Gera Tools

AI Cost-per-User Calculator

Calculate your LLM cost per monthly active user

Enter total monthly API spend and your active user count to compute LLM cost per user, with a free-vs-paid tier breakdown and a SaaS gross-margin impact analysis against your ARPU. It runs free in your browser on Gera Tools, with nothing uploaded.

Last updated Source: Gera Tools

What is AI cost per user?

It is your total LLM API spend divided by your active users for the same period. It tells you how much each user costs you in model inference, which is a core input to SaaS unit economics.

AI cost-per-user calculator

Turn a single API invoice into real unit economics. Enter your monthly API spend, your active users, the fraction who pay, and your ARPU, and the calculator returns the blended cost per user, the sharper cost per paying user, and exactly how much of each paying user’s revenue is eaten by model inference.

How it works

Two cost figures matter:

blended_cost_per_user = api_spend / total_MAU
cost_per_paid_user    = api_spend / (total_MAU × paid_fraction)
margin_impact         = cost_per_paid_user / ARPU

The blended figure flatters you because free users dilute it. The paid figure is the honest one: it concentrates the whole bill onto the people actually paying, which is what determines whether the product makes money. Comparing cost_per_paid_user to ARPU shows the percentage of revenue lost to AI before hosting, support or payment fees.

Why the free-to-paid split changes everything

Consider a product with $5,000 in monthly API spend, 10,000 MAU, and 10% paying users.

  • Blended view: $5,000 / 10,000 = $0.50 per user. Looks manageable.
  • Paid-user view: $5,000 / 1,000 = $5.00 per paying user.
  • If ARPU is $20/month, AI costs 25% of revenue before any other costs.

That 25% eats deeply into gross margin once you add hosting, payment processing fees, and support. The blended $0.50 hides the problem until you look at it from a unit-economics perspective.

What drives cost-per-user up

Several usage patterns can push cost-per-paying-user higher than expected:

Heavy free-tier usage. If free users use the product intensively, they generate API spend but zero revenue. A product with a generous free tier and a low conversion rate can produce a cost-per-paying-user that looks fine in aggregate but is unsustainable per cohort.

Feature mix skewed toward expensive tasks. Not every feature has the same AI cost. A product where the cheapest feature (quick classification) is used heavily by free users, and the expensive feature (long document analysis) is used by paid users, may have a lower blended cost but a high cost on the paid tier that matters most for margin.

Low model utilisation efficiency. Calling a frontier model for tasks a smaller model could handle well drives cost up across all users. Many applications can route 80% of queries to a cheaper model and only escalate complex ones.

Rules of thumb for AI SaaS economics

AI cost as % of ARPUAssessment
Below 5%Comfortable margins even after stacking other COGS
5–15%Typical range for healthy AI products; manageable
15–25%Tight; requires discipline on model routing and free-tier limits
Above 25%Difficult to reach acceptable gross margins without significant ARPU increase or cost reduction

Tips for healthy AI unit economics

  • Target single-digit ARPU percentage. If AI costs exceed 15–20% of ARPU, margins get thin fast once other COGS stack up.
  • Cap free-tier usage. Free users with no rate limits can quietly dominate spend and wreck your blended cost.
  • Route by tier. Serve free users a cheaper model and reserve premium models for paying tiers to protect the paid-user margin.
  • Re-run after pricing changes. A price increase that raises ARPU by 20% dramatically improves AI cost as a percentage even if API spend stays flat.