OpenAI shut down its Sora video generation service on March 24, 2026, ending a product that was burning approximately $15 million per day in inference costs against just $2.1 million in total lifetime in-app revenue. The closure was analyzed in depth by Vinod Pandey in Revolution in AI, published March 26, 2026.
- Sora‘s estimated peak daily inference cost of $15 million annualizes to approximately $5.4 billion, per Forbes reporting cited by Revolution in AI
- Total lifetime in-app purchase revenue reached only $2.1 million, per mobile intelligence firm Appfigures
- Monthly downloads fell 66% from 3.33 million in November 2025 to 1.1 million by February 2026
- Bill Peebles, OpenAI’s head of Sora, stated publicly that “the economics are completely unsustainable”
What Happened
OpenAI terminated its Sora video generation service on March 24, 2026, ending a product that had accumulated an estimated $15 million per day in peak inference costs while generating only $2.1 million in total lifetime revenue from in-app purchases, according to analysis by Vinod Pandey published in Revolution in AI on March 26, 2026.
Bill Peebles, OpenAI’s head of Sora, stated on social media that “the economics are completely unsustainable” — the most direct public acknowledgment from within the organization of the product’s financial structure.
The standalone Sora app had launched in September 2025, initially reaching the top of the iOS App Store’s Photo and Video category within its first day, a debut that set user expectations the product’s revenue trajectory never came close to matching.
Why It Matters
Sora’s closure highlights a structural challenge in consumer AI video generation that extends beyond OpenAI: at current compute prices, video inference costs cannot be recovered through the per-generation or credit-based pricing models that consumer products typically require to achieve mass adoption.
OpenAI’s ChatGPT operates on a recurring subscription model and has reached 900 million weekly active users — a fundamentally different cost-to-revenue structure that Sora’s in-app purchase model never approximated.
Competing video generation tools from Google and independent providers such as Runway face similar questions about sustainable unit economics at consumer scale, making Sora’s closure a data point for the broader AI video market rather than an isolated failure.
Technical Details
Video generation is computationally more intensive than text output because every second of rendered video requires the model to solve spatial reasoning problems — motion, lighting, physics, and temporal consistency — across hundreds of individual frames, making per-output inference costs categorically higher than language model responses.
Forbes, cited by Revolution in AI across multiple industry outlets, estimated OpenAI’s peak daily inference spend on Sora at approximately $15 million, which annualizes to roughly $5.4 billion solely to keep the servers operational.
Against that figure, mobile analytics firm Appfigures tracked Sora’s total in-app purchase revenue at $2.1 million across the product’s entire commercial lifetime — revenue generated through a credit-based system that allowed users to purchase additional video generation capacity. The gap between those two numbers was not closable by any realistic adjustment to the credit pricing model.
Monthly downloads peaked at 3.33 million across iOS and Google Play combined in November 2025, then fell to just over 1.1 million by February 2026 — a decline of approximately 66% in three months. Monthly active users peaked in December 2025 and declined through early 2026.
Who’s Affected
The shutdown removes Sora as an available tool for content creators, marketers, and developers who had built video generation workflows around the platform, and ends a reported $1 billion enterprise partnership with Disney that had been described as a flagship commercial use case for the product.
KeyBanc Capital Markets analyst Justin Patterson noted in a research note, cited by Revolution in AI, that Sora failed to attract and retain an engaged user base despite its strong initial App Store performance — an assessment consistent with the Appfigures download data showing the 66% decline in three months.
What’s Next
OpenAI has not announced a replacement consumer video product, and the cost constraint that drove the shutdown — video inference expenses that outpace what consumer subscription pricing can recover — remains unresolved for the AI video generation industry as of April 2026.
Revolution in AI’s analysis indicates that OpenAI’s internal strategic focus may be shifting toward robotics applications, where video generation technology operates under different deployment and monetization conditions, though OpenAI had not made formal statements confirming this direction at time of publication.
Any future AI video offering from OpenAI would need either significantly reduced inference costs through hardware improvements or model efficiency gains, or an enterprise pricing structure capable of supporting higher per-query costs without depending on consumer volume to bridge the gap.
