On Monday, March 23, OpenAI published a blog post titled "Creating with Sora safely." It covered watermarking, content provenance metadata, updated consent controls, stronger protections for teen users. It read like a company settling in for the long haul.
Twenty-four hours later, Sora was dead.
On Tuesday, March 24, Sam Altman told OpenAI staff the company was killing Sora entirely — the app, the API, and the underlying video model. Teams from Disney and OpenAI had met that same morning to discuss the ongoing partnership. Thirty minutes after that meeting ended, Disney was informed OpenAI was pulling the plug.
The $1 Billion Deal That Lasted Three Months
The Disney partnership had been a marquee moment for OpenAI's consumer ambitions. In December 2025, then-CEO Bob Iger announced a three-year, $1 billion agreement with OpenAI. Disney would take a major stake in the company and license more than 200 of its characters to Sora. Mickey Mouse, Darth Vader, Baby Yoda, Deadpool — all fair game for AI-generated fan videos.
The vision was expansive. Users would generate personalized clips using Disney, Marvel, Pixar, and Star Wars IP. Some videos could stream on Disney+. Disney employees would get access to ChatGPT. Iger framed it as a bold step toward interactive, AI-powered entertainment.
No money ever changed hands.
Disney's spokesperson issued a measured statement following the Tuesday announcement: "As the nascent AI field advances rapidly, we respect OpenAI's decision to exit the video generation business and to shift its priorities elsewhere. We appreciate the constructive collaboration between our teams and what we learned from it, and we will continue to engage with AI platforms to find new ways to meet fans where they are while responsibly embracing new technologies that respect IP and the rights of creators."
The statement was diplomatic. The situation was not.
Six Months, $2.1 Million, $15 Million Per Day
Sora launched as a standalone app on September 30, 2025 alongside the second-generation model, which added audio capabilities. Within days it had topped the overall US App Store. By November it had reached 3.3 million monthly downloads.
Then it started bleeding users.
By February 2026, monthly downloads had fallen to roughly 1.1 million — a 67%** collapse** in three months. Total lifetime in-app revenue across its entire run came to approximately $2.1 million. The estimated compute cost to run the service: $15 million per day.
The math was not ambiguous.
Video generation is exponentially more compute-intensive than text. While ChatGPT and Claude trade tokens by the billion, Sora had to render photorealistic frames, model physics, animate coherent motion across entire scenes. Every video cost OpenAI far more than any user ever paid.
OpenAI said it needed to make hard trade-offs as compute demand grows. The company's spokesperson described the Sora research team as continuing to focus on "world simulation research to advance robotics that will help people solve real-world, physical tasks" — a quieter acknowledgment that the underlying technology may still have a future, just not as an entertainment product.
The Safety Timeline That Raised Questions
The timing of the shutdown rattled some observers. One day before the announcement, OpenAI had published what looked like a comprehensive safety framework for Sora. That post was live on the website when the shutdown was announced on March 24.
Behind the scenes, something else was changing. Sam Altman reorganized safety oversight at OpenAI, with safety responsibilities moving under Mark Chen, OpenAI's chief research officer, and security moving under president Greg Brockman. Altman himself stepped back from direct safety oversight to focus on what he described as capital raising, supply chain management, and building data centers at scale.
Fidji Simo, who had been CEO of Applications, saw her role renamed to CEO of AGI Deployment. She told staff in an earlier address that the company could not afford to be distracted by what she called "side quests." Sora, it turned out, was one of them.
The same meeting that killed Sora also folded the Atlas browser project into a planned super app. OpenAI's Jony Ive-designed consumer device, already pushed back to 2027, was another casualty of the retreat. An OpenAI that had spent 2025 launching products in every direction was pulling back hard.
The Safety Problems Sora Never Solved
Even as a technical achievement, Sora spent its short life generating controversy.
The app's signature feature — originally called "Cameo" before the celebrity video platform of the same name sued OpenAI over the trademark and won a court order forcing a rename to "Characters" — let users scan their faces and insert themselves into AI-generated videos. OpenAI said it would block deepfakes of public figures who had not opted in. Those guardrails failed repeatedly.
Deepfakes of Martin Luther King Jr. and Robin Williams circulated on the platform, prompting complaints from their daughters. Advocacy groups documented the proliferation of nonconsensual intimate imagery and what critics called "AI slop" — a flood of low-quality, often offensive generated content.
The Motion Picture Association, SAG-AFTRA, and the Writers Guild of America had all raised alarms. SAG-AFTRA executive director Duncan Crabtree-Ireland had stated plainly that "everyone in the entertainment industry, especially all the creative talent, are incredibly worried about what the implications are."
Disney's creative guilds added their own internal pressure. Sources told outlets that Disney had demanded ironclad guarantees about the provenance of Sora's training data — guarantees OpenAI was apparently unable to provide. That gap, combined with the political cost of a deep AI integration with Hollywood unions watching closely, made the partnership increasingly precarious even before Tuesday's call.
Sora's History: A Timeline
What OpenAI Is Betting On Instead
The resources freed up from Sora are headed somewhere specific.
OpenAI has completed initial development on a new model internally codenamed "Spud." Altman has told staff internally he expects the company to have a "very powerful model" ready in "a few weeks" — one he described as capable of meaningfully accelerating the overall economy. Whether Spud becomes GPT-5.5 or GPT-6 remains publicly unclear.
The pivot tracks with OpenAI's IPO preparation. The company is targeting a public offering in the fourth quarter of 2026, with reports placing its expected valuation between $830 billion and $1 trillion. Institutional investors evaluating that kind of valuation do not want to see $15 million daily compute bills for a service that generated $2.1 million in total lifetime revenue.
OpenAI also recently raised $10 billion in fresh funding, bringing its total raised to roughly $120 billion. The company's foundation pledged $1 billion in grants spanning life sciences, jobs and economic impact, and AI resilience programs. The narrative being built is less "social media meets AI art" and more "critical infrastructure for civilization."
Sora did not fit that narrative.
A Window for Everyone Else
The competitive question now is who absorbs the gap Sora leaves behind.
Runway, Kling, and Google's Veo remain active. None have Sora's brand recognition or OpenAI's distribution, but all are still operating. OpenAI's exit from the consumer video space does not eliminate the technology — it hands the field to competitors who were already closing fast.
For Hollywood, the reaction is complicated. On one hand, the guilds who fought hardest against AI video generation just watched their most prominent adversary exit the space. SAG-AFTRA and the WGA can claim some credit for making the Disney deal politically toxic.
On the other hand, Runway and Kling are still generating studio-quality video. The threat has not disappeared. It has simply changed corporate logos.
The Bottom Line
Sora lasted six months as a public product. It topped the App Store, generated deepfakes of civil rights icons, and nearly handed a billion dollars to the world's most valuable entertainment company — all while hemorrhaging $15 million a day on compute it could not monetize.
The shutdown is a reminder that viral adoption and sustainable economics are different problems. OpenAI solved the first one with Sora. It never got close to the second.
New CEO Josh D'Amaro, just six days into the job, inherited the fallout of a deal his predecessor Bob Iger announced and never closed. Disney's statement — diplomatic, measured, noncommittal about what comes next — signals that the company is not done with AI video. Just done with this version of it.
Altman's framing to his staff was direct: video was a side quest. The main quest, apparently, is everything else.
Sources
- OpenAI Will Shut Down Sora Video App; Disney Drops Plans for $1 Billion Investment (March 24, 2026)
- Disney's $1B Investment In OpenAI DOA As Sam Altman Pulls Sora Plug (March 24, 2026)
- Disney Exits OpenAI Deal After AI Giant Shutters Sora (March 24, 2026)
- OpenAI shuttering Sora video-generating service (March 24, 2026)
- OpenAI Fumbled Its $1 Billion Deal With Disney (March 25, 2026)
- Sora Was Shut Down a Day After Publishing Its Safety Standards (March 25, 2026)
- OpenAI Kills Sora, Preps Spud Model (March 25, 2026)
- Why OpenAI and Disney Ended Their Deal, Bob Iger Big Strategic Move (March 25, 2026)
- OpenAI pulls the plug on Sora, the viral AI video app that sparked deepfake concerns (March 25, 2026)
- OpenAI shuts down Sora video app amid deepfake concerns (March 25, 2026)
- OpenAI just killed Sora as company readies IPO and new 'Spud' model (March 25, 2026)
- Josh D'Amaro's Disney Moves After OpenAI Deal Ends (March 25, 2026)