AI unit economics · free calculator
AI App → Profit
Your best customer might be your most expensive one. See what heavy usage does to your price, margin and limits.
Hypothetical USD examples. No live provider-price feed. Your numbers stay in this browser until you send a mentor chat.
One plan, one month
Usage and AI cost
Fees, other costs and target
Contribution margin · before overhead
$2,000.00 monthly revenue leaves $1,295.00 after the entered variable costs. After $100.00 overhead: $1,195.00, before acquisition costs and tax.
The average can hide a loss.
Each segment includes its share of free-user AI costs. This is a monthly snapshot, not a retention or usage forecast.
Typical user
100 completed tasks / month
contribution / user
Heavy user
1,000 completed tasks / month
contribution / user
| Monthly line item | USD |
|---|---|
| Revenue | $2,000.00 |
| AI costs (paid + free) | $525.00 |
| Payment / platform fees | $130.00 |
| Other paid-user service costs | $50.00 |
| Contribution before overhead | $1,295.00 |
| Entered overhead | $100.00 |
| Remainder before acquisition and tax | $1,195.00 |
The cap is a cost ceiling at your target margin, not a recommendation about customer value. It holds free usage and provider cost constant. If no payers exist, free costs cannot be allocated. If fees plus target margin leave no room, the target price is unavailable.
Design an offer you can sustain
Where should the usage limit sit?
Take the economics to Jon Yongfook and Marc Louvion’s AI mentors. Review packaging, user value and what to measure before changing the price.
Review my price and usage cap ↗A reproducible example
$20 is the price.
Usage decides the margin.
At $0.02 per attempt and 1.25 attempts per completed task, a task costs $0.025. A 90% / 10% mix of 100-task and 1,000-task payers averages 190 tasks. For 100 payers plus 400 free users doing five tasks each, total AI cost is $525.
With $130 payment fees and $50 other service costs, $2,000 revenue leaves $1,295 contribution: 64.75%. Another $100 overhead leaves $1,195 before acquisition and tax. These are invented inputs, not provider prices.
Use real billable costs
Choose a representative usage period. Divide actual AI costs by billable attempts, then include attempts per completed task. If you instead divide the invoice by completed tasks, use an attempts factor of 1 to avoid counting retries twice. Include model calls, tool fees, generated media and storage in the appropriate inputs. Different task types may need separate runs.
Anthropic’s pricing documentation separates input, output, caching and other feature charges. Check the current invoice for every provider you use. This calculator does not call a model or fetch prices.
Price and usage-cap formulas
Target price = (average AI cost + allocated free-user AI cost + other per-payer service cost + fixed payment fee) ÷ (1 − percentage fee − target margin). The cap spends only the remaining per-payer budget on completed tasks, rounding down. Both assume the same mix and cost per task. A target with no available budget is shown as unavailable, not zero-cost usage.
Limits of this monthly snapshot
One plan, one currency, one monthly payment per payer. No trials, refunds, discounts, annual billing, tier migration, demand response or retention forecast. The fee is an assumption, not a universal processor rate. Include other costs in the overhead field or calculate separately. Contribution margin is not net profit; the remainder still excludes unentered expenses and tax.
Before offering “unlimited”
Work through the heavy-user margin and usage-cap example. It separates a cost ceiling from a useful offer.
Find the actual useful task before measuring cost per task: measure the second real API job. To compare how a campaign recovers acquisition costs, use Ads → Payback.