OpenAI just shut down Sora. The app hit the top of the App Store charts, hit a million downloads in five days, and had a billion-dollar deal with Disney.
Poof. All gone.
What Happened
In case you missed it, OpenAI announced yesterday that it's discontinuing Sora, the AI video creation app and API, and sending the team to work on "world simulation research for robotics."
Translation: This thing was losing money hand over fist and no one was using it anymore.
The Timeline That Should Make You Nervous
Sora 2 launched in September 2025. By October, it was the #1 app in the US App Store. A million downloads in under five days. The hype was unreal.
By December, downloads dropped 32%.
By January 2026, they dropped another 45%. The app fell to #101 on the App Store and #181 on Google Play.
From #1 to irrelevant in three months.
And the Disney deal â a billion-dollar, equity-based agreement. Licensing 200 characters from Marvel, Pixar, and Star Wars. Disney short videos with AI-generated characters of their IP with plans to launch on Disney+.
Disney has confirmed they've walked away. The entire project is dead.
This Should Scare Every AI Builder
Sora wasn't some plucky startup that couldn't make it to the other side of runway. This was OpenAI. The company is worth hundreds of billions. They have the best model, the biggest brand, and the deepest pockets in the world for generative AI.
And it still wasn't enough to figure it out.
The problem wasn't the tech. The demos were actually incredible. The problem was the canyon of a gap between "wow, cool video" and "I'd like to pay for this every month."
That gap has killed more AI products than poor modeling ever will.
The Demo Trap ðŠĪ
Right now, every AI dev that's shipped a feature knows this gap intimately. You ship the bells-and-whistles demo. You get on Twitter. TechCrunch write-up. Product Hunt features you in their newsletter.
Ninety days later, your user charts are Sora's App Store standing.
Dem videos do not come cheap. Meanwhile, the AWS/GCP bills keep ticking along and they don't care if anyone is using your service.
Where the Money Actually Is
AI is a dangerous trap, which is why Altman's pivot is towards AI agents and a new model known as "Spud." That should tell you where people actually are spending money in AI today.
â Not in content creation, but in task completion
â Not in "wow, look at this" but in "this saved me 20 minutes"
â AI pays if it does your work. AI does not pay if it makes cartoons.
The Question You Should Be Asking
So then, the question you should be asking yourself if you're building an AI product right now is this:
"If our AI feature disappeared tomorrow, would our users notice it was gone the next day?"
If the answer is "no," you should probably brace yourselves. You may just have built the next Sora. Beautiful demo, nobody cares.
The best AI products I've seen recently aren't the flashy ones. They're the boring ones. The ones that auto-fill a form, catch a bug in CI, or draft a response you actually send.
Nobody tweets about those. But nobody cancels them either.
Sora proved that you can have the best tech, the biggest brand, and a billion-dollar partner â and still fail if you're solving a problem people don't actually have.
What's the most overhyped AI feature you've seen ship this year? ð
Top comments (0)