What makes a company kill its own product half a year after launch?
OpenAI pulled the plug on Sora last week, retiring its AI video generation tool roughly six months after making it publicly available. The company hasn't offered a fully satisfying explanation, and the timing has people asking uncomfortable questions.
From Flagship Demo to Shutdown
Sora launched with enormous fanfare. OpenAI positioned it as a leap forward in AI video generation - the tool could produce short, surprisingly coherent video clips from text prompts. It quickly became one of the most visible demonstrations of what generative AI could do beyond text and images.
But six months is an absurdly short lifespan for a product from a company valued at over $150 billion. Products at this scale typically get years of iteration before being sunset, not months. Something clearly went wrong, and OpenAI's public statements haven't filled in the picture.
The Face Upload Problem
The most pointed suspicion centers on Sora's face upload feature. The tool invited users to upload photos of their own faces to generate personalized video content. On the surface, that's a logical feature for a video generation tool. In practice, it created a pipeline for collecting biometric data - face geometry, skin texture, expressions - from millions of users who clicked "agree" on a terms of service document they almost certainly didn't read.
Was Sora primarily a data collection exercise? That's probably too cynical. Building a video generation model requires massive compute resources, and OpenAI likely had legitimate product ambitions. But the face data question won't go away, especially as regulators in the EU and several US states have been tightening rules around biometric data collection throughout 2025 and 2026.
The More Boring (and More Likely) Explanation
The less conspiratorial read is simpler: Sora was expensive to run and didn't generate enough revenue to justify the compute costs. Video generation is orders of magnitude more resource-intensive than text or image generation. Each clip requires processing hundreds of frames, and users expected results in seconds, not hours.
OpenAI is reportedly burning through cash at a staggering rate, and every product needs to justify its server time. If Sora wasn't converting free users into paying subscribers at a rate that covered its infrastructure costs, the math just doesn't work - no matter how impressive the demos looked.
The real answer is probably some combination of both factors: regulatory risk around face data made the product legally complicated, while the economics made it financially painful. When a product is both a liability risk and a money pit, the decision to shut it down gets a lot easier.
For anyone who was building workflows around Sora, the lesson is familiar: don't depend too heavily on any single AI tool, especially one that's free or underpriced relative to what it actually costs to run.