating if you zoom out. The app launched in September 2025, hit the top of the App Store within a day, and reached 1M downloads faster than ChatGPT did. By January, downloads had dropped 45%, and the whole thing had made roughly $2.1M in in-app purchases over its lifetime.
What killed it was a combination of deepfake controversies that forced OpenAI to keep restricting what you could create, copyright chaos, and a compute cost that clearly wasn't justified by the retention numbers. The resources are now going toward coding tools, reasoning, and enterprise products ahead of their IPO.
On this, a learning for me is that hype and retention are completely different metrics, and "most downloaded app in X category" means nothing if you can't solve the moderation and trust problem at scale.
Do you think this is a story about AI video being too early, or about Sora specifically being the wrong product for OpenAI to own? And if you're building in the AI space, how are you thinking about your own "is this compute worth it?" question?
Replies
Totally agree. Some apps burn incredibly bright and then vanish just as fast.
@Periscope Producer back in the day, and more recently @Clubhouse, are perfect examples of how hype alone doesn’t translate into retention for the product.
In a hyper-competitive AI landscape where new products launch every week, what really matters is how sticky the product is, how fast the innovations are shipped, how compelling the ongoing use cases are, and whether reviews reflect real, repeatable value, not just attention during the peak launch moment.
@rohanrecommends 100%! I actually completely forgot about Clubhouse, and I was a huge fan of the platform in its early weeks/months 😅
Do you think we can spot a pattern for this situation? Regarding the apps that shine bright like a diamond in their early days, and then they suddenly disappear? Is there something more they could've done to avoid this?
This was a classic case of biting more than you can chew. In theory this was an amazing development, but factor in the constant struggle of IP, copyright infringement, etc this became a headcase fast. In 10 years we may look back at Sora as being "ahead of its time" if we can figure out how to agentize IP and have them communicate with LLMs etc.
sora was never a product, it was a demo that got shipped. nobody who actually makes video content used it for anything real because the workflow is completely wrong - you dont need a magic box that generates 10 second clips from text, you need image gen + selective animation + ffmpeg. the people making actual AI youtube channels spend like $2 per video using cheaper models and only animate maybe 20% of scenes. sora was solving a problem content creators never had