Most indie hackers I know treat financial modeling like a dentist appointment — uncomfortable, avoidable, and easy to postpone until something actually hurts. The problem: by the time it hurts, you've already burned three months building something that was never going to be profitable. I've seen this pattern too many times. Someone ships an AI SaaS, gets 50 signups, and then realizes their OpenAI bill eats 40% of revenue at scale. Or they price at $9/month because "that feels right," not because the math works. This article gives you the framework to run the numbers before you write a single line of code. Why AI SaaS Changes the Math Entirely Traditional SaaS had a brutally simple unit economics model: infrastructure costs were nearly fixed, so margins improved as you scaled. AI SaaS breaks this assumption. Every request to GPT-4o, Claude, or Gemini is a variable cost that scales linearly with usage. If your power users hammer your app, your gross margin doesn't improve — it can actually deteriorate .…