OpenAI’s move last week to shut down Sora, its AI video-generation tool, just six months after launching it, immediately sparked speculation. Because the app let users upload their own faces, some wondered if it had been an elaborate data-collection play. But a new WSJ investigation suggests a far simpler reason: Sora was expensive to run, struggled to retain users, and was pulling resources away from OpenAI’s broader AI efforts.

 

So what went wrong? After an attention-grabbing debut, Sora’s global user base peaked at around one million before dropping to under 500,000. At the same time, it was reportedly costing about $1 million per day to operate—not due to overwhelming demand, but because generating video is inherently resource-intensive. Every user session consumed a limited pool of high-performance AI chips.

 

While OpenAI dedicated significant resources to keeping Sora afloat, competitors like Anthropic were gaining ground with developers and enterprise customers. Products such as Claude Code were steadily attracting the audience that drives long-term revenue.

 

In the end, CEO Sam Altman opted to shut Sora down, reallocate compute power, and refocus the company’s priorities. The abruptness of the decision caught even major partners off guard. According to the WSJ, Disney—despite committing $1 billion to the partnership—was informed less than an hour before the public announcement, effectively ending the deal overnight.