OpenAI confirmed on March 24, 2026, that it is shutting down Sora, its AI video generation tool, with the standalone app closing April 26 and the API shutting down September 24. The decision simultaneously collapsed a $1 billion Disney partnership announced just three months earlier — Disney learned Sora was being terminated less than one hour before the public announcement, according to Deadline.
The Economics That Killed Sora
Sora’s financial profile was unsustainable by any measure. At peak usage, inference costs ran approximately $15 million per day. A single 10-second video clip cost roughly $1.30 to generate. Total lifetime revenue from in-app purchases: $2.1 million. Downloads peaked at 3.33 million per month in November 2025 on iOS and Google Play combined, then collapsed to 1.1 million by February 2026 — a 66% decline in three months.
The Sora team will continue as a research unit focused on world simulation, working on a new model codenamed “Spud” for robotics and physical environment modeling. OpenAI is rerouting the freed compute capacity toward its core language model development, where competitive pressure from Anthropic’s Claude Code has intensified.
Who’s Still Standing
Sora’s exit reshapes the AI video landscape. The remaining competitors and their current capabilities as of March 2026:
- Google Veo 3.1 — True 4K at 60fps with synchronized audio. The current technical leader.
- Runway Gen-4.5 — Professional-grade video editing and generation. Strong creative community.
- Kling 3.0 — Up to 2-minute clips starting at $6.99/month. Competitive pricing.
- Pika 2.5 — Consumer-focused with social sharing features.
- Seedance — ByteDance’s open-source alternative.
The lesson from Sora’s brief life is that AI video generation works technically but fails economically at consumer scale. Generating video requires orders of magnitude more compute than text or images, and no pricing model has yet bridged the gap between what inference costs and what consumers will pay. The companies that survive will be those targeting professional and enterprise use cases where the output value justifies the compute cost.
