DX Today | No-Hype Podcast & News About AI & DX

OpenAI Kills Sora: The $15M-a-Day Video App That Couldn't Survive the IPO Math - March 26, 2026

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 7:28

Send us Fan Mail

Join Chris and Sarah for a deep-dive conversation into OpenAI's stunning decision to shut down Sora.Today's Topic:OpenAI is discontinuing Sora — its consumer video app, the developer API, and all video generation in ChatGPT. The Disney deal dissolves. A billion-dollar investment evaporates. Chris and Sarah break down why the unit economics were "completely unsustainable," how the IPO calculus forced the decision, what it means for the entertainment industry, and why the consumer AI novelty era may be over.The DX Today Podcast brings you daily deep dives into the most consequential stories in the AI ecosystem.Hosted by Chris and Sarah.
SPEAKER_01

Welcome to the DX Today Podcast, your daily deep dive into the AI ecosystem. I'm Chris, and joining me as always is Sarah. And Sarah, OpenAI just killed Sora.

SPEAKER_00

They did. Tuesday afternoon, Sam Altman told employees that OpenAI is shutting down the Sora Consumer App, the Sora API for developers, and all video generation capabilities in ChatGPT. Done.

SPEAKER_01

This is wild to me. Six months ago, Sora launched, and it was the hottest thing in AI. Text to video that actually looked good. A social feed. Disney signed a deal. And now it's gone.

SPEAKER_00

And it's not just winding down someday. The Sora account posted on X. We're saying goodbye to Sora. To everyone who created with Sora, shared it, and built community around it. Thank you. What you made with Sora mattered, and we know this news is disappointing. That's a eulogy, not a pause.

SPEAKER_01

Okay, so let's unpack this. What's the Disney piece? Because that was a massive deal just three months ago.

SPEAKER_00

In December, OpenAI and Disney announced a three-year licensing agreement. Over 200 Disney characters available in Sora. Mickey Mouse, Cinderella, Yoda, the whole catalog. And Disney pledged a billion-dollar investment in OpenAI as part of the arrangement. That was seen as a landmark moment for AI in entertainment.

SPEAKER_01

And now?

SPEAKER_00

The LA Times is reporting that Disney's investment remained unpaid. No money had actually changed hands yet, and OpenAI hadn't paid any licensing fees for the characters either. A source familiar with the negotiations says no formal deal had been finalized, so the whole thing dissolves. Disney issued a very diplomatic statement saying they respect OpenAI's decision to exit the video generation business.

SPEAKER_01

That's corporate speak for we're not happy, but we'll deal with it. A billion dollars walking away is not nothing.

SPEAKER_00

Right. Though Disney framed it as a learning experience and said they'll continue engaging with other AI platforms.

SPEAKER_01

So why did OpenAI actually do this? They haven't said explicitly, right?

SPEAKER_00

Not explicitly, but the signals are pretty clear. Fiji Simo, OpenAI's chief of applications, told employees in an internal note seen by the Wall Street Journal, we realized we were spreading our efforts across too many apps and stacks, and that we need to simplify. That fragmentation has been slowing us down and making it harder to hit the quality bar we want.

SPEAKER_01

So it's a focus play.

SPEAKER_00

Partly. But the bigger driver is economics. Analysts estimated Sora's daily computational costs at around $15 million. That's billions annualized for a consumer app that was either free or charging modest subscription fees.

SPEAKER_01

$15 million a day. Let me just sit with that for a second.

SPEAKER_00

Video generation is extraordinarily GPU intensive. Every video a user creates requires massive compute. And unlike a text model where the cost per query has been coming down, video scales in the wrong direction, longer videos, higher resolution, more users all multiply the costs. Bill Peebles, who led Sora, had already been limiting how many videos users could create because they were running out of chips.

SPEAKER_01

So this wasn't a failure of the product, it was a failure of the business model.

SPEAKER_00

Exactly. The demand exceeded expectations. Users loved it. But the unit economics were, quote, completely unsustainable. You can't build a consumer business where every interaction costs significantly more than what the customer pays.

SPEAKER_01

Here's the IPO angle, right?

SPEAKER_00

That's the context that makes everything click. OpenAI is preparing for a potential IPO later this year. They generated around $13 billion in revenue last year, but expect to spend roughly $100 billion over the next four years on data center expansion. When you're about to open your books to public investors, a service hemorrhaging cash with no clear path to profitability is exactly what you cut.

SPEAKER_01

So Sora was an IPO liability.

SPEAKER_00

Every dollar of compute going to Sora is a dollar not going to coding tools, reasoning models, enterprise features, the things that actually generate revenue and show a growth story Wall Street wants.

SPEAKER_01

Let me push back though. They made it a marquee launch, they signed Disney, and then three months later they kill it? That feels like more than just focus.

SPEAKER_00

Fair. A few things collided. First, the competitive landscape shifted fast. Google's Vio, Runway's Gen 3, Chinese video models, the market got crowded. Sora wasn't as differentiated as ChatGPT was in the language space. Second, the content moderation burden was enormous. Deep fakes, copyright issues, inappropriate material. That's an ongoing cost center. Third, OpenAI is strategically pivoting toward coding and business users. Anthropic doesn't offer video generation at all, and they're eating OpenAI's lunch with enterprise customers.

SPEAKER_01

That's an interesting point. Anthropic is gaining enterprise share fast. Ramp data showed them capturing over 73% of first-time AI spending among companies. Maybe OpenAI looked at that and thought, we're subsidizing a social video app while our core business is under threat.

SPEAKER_00

Exactly. You can't afford to lose the enterprise race while burning 15 million a day on consumer video creation.

SPEAKER_01

What happens to the underlying research?

SPEAKER_00

They're keeping that. OpenAI said the Sora research team will continue focusing on world simulation research to advance robotics. So the technology, using video to teach AI about the physical world, lives on. It just won't be a consumer product.

SPEAKER_01

Kind of poetic. Sora started as a research project about world models, briefly became a consumer sensation, and now goes back to being a research project.

SPEAKER_00

Full circle. And honestly, video as training data for robots and embodied AI agents is arguably a bigger market long term than consumer video creation.

SPEAKER_01

Zooming out, what does this mean for the ecosystem?

SPEAKER_00

Three big takeaways. First, compute allocation is the central strategic decision in AI now. Every GPU hour is a choice, and companies are getting ruthless about what earns those resources. Second, the consumer AI novelty era may be ending. The look what AI can do phase is giving way to show me the business model. Third, the entertainment industry just learned that AI partnerships can evaporate overnight. Disney structured a billion-dollar deal that never closed.

SPEAKER_01

That last one is going to make every studio rethink how they engage with AI companies. And for builders, if you built on the Sora API, you just had the rug pulled. Platform risk is real.

SPEAKER_00

100%. Expect more demand for open source video models and multi-provider strategies where you're not dependent on any single vendor.

SPEAKER_01

Great conversation. The Sora shutdown is one of those inflection point moments, away from spectacle, towards sustainable business.

SPEAKER_00

Agreed. End of an era, start of a much more pragmatic one.

SPEAKER_01

That's all for today's episode of the DX Today podcast. Thanks for listening, and we'll see you next time.