Category
Seat, usage, outcome or hybrid — each AI SaaS pricing model decides who absorbs the inference variance. A fractional CFO compares all four through the one lens that protects your gross margin.
Stability AI spent ~$99M on compute to earn ~$11M, burning $8M a month. A forensic CFO autopsy of the burn rate, the AWS bill, and the 2024 down round that turned its runway into a cliff.
Customer-facing AI inference is cost of revenue; internal tools are OpEx; training books separately. A CFO's decision guide with the AI Inference COGS Test, a decision table, and a worked margin example.
Inference, GPU, vector-DB and human-in-the-loop costs behave like COGS, not overhead. Here is how to build the AI startup financial model that layers them, runs the scenarios, and restates LTV:CAC on the real AI-layer margin.
Your blended gross margin hides the AI layer underneath. Learn what AI Layer Gross Margin is, how to calculate it, and why it runs near a third, not 80%.