Last updated: 30 May 2026
OpenAI officially shut down Sora on guide March 24, 2026, ending the company‘s highest profile consumer AI product after it burned approximately $1 million per day in operating costs. Total lifetime revenue from the video generation tool was just $2.1 million. A planned $1 billion partnership with Disney collapsed after the entertainment giant learned of the shutdown less than an hour before the public announcement.
Key Highlights
- Sora shut down on March 24, 2026 with app closure on April 26 and API closure on September 24
- Operating costs estimated at approximately $1 million per day, with peak inference costs cited as high as $15 million daily
- Total lifetime in app purchase revenue: $2.1 million
- User count peaked at approximately 1 million, then dropped below 500,000
- Disney had committed $1 billion to a Sora partnership before learning of the shutdown at the last minute
- Sam Altman is redirecting compute resources toward world simulation research for robotics
The Numbers Tell the Full Story
Sora’s economics were never sustainable. At $1 million per day in operating costs against $2.1 million in total lifetime revenue, the product was running at a loss ratio that no amount of growth could have corrected on the existing architecture. Some estimates put peak inference costs as high as $15 million per day during periods of heavy usage, though OpenAI has not confirmed that figure publicly.
User engagement told the same story from the demand side. After launching to significant media attention, Sora attracted roughly 1 million users. That number subsequently fell below 500,000, indicating that the product failed to retain users beyond the initial curiosity phase. For a tool backed by the most recognized name in AI, that retention curve was a clear signal.
The Disney Collapse
The most consequential fallout from Sora’s shutdown is the death of a $1 billion Disney partnership. Disney had committed to integrating Sora’s video generation capabilities into its content production pipeline, a deal that would have been the largest single enterprise commitment to generative AI video.
Disney reportedly learned of the shutdown less than an hour before OpenAI’s public announcement. The timing suggests that either OpenAI made the decision rapidly or chose not to inform its largest prospective partner in advance. Neither interpretation reflects well on the partnership management side of the business.
For Disney, the Sora collapse is a setback but not a crisis. The company has relationships with multiple AI labs and was already evaluating alternatives. For OpenAI, losing a billion dollar anchor partnership on the eve of a potential IPO introduces a trust question that prospective enterprise customers will now factor into their own evaluations.
What Altman Is Betting On Instead
Sam Altman announced that the compute resources previously allocated to Sora will be redirected toward “world simulation research” focused on robotics applications. This pivot suggests OpenAI views physical world modelling as a more viable commercial path than consumer video generation.
The logic is defensible. Robotics simulation requires the same core capability, generating physically plausible visual sequences, but serves an enterprise market with higher willingness to pay and clearer integration paths. The consumer video market, by contrast, proved that users were willing to experiment with AI video but not willing to pay enough to cover the compute cost of generating it.
What This Means for Generative Video
Sora’s shutdown does not mean generative video is dead. It means generative video at consumer pricing is economically unviable with current inference costs. The technology works. The cost structure does not.
This is a recurring pattern in AI deployment. Capabilities advance faster than the economics that make them commercially sustainable. Sora could generate impressive video. It could not generate impressive video at a price that users or advertisers would cover. That gap is the central challenge for every generative AI product operating at the frontier of compute intensity.
