OpenAI’s Q1 2026: $5.7 billion revenue (3x YoY). $3.7 billion cash burn (3x YoY). For every dollar earned, $1.65 spent. The ratio didn’t improve.
With $73 billion in cash, OpenAI has ~5 years of runway. The IPO pressure eases β but the unit economics question remains: can inference revenue grow faster than compute costs?
Why this matters:
- $5.7B revenue (3x YoY) but $3.7B cash burn (also 3x YoY). Ratio didn’t improve
- $73B cash on hand = ~5 years runway. IPO pressure eased but economics unproven
- Microsoft exploring DeepSeek as cheaper model. The cost pressure is industry-wide
The Dynamo Doctrine’s factory multiple: AI factories are fab-like, not SaaS-like. Negative margins during buildout phase β same arc as electrification.









