Tool

AI Cost & Margin Calculator

What AI costs this business today, what it costs at plan volume, and how much of that is recoverable

Inference is a variable cost. It scales with usage, while most software pricing scales with seats or contracts. Where usage grows faster than revenue, AI's share of revenue rises and the difference comes out of gross margin — so growth amplifies the problem rather than fixing it. A company can look healthy at current volume and be structurally unprofitable at three times the volume, and none of that is visible in the current-period financials.

This models both points on that curve. Enter today's figures and the growth assumptions in the plan.

Inputs

$
$

Model provider bills, GPU hosting, and any human review of AI output — that review is a cost of revenue, however it is booked.

×
×

Equal multiples describe usage-based pricing, where cost and revenue move together. Usage growing faster than revenue is what per-seat and flat pricing look like on an AI product.

%

Optimization ranges are blended benchmarks, not a quote. What is actually available depends on how much has already been captured — which is the first thing an assessment establishes.

Method

Where these numbers come from

The mechanism behind the margin projection is in When AI-Powered Quietly Breaks Gross Margin, the tiering estimate in Model Tiering: The Fastest AI Cost Win Available, and the benchmarking approach in What Should AI Cost?. For the non-AI half of the same infrastructure estate, use the cloud savings calculator.