OpenAI kills Sora and ditches Disney: when spectacle stops paying the bills

$1 billion. That's what Disney was about to pour into OpenAI. Last night, it all went up in smoke.

On Monday, March 24, OpenAI announced it's shutting down Sora, its AI video generator. Not pausing it. Not pivoting it. Shutting it down. Disney pulled its investment and walked away from the partnership within hours.

Here's the thing: Sora wasn't some forgotten lab project gathering dust. It was OpenAI's second consumer hit after ChatGPT. And its death tells a story way bigger than the end of one video tool.

From viral demo to App Store #1

Rewind a bit to measure the fall.

February 2024: OpenAI drops a Sora demo. A few seconds of AI-generated video from text prompts. A woman walking through snowy Tokyo streets. A golden retriever hosting a podcast. The internet loses it. Social media declares it an "iPhone moment" for video. The hype is real.

Ten months later, December 2024, the first public version ships. Sora 1 works, even if the demo magic collides with reality in the form of six-fingered hands and wonky physics.

Then September 2025: Sora 2 launches as a standalone app. It hits #1 on the App Store. A million downloads in record time. Generated clips flood TikTok and Instagram.

The Disney deal: summit before the cliff

In December 2025, the Disney partnership looked like the coronation. A three-year license covering 200+ characters: Marvel, Pixar, Star Wars, the Disney classics. The plan: let users create "fan-inspired" videos that would stream on Disney+.

Bob Iger, Disney's CEO, was still hyping the deal in February 2026. The hottest AI company meets the world's most powerful IP catalog.

Except in the hallways at OpenAI, the mood had already shifted.

Code red: the real war is elsewhere

March 2026. Internal all-hands. Fidji Simo, OpenAI's head of consumer apps, says the words "code red." Her directive is crystal clear: kill the side quests.

Here's what was happening while Sora was racking up likes and downloads: the competition was moving on entirely different terrain. Anthropic's Claude was eating market share in enterprise and coding. Google's Gemini was dominating benchmarks by late 2025. Both were attacking where the actual contracts live: productivity, software development, enterprise integration.

OpenAI found itself in a paradox. The most viral AI product on the planet, yet losing strategic ground in the segments that actually pay.

The math became simple, almost brutal. Video generation devours massive compute. Every second of Sora video is GPU cycles not training next-gen language models. And those models are the core business.

Sam Altman put it bluntly: shutting down Sora frees resources for what matters. OpenAI is building a "super app" merging ChatGPT, Codex, and Atlas, refocused on productivity and enterprise. Show's over, time for plumbing.

Disney gets thirty minutes' notice

Disney's team was in a meeting with OpenAI thirty minutes before the public announcement. Thirty minutes. To learn the whole thing was going in the trash.

Disney's statement is a masterpiece of corporate ice: "We respect OpenAI's decision to exit the video generation business." When a company "respects" your decision in a press release, they're furious.

Josh D'Amaro, Disney's new CEO, inherits a beautiful mess. On one side, a flagship partnership evaporating overnight. On the other, open legal proceedings against basically every generative AI player: cease-and-desist letters to Google and Meta, lawsuits against Midjourney, Minimax, ByteDance and its Seedance tool.

Disney wanted to control AI that touched its characters. The strategy boiled down to two words: license friends, sue everyone else. Except the main friend just left the table.

Spectacle gets likes, utility gets budgets

What's happening here goes way beyond OpenAI and Disney.

We've spent two years measuring AI progress by its ability to impress us. Generated video is the perfect symbol: it's spectacular, it's shareable, it racks up millions of views. But it doesn't solve anyone's specific problem.

Meanwhile, tools that help developers code faster, sales reps draft proposals, or legal teams analyze contracts don't trend on X. But they justify $200/month per-seat subscriptions.

Sora had the downloads and the Disney deal. But by January 2026, the excitement was fading and problems were piling up: violent content, deepfakes, unresolved copyright questions. The GPU cost was astronomical, so were the risks.

This feels like an inflection point for the whole industry. When the most prominent AI company admits the "wow" isn't enough, that's a signal nobody can ignore.

What's next

AI video generation isn't dying. ByteDance, Minimax and others are still in the game. But they're advancing under mounting legal pressure, without the shield the Disney deal gave OpenAI.

For OpenAI, the bet is clear: win the race against Anthropic and Google where margins actually exist. Transform ChatGPT from conversational novelty to indispensable work tool. It's less exciting than thirty-second Pixar shorts. But it's probably more viable.

The lesson applies to everyone in this industry. Building what impresses is easy. Building what's useful is the real work. And sometimes, to move forward, you need the guts to kill what shines.