On March 24, 2026, OpenAI posted a brief message on its official X account that caught much of the industry off guard: "We say goodbye to Sora. To everyone who created with Sora, shared it and built a community around it: thank you. What you created mattered, and we know this news is disappointing."
There was no press conference, no lengthy statement. Just that text, and the announcement that Sora — the app, the API and the website — would shut down completely. The company said it would share exact timelines for the closure and instructions for users to save their work. Nothing more.
Six months after its launch, one of the year's most talked-about AI tools was leaving the market.
What Sora was and how it got here
OpenAI introduced Sora in February 2024 as a model capable of generating high-quality video from text prompts. The first clips that circulated — cinematic scenes, fluid camera movements, sustained visual coherence — had an immediate impact on the industry. It was, at that moment, the most advanced video generation AI that existed.
For more than a year, access was restricted to selected researchers and creators. The public launch came in September 2025, when OpenAI released Sora as a standalone app — the first independent application the company had launched outside of ChatGPT. The pitch was ambitious: a kind of AI-powered TikTok where users could create, share and remix AI-generated clips.
The takeoff was explosive. Sora hit one million downloads in under five days, outpacing the adoption speed of ChatGPT itself at launch. For days it was the most downloaded free app on Apple's App Store.
The problems that never got solved
The early enthusiasm didn't take long to collide with reality. From day one, the app faced two problems that proved impossible to fix without destroying what made it appealing.
The first was legal. Copyright holders — studios, musicians, production companies — immediately raised concerns about the use of their intellectual property in user-generated videos. What made Sora fun and viral was exactly what made it problematic: the ability to generate any video, with any character or person, in any context. When OpenAI began tightening restrictions to limit that use, the app's main appeal evaporated.
The second was user retention. At its peak, Sora reached around 3.3 million downloads, but by February 2026 that figure had dropped to roughly 1.1 million. The decline reflected a familiar pattern: many people download an AI app out of curiosity, try it once and don't return. Without a clear everyday use case and with increasingly strict content restrictions, Sora couldn't find a way to keep users coming back.
The Disney deal that never closed
The most revealing chapter in Sora's story is not its closure. It's what never happened.
In December 2025, OpenAI announced with considerable fanfare a licensing deal with Disney that would have allowed Sora to generate videos featuring more than 200 characters from Disney, Marvel, Pixar and Star Wars. The deal also included an investment of close to one billion dollars from Disney in OpenAI — entirely in stock warrants, not cash.
It was the most ambitious agreement between an AI company and the entertainment industry to date. The Writers Guild of America had already publicly criticized the deal, warning it would continue defending creative rights against AI's advance.
The transaction never closed. With Sora's shutdown, Disney confirmed the alliance was dissolved. The company's spokesperson was brief and direct: "We respect OpenAI's decision to exit the video generation business and redirect its priorities." Disney walked away unscathed. OpenAI lost both the money and the product.
Why OpenAI shut Sora down now
The company didn't offer a full public explanation. But sources who spoke with Bloomberg and other specialist publications painted a clear picture.
The first factor is economic. Sora carried very high compute costs — generating high-quality video consumes orders of magnitude more resources than generating text. With OpenAI preparing for a potential IPO in the fourth quarter of 2026, pressure to demonstrate profitability and efficient resource use is higher than ever. Maintaining a product with high operating costs and declining users was not sustainable.
The second factor is strategic. An OpenAI spokesperson confirmed to Bloomberg that Sora's research team will continue working, but with a completely different focus: real-world simulation to advance robotics. Instead of making entertaining videos for social media users, Sora's underlying technology is being redirected toward training robots to understand and operate in the physical world. It's a shift in priorities that says a great deal about where OpenAI is heading.
The third factor is resource concentration. OpenAI is betting increasingly heavily on Codex — its code generation tool — and on improvements to the ChatGPT ecosystem. Sora was competing internally for computational resources with products that generate more revenue and have stronger commercial traction.
What this says about generative AI
Sora's closure is not a story of technological failure. The technology worked — and its second version, launched in September 2025, was genuinely impressive. The problem was not what Sora could do. It was that what it could do couldn't find a sustainable business model.
The most downloaded app from a company valued at close to $730 billion lasted six months. That raises an uncomfortable question for the entire industry: how many AI products that seem inevitable today will follow the same path?
Generative AI advances at a speed that outpaces companies' ability to build businesses around it. Enthusiasm is not a strategy. And Sora, for now, is the clearest proof of that.
Comments
💬 Log in to comment💬 Join the conversation and log in to comment.