Killing a product that ranked #1 in the App Store and had a Disney deal – that sounds like a mistake. OpenAI did exactly that with Sora. And I think it's one of the most consequential decisions a tech company of this size can make.
Sora was technically impressive. Video generation at a level that made the industry take notice. Still, it got killed. The reason: the compute capacity is now needed for "Spud" – the next major model that, according to Sam Altman, will come in a few weeks and "actually accelerate the economy." Internally, employees called Sora a "drag on resources." That's not failure. That's prioritization.
#Focus Isn't a Given
Applications CEO Fidji Simo had communicated internally weeks earlier: no more "side quests." Sora was exactly that – an impressive detour, but not a strategic core.
OpenAI is currently fighting head-to-head against Anthropic for enterprise customers. That's not a market where you can afford distractions. If you want to win there, you need focus on what enterprise actually needs: powerful models, reliable APIs, clear roadmaps.
Video generation doesn't fit in there right now. Not because it's worthless, but because it's the wrong construction site at the wrong time.
#The Sora Paradox
Here's the real tension: Sora worked. It had users, an App Store hit, a billion-dollar partner. That doesn't make the decision easier – it makes it braver.
It's easy to kill a product that flops. Then you're just cleaning up what wasn't working anyway. It's harder to end something that has traction but is pulling in the wrong direction.
The billion-dollar Disney deal is now on ice. That's painful. But apparently the pain was smaller than the opportunity cost of continuing to pour compute resources into Sora while Spud waits.
#What I See in Projects Again and Again
I work a lot with smaller teams and founders. And the pattern is the same everywhere: features get built because they're cool. Not because they solve the core problem.
An online shop builds a sophisticated recommendation system before the checkout works. A SaaS app gets a dashboard with ten widgets before the core use case is solid. A freelancer builds an elaborate portfolio section for a service they don't even want to offer anymore.
That's human. New features feel like progress. But progress in the wrong direction isn't progress.
#The Counter-Argument – and Why It Doesn't Hold
You could say: OpenAI is giving up a market. Video generation is a growth field. Runway, Kling, Pika – the competition isn't sleeping.
That's true. But Sora chief Bill Peebles himself communicated where the journey is headed: "world simulation" for robotics. The team isn't retiring, it's switching contexts. And that context is strategically far more interesting than consumer video.
Besides: giving up markets doesn't mean losing them forever. OpenAI can come back when Spud is running and resources can be allocated differently again. Trying to win everywhere simultaneously often means winning nowhere.
#What You Can Actually Take Away
Three questions I ask myself about every project – and that you should ask yourself too:
Does this feature solve the core problem? Not: Is it cool? Not: Could someone use it? But: Does it help the user with the actual job?
What does it really cost? Not just in money. In attention, in maintenance overhead, in complexity. Sora ate compute that was needed elsewhere. What is your feature eating?
Is it pulling in the right direction? A working product can still head in the wrong direction. Traction isn't proof of relevance.
Deleting feels like failure. Most of the time it's the opposite. It means you know what you want – and are willing to give up something that works to get there.
That's not giving up. That's clarity.
