When OpenAI pulled the plug on Sora last week, the conspiracy theories started immediately. Face uploads? Must be a data harvesting scheme, right? Turns out the truth is way less interesting and way more instructive for anyone building in the AI space.
Sora was hemorrhaging a million dollars every single day. Not because it was wildly popular. Because video generation is brutally expensive to run, and hardly anyone was actually using it.
The Math That Killed a Flagship Product
Here’s what happened after the splashy launch: user count peaked at around a million, then cratered to under 500,000. Meanwhile, every person who wanted to insert themselves into some AI-generated fantasy world was burning through precious GPU cycles. The kind of compute that OpenAI desperately needs for products that actually make money.
I’ve seen startups die over unit economics before, but this is OpenAI we’re talking about. They’ve got more funding than most countries’ GDP. The fact that they couldn’t justify keeping Sora alive tells you everything about how the economics of video generation work right now.
Think about it. While an entire team was babysitting Sora and watching it bleed cash, Anthropic was quietly eating their lunch with Claude Code. The enterprises and developers who actually pay bills were choosing Claude over ChatGPT. That’s not a minor problem. That’s an existential threat.
Disney Found Out An Hour Before You Did
The WSJ’s reporting on the Disney situation is genuinely wild. A billion dollar partnership commitment. The entertainment giant probably had teams already planning integration workflows, pitching internal stakeholders, maybe even building proof of concepts.
They got less than an hour’s notice before the public announcement. The deal just evaporated. That’s not how you normally treat a partner who’s committed ten figures to your platform.
It shows how sudden and urgent this decision was. Sam Altman looked at the numbers and realized they were losing the war while throwing resources at a product that wasn’t moving the needle. Sometimes in tech, you have to kill your darlings before they kill you.
What This Means For The Rest Of Us
The Sora shutdown is a masterclass in what happens when hype meets reality in the AI space. Video generation is cool. It gets great demos. People tweet about it. But does it drive revenue? Can you scale it profitably? Will developers build businesses on top of it?
For Sora, the answer to all three was no.
I think we’re going to see more of this in 2026. The era of shipping AI features just because they’re technically impressive is ending. The companies that survive will be the ones that can answer the boring questions: What’s the cost per inference? Who’s actually paying for this? How does this help us compete where it matters?
OpenAI made the right call here, even if it was embarrassingly late. They freed up compute, killed a money pit, and can now focus on not losing the developer market to Anthropic. But the fact that they launched Sora at all, with those economics, makes you wonder what other experiments are running right now that will meet the same fate six months from now.