OpenAI Shuts Down Sora: $15M Daily Inference Costs Expose Fundamental Economics Problem
Key Takeaways
- ▸Sora's $15M/day inference cost versus $2.1M lifetime revenue demonstrates video generation's unviable economics at consumer scale
- ▸User engagement collapsed 66% in three months (November 2025 to February 2026), from 3.33M to 1.1M monthly downloads, signaling market rejection before official shutdown
- ▸Video generation requires fundamentally different compute economics than text-based AI due to per-frame rendering, motion physics, and temporal consistency requirements
Summary
OpenAI discontinued its Sora video generation service on March 24, 2026, after the product became economically unsustainable. The shutdown reveals a stark financial reality: Sora was burning approximately $15 million per day in inference costs at peak operation, while generating only $2.1 million in total lifetime revenue from in-app purchases. This represents an annualized cost of around $5.4 billion against essentially zero meaningful revenue, making the business model fundamentally broken from inception.
The failure was compounded by severe user engagement collapse that preceded the official shutdown. While the standalone Sora app achieved top rankings in the iOS App Store upon launch in September 2025, downloads plummeted 66% within three months—from 3.33 million monthly downloads in November 2025 to just 1.1 million by February 2026. Monthly active users similarly peaked in December 2025 before declining through early 2026, demonstrating that market demand could not support the product's cost structure.
The shutdown underscores a critical challenge in generative AI economics: video generation's computational intensity makes it fundamentally different from text-based AI products like ChatGPT. While ChatGPT supports 900 million weekly active users through subscription revenue, Sora could never reach comparable scale at current compute prices. OpenAI's own head of Sora acknowledged the reality, stating that "the economics are completely unsustainable."
- The shutdown reflects broader challenges in monetizing computationally expensive generative AI products without enterprise-scale deployment or significant pricing increases
Editorial Opinion
Sora's failure represents a cautionary tale about the gap between technological capability and economic viability in generative AI. While the video generation quality may have impressed observers, the underlying math was always unforgiving—a $5.4 billion annual cost structure cannot be sustained by consumer in-app purchases. This shutdown should prompt the industry to reconsider which AI applications are genuinely viable at scale versus which are demos in search of a business model.
