OpenAI shut down its standalone video generation app Sora on March 24, 2026, without providing an official reason at the time of announcement. The product initially gained traction after its debut in February 2024, which showcased high-quality video clips generated by AI.
Sora’s official account posted the following to X: “We’re saying goodbye to the Sora app. Thank you to everyone who created with Sora, shared it, and built a community around it. What you created with Sora was important, and we know this news is disappointing.”
Sora’s demo video showed a woolly mammoth moving through the snow, a cinematic street scene in Tokyo, and complex visuals generated solely from text prompts. The clip received a lot of attention across the technology and entertainment industries.
Film director Tyler Perry said he is reconsidering his plans for an $800 million expansion of the studio after seeing the capabilities of this model. In September 2025, the company released Sora 2 as a standalone app with improved motion, audio, and social sharing feeds, briefly taking the top spot in the app store rankings.
Lower usage, lower retention, higher computing costs
The early momentum didn’t last. According to Appfigures, the number of downloads decreased from approximately 3.33 million across iOS and Google Play in November 2025 to approximately 1.13 million by February 2026, a decrease of approximately 66%.
Appfigures also reported lifetime gross revenue from in-app purchases of approximately $2.1 million, but retention remained low.
The economics of video generation was a key constraint. Video generation requires much more computing than text because each frame must account for movement, lighting, and visual consistency.
A Forbes report estimates that producing a 10-second clip costs about $1.30 and requires significant GPU usage. At peak usage, this equates to approximately $15 million per day, or approximately $5.4 billion per year in computing costs. In comparison, total lifetime income remained limited.
Bill Peebles, head of Sora at OpenAI, said on social media in October 2025 that the economic situation is “completely unsustainable.”
According to the Wall Street Journal, the system was consuming about $1 million every day. All GPUs running Sora requests were unavailable for ChatGPT queries, coding tools, or enterprise API calls. All of these directly generate revenue. Fiji Simo, OpenAI’s CEO of applications, told employees that the company could no longer afford to pay for what she called “side quests.” Sam Altman called for shutting down Sora and redirecting computing to higher priority products.
Partnerships, moderation issues, and next steps
The closure also affected the company’s $1 billion partnership with the Walt Disney Company, which had plans to integrate the franchise’s content. According to the Wall Street Journal, Disney was informed shortly before the announcement and no funds were exchanged.
The app faced content management challenges during production. According to the Washington Post, Martin Luther King Jr.’s estate has asked OpenAI to restrict the creation of videos using his likeness. Users also generated unauthorized videos of Robin Williams, prompting complaints from his estate.
OpenAI said the Sora team will continue as a research unit focused on world simulation with applications in robotics and physical environment modeling. The company said the standalone app and sora.com will be shut down on April 26, 2026, followed by the API on September 24, 2026, after which user data will be permanently deleted. The Sora 2 model remains available within ChatGPT’s paid tier.
OpenAI is also preparing for an initial public offering at a reported valuation of about $300 billion, and the shutdown will see the retirement of a product that is estimated to spend about $5.4 billion a year on computing and generate about $2.1 million in lifetime revenue.
Competitors such as Google DeepMind (Veo), Runway, and Luma AI continue to develop video generation tools. However, ByteDance’s Seedance platform has paused expansion due to copyright concerns. This closure shows that while AI video generation technology is advancing, its feasibility as a large-scale consumer product remains constrained by cost and infrastructure limitations.
