Sora AI Video Economics: $20 User Costs OpenAI $65 in Compute

Sora's Unit Economics
According to analysis of OpenAI's Sora AI video generation app, each $20/month subscriber generating 50 videos costs OpenAI approximately $65 in compute. The more users engaged with the product, the faster it lost money.
Cost Breakdown
- Peak compute costs: $15 million per day (analyst estimate)
- Total lifetime revenue: $2.1 million across all platforms over six months
- Conservative estimate: ~$1 million daily net burn rate after throttling and optimization
- Cost ratio: Generating 10-second AI video costs roughly 160 times more than equivalent text generation
Technical Details
Each 10-second video costs about $1.30 in GPU compute, requiring approximately 40 minutes of total GPU time across 4 H100s in parallel. At peak usage, with 25% of Sora's 4.5 million cumulative users publishing 10 videos daily, this resulted in 11.3 million videos per day at $1.30 each.
The structural cost issue stems from video's computational requirements: a 10-second 720p video requires processing roughly 80,000 small patches of pixels across space and time, compared to a few hundred tokens for typical ChatGPT prompts. Compute needs grow with the square of patch count - double the resolution quadruples the cost.
Market Context
No AI video company has proven net profitability according to the analysis. Runway reported -$155M EBITDA, Pika had $7.6M revenue on $80M raised, and while Kling reported $240M ARR, they haven't published profit data. Sora showed 1% day-30 retention versus TikTok's 32%, with users typically generating videos, watching them once, and not returning.
📖 Read the full source: HN LLM Tools
👀 See Also

Ubuntu Linux to Integrate AI Features Over the Next Year, Starting with Local Inferencing
Canonical announces a multi-year AI push for Ubuntu, focusing on local inferencing, agentic workflows, and context-aware OS capabilities, with features rolling out throughout 2026.

AI Didn't Delete Your Database — You Did: Accountability in the Age of AI Coding Agents
A viral story blamed an AI agent for deleting a production database, but the real issue is exposing destructive API endpoints and lack of process—not the tool.

UW Researchers Plan to Use Teacher-Worn Cameras for AI Training, Parents Opt-Out
University of Washington researchers planned to have preschool teachers wear first-person cameras to record children for AI model training, with an opt-out consent model.

Ohio Suspends Data Center Tax Break: AI Cost Pressures Mount for Tech Firms
Ohio halts a sales tax exemption on equipment for new data centers, including those powering AI. The move signals growing state-level scrutiny of tax incentives as AI infrastructure demands surge.