OpenAI Discontinues Sora, Citing Unsustainable Video Generation Costs

The shutdown signals that not all AI capabilities can scale economically at current pricing.

Illustration: OpenAI Discontinues Sora, Citing Unsustainable Video Generation Costs
AI-generated illustration · Sylvaris

App Shuts April 26, API Follows in September

On March 24, OpenAI announced it will discontinue Sora, its text-to-video generation service. The Sora web app and mobile experience will shut down on April 26, 2026, while the API will remain available until September 24, 2026.

The company cited unsustainable economics as the core reason. According to internal descriptions, the cost per generated minute of high-quality video was described as economically irreconcilable with public API pricing that users would pay.

Peak Users, Falling Revenue

Sora's user base peaked at around one million users before collapsing to fewer than 500,000, while burning through roughly one million dollars per day in compute costs. In-app purchases generated only 2.1 million dollars, far too little to justify the infrastructure expense.

With an IPO on the horizon, OpenAI is redirecting compute resources toward coding tools and enterprise customers where unit economics are stronger. The decision reflects broader pressures: video generation requires massive compute that few companies can afford to subsidize at scale.

What It Means for AI Economics

The shutdown reinforced that AI infrastructure winners will not all rise together. Software optimization can quickly reshuffle which hardware layers capture value, and some AI workloads may never reach economic viability at consumer pricing.

For organizations building on AI vendor platforms, the episode is a reminder that product lifecycles can be short, and reliance on a single provider carries continuity risk.

sources
more in Artificial Intelligence
Text-to-SQL benchmarks fail to address real-world data store complexities AI code generation tools struggle with messy production databases that lack the clean schemas found in test environments. Meta launches Content Seal watermarking system for AI-generated content detection Meta's new invisible watermarking technology addresses platform accountability for AI-generated content, though it remains less accessible than Google's existing SynthID solution. MCP servers fail agent usability testing, one-third score D or F grades Poor server design undermines the Model Context Protocol's promise to standardize AI agent tool access, creating friction in production deployments.