On Saturday, April 26, at some point that OpenAI has not bothered to specify, the Sora web and mobile apps will stop working. Your videos will disappear unless you export them first. The API gets a longer goodbye, dying September 24. But the product is already dead. It has been dead for weeks.

Sora is the first major AI product to die at scale, and the autopsy reveals a pattern that every AI company should study.

The Numbers That Killed It

Sora peaked at around one million users after its September 2025 public launch. That number collapsed to fewer than 500,000 before OpenAI pulled the plug. The cost: roughly $1 million per day to operate, according to The Next Web's reporting. At no point did Sora generate enough revenue to justify those economics.

The math was never going to work. Video generation requires orders of magnitude more compute than text or image generation. Each Sora output consumed GPU time that could have been sold to enterprise customers at vastly better margins. OpenAI was subsidizing a consumer novelty at the expense of its core business.

The Executive Exodus Tells the Story

On April 17, OpenAI lost three senior leaders in a single day. Kevin Weil, the chief product officer who championed Sora. Bill Peebles, the researcher who led Sora's technical development. Srinivas Narayanan, the enterprise CTO. Three departures, one day, and the common thread was a company pivoting away from everything they built toward.

CEO Fidji Simo's directive was clear: kill the side quests. Sora was the biggest side quest. A $1 billion Disney deal, dead. An entire video generation research team, scattered. The integrated social media features that were supposed to make Sora sticky? Nobody cared.

Why Sora Failed Where ChatGPT Succeeded

ChatGPT solved a problem people had every day: writing, research, coding, analysis. You could use it for work. It saved time. It was measurably useful. Sora solved a problem almost nobody had: generating short AI videos from text prompts. Fun to try once. Hard to justify paying for monthly.

The enterprise use case never materialized either. Hollywood wanted control, consistency, and long-form generation. Sora offered none of those at production quality. Marketing teams experimented and went back to stock footage. Social media creators found the output too uncanny for audiences that were already developing AI fatigue.

The Market Sora Left Behind

Sora's death does not mean video AI is dead. It means OpenAI decided the market wasn't worth the cost. Runway still has deep Hollywood relationships and a production-grade workflow. Google's Veo has YouTube distribution, the largest video platform on earth. Kling AI in China competes on speed and price. Alibaba's Happy Oyster model is building text-to-interactive-3D-worlds, which is arguably where the medium is heading.

The $12 billion video AI market didn't shrink because Sora left. It just lost its most famous entrant. And every competitor is better positioned to serve actual customers.

The Lesson for Every AI Company

Sora is proof that technical capability is not a business model. OpenAI built one of the most impressive video generation systems ever created. The demos were stunning. The research was genuinely novel. None of that mattered because the economics were upside down.

For OpenAI, Sora's death is actually good news. One fewer money-losing product ahead of an IPO. One fewer distraction from the enterprise push that Anthropic forced by eating their lunch on professional AI spending. GPT-5.5 launched two days ago at double the API price. That is where OpenAI's future lives. Not in 15-second videos of cats walking through Times Square.

But for users who built workflows around Sora, and for creators who invested time learning the tool, this is a reminder that every AI product from every vendor might vanish tomorrow. Download your data before Saturday. OpenAI will not keep it for you.