Monday they posted the goodbye. Short. Almost breezy about it. “We’re saying goodbye to Sora” — and then some warm language about the community, and then something about details coming soon.
Details coming soon. Right.
I’ve been at this long enough to know what that phrase actually means. It means nobody inside has agreed on the story yet. It means the post-mortem is still being written and whatever comes out the other end will be carefully worded and mostly useless. It doesn’t mean answers are coming. It means they’re buying time.
Six months the thing lasted. Half a year.
February 2024 feels like a long time ago now. OpenAI drops the first footage and — okay, fine, I’ll say it — it was genuinely good. Not “good for AI” good. Actually good. The clips had a physical weight to them that earlier stuff completely lacked. Things moved right. Light behaved. Tyler Perry watched it and shut down an $800 million construction project in Atlanta on the spot. Studio expansion, crews, equipment, all of it — he just stopped it. He watched that footage and thought: if this is real, none of what I’m building makes sense anymore.
Hard to argue with that reaction in the moment.
The coverage went everywhere. Everyone was writing about what it meant for film, for visual effects, for advertising. This enormous cultural conversation, all of it happening before anyone could even try the product.
Then OpenAI waited eleven months to ship it.
I don’t know how else to say that. The demo lands in February 2024. The app comes out in December 2024. Eleven months. The hype peaked, plateaued, started fading, competitors caught up, and then they finally released the thing into whatever was left of the moment. Someone made that call. Several people, probably. Signed off on it at multiple levels.
When it did come out the design felt off from the jump.
Scrollable video feed — okay, sure. Face scanning to put yourself in scenes — weird, but I see what they were going for. And then the IP thing. Built-in tools, on launch day, for generating content using characters and properties belonging to other companies. Not a bug that snuck through. A feature someone decided to ship.
Who approved that? Genuinely asking. What did legal say? Did legal say anything? These are questions I’d love answered.
Downloads were fast regardless. App Store charts within a day. Three million in a few weeks. Altman tweeting about it.
Then TechCrunch called it the creepiest app on the phone and the whole vibe shifted. The face scanning broke almost immediately — deepfakes started showing up, Mario smoking, Pikachu ASMR, Naruto content that would make several rights holders very unhappy. Week after week of this. The tech underneath was functional. Everything surrounding it kept catching fire.
Then somehow, December, there’s a Disney deal.
Three years. Almost 200 characters — the Marvel stuff, the Pixar stuff, Star Wars. Nearly a billion dollars in planned investment. An exec at OpenAI compared it to movies getting sound for the first time.
Someone said that. To journalists. Knowing it would be printed.
Dead now. Nothing was ever paid. Disney put out a statement about respecting the decision and continuing to engage with AI broadly — and if you read that sentence and think it sounds like something a PR team wrote specifically to avoid saying what they actually thought, you’re reading it right. Then Axios reported that Disney’s own people found out the night before the announcement. For a deal signed roughly twelve weeks before that.
Twelve weeks.
Downloads: 3.3 million in November, down to 1.1 million by February. Not a slow decline. A drop. People showed up out of curiosity, poked around, didn’t find a reason to stay. Revenue for the whole run of the product came out around $2.1 million total.
This kind of generation at scale — the GPU costs, the infrastructure — does not run on $2.1 million. Not even close. And the losses weren’t stabilizing, they were widening, because the people who sign up in month one for a buzzy new app are also the people most likely to ghost it by month three.
Bill Peebles had already been throttling generation limits because they didn’t have enough chips. So: losing money, losing users, and making the product worse for the users who remained. All at once.
A Wall Street analyst wrote afterward that Sora couldn’t hold an audience even with everything OpenAI had behind it. That’s a diplomatic way of saying the retention curve looked like a cliff.
The robotics line they’re using — “world simulation research to advance robotics” — is probably true in a literal sense. It’s also the version of “we’re cutting this loose” that sounds like a strategic pivot instead of a retreat. The compute is worth more elsewhere. Consumer video wasn’t working. So it’s gone.
The IPO is in the background of all this. $110 billion raised, heading toward a public offering in the $700 billion range. People inside have been talking about cleaning up before that — getting rid of expensive products with messy numbers and a habit of making the news for bad reasons. Sora fit every part of that description.
Anthropic didn’t touch consumer video. Just kept building coding tools and signing enterprise deals. Quiet work. It paid off. That comparison exists whether anyone wants to make it or not.
The people I feel for are the ones who committed to this thing. Who built workflows, built content strategies, built audiences around it — because why wouldn’t you? There was a Disney deal. There was App Store dominance. There was every signal that OpenAI considered this a serious long-term product. Reasonable people made reasonable decisions based on those signals.
Then it was gone in six months.
They say they’re figuring out content export. No timeline on that.
The real story here isn’t whether the technology worked. It mostly did. The story is whether you can build something people actually keep paying for, keep coming back to, build habits around — and that problem is still wide open. OpenAI took a run at it, it didn’t come together, and they decided the cost of another run wasn’t worth it.
Whether they had any real answer to that problem when they launched, or whether they were improvising the whole time and hoping something clicked —
You know what I think. But it’s your call.
Read more of our blogs here
Leave a Reply