Skip to main content
AI Biz Hub
Menu

Run the numbers

AI Product Margin Calculator

Calculate per-user margin for AI products from subscription price, API token costs, hosting, and other per-user expenses. See margins at 100, 1K, and 10K users.

Your numbers

Start from an example

$
/day
tok
tok
$
$

Result

Status: in range.Gross margin
74.7%
Cost per user per month
$7.35
Status: in range.Margin per user per month
$21.65
AI API share
89.8%
Key insight

AI API is 89.8% of your per-user cost ($6.60 a month). At 10K users, that is $66,000 a month on API alone. Caching responses or using a tiered model approach could significantly improve margins.

Margin at scale

Total monthly profit at 100, 1K, and 10K users.

100 users
$2,165/mo
1K users
$21,650/mo
10K users
$216,500/mo

Dominant cost driver: AI API. Token costs are list prices — adjust to reflect prompt caching or negotiated rates if applicable.

Figures from Anthropic, OpenAI and Google Gemini, checked on May 25, 2026.

General business information, not legal, tax or financial advice. The result comes only from your inputs and the published method. How we check our numbers.

How to use it

  1. Enter your subscription price per user per month. This is the revenue side of the margin equation and the number that determines whether your AI product is financially viable at scale. AI product pricing balances two pressures: token costs create a per-user cost floor that traditional SaaS does not have, while customers compare your price with the software they already pay for. A premium over non-AI alternatives holds only while the AI functionality delivers clear, measurable value.
  2. Configure your AI API usage with four critical parameters: average API calls per user per day, average input tokens per call, average output tokens per call, and your model's pricing per million tokens for both input and output. These four numbers determine your dominant cost driver and are where most AI product founders make critical estimation errors. Token economics vary dramatically by model: Claude Sonnet 5.5 costs $2 per million input tokens and $10 per million output tokens, Claude Haiku 5.5 costs $0.10/$0.50, Claude Opus 5.5 costs $4/$20, and GPT-5.4 Mini costs $0.75/$4.50. A single user interaction averaging 1,500 input tokens and 800 output tokens costs approximately $0.011 with Claude Sonnet 5.5 but only $0.0047 with GPT-5.4 Mini. At 10 calls per user per day, the monthly per-user AI cost is approximately $3.30 with Claude Sonnet 5.5 versus $1.42 with GPT-5.4 Mini, which is a meaningful difference in unit economics at a $10/month price point.
  3. Add hosting cost per user per month and any other per-user variable costs including payment processing fees (at Stripe's standard US card rate of 2.9% + $0.30, a $10/month subscription pays about $0.59, 5.9% of revenue), customer support tool costs per user, third-party API costs beyond the primary AI model, and storage costs for user data and conversation histories. The total per-user cost across all categories determines your gross margin floor. If the margin left cannot pay for support, marketing, development and your own pay, the fix is a higher price, a cheaper cost structure, or a use case customers value more.
  4. Read the gross margin percentage, cost and margin per user, the AI API share of per-user cost, and the dominant cost driver named under the chart. When the AI API share is the largest part of per-user cost, model selection and prompt design are the highest-impact levers. Common techniques include semantic caching (storing and reusing responses for similar queries), model routing (using cheap models like GPT-5.4 Mini or Gemini 3.5 Flash-Lite for simple queries and expensive models like Claude Sonnet 5.5 only for complex ones), prompt engineering to reduce input token count (shorter system prompts, fewer few-shot examples), and batching multiple user interactions into single API calls where the product design allows it.
  5. Check the margin-at-scale chart showing monthly profit at 100, 1,000, and 10,000 users. AI product margins generally remain stable across scale because both revenue and costs scale linearly per user, unlike traditional SaaS where margins improve with scale as fixed costs are amortized. Negotiated model or payment-processing rates and larger hosting plans can lower per-user cost at scale, but plan on list prices. If your margin at 10,000 users cannot carry your operating costs, the primary remediation strategies in order of impact are: switch to a cheaper model tier for routine queries, implement response caching, optimize prompt token usage, raise prices, or renegotiate vendor contracts.

Questions people usually ask

How is AI API cost per user calculated?

The formula is: monthly AI cost per user = (input tokens per call x input price per million + output tokens per call x output price per million) x calls per user per day x 30 days. For example, with Claude Sonnet 5.5 at $2/M input and $10/M output, an interaction averaging 1,000 input tokens and 500 output tokens costs ($2 x 1,000/1,000,000 + $10 x 500/1,000,000) = $0.007 per call. At 10 calls per user per day, monthly cost is $0.007 x 10 x 30 = $2.10 per user. With GPT-5.4 Mini at $0.75/$4.50, the same interaction costs $0.0030 per call, or $0.90/month per user. That 2.3x difference is why the model choice is the first number to get right.

What gross margin should AI-powered products target?

Enough to pay for everything below gross margin: support, marketing, development and your own pay. There is no published benchmark for AI-product gross margin we would stand behind, so work it backward from your operating costs: if they take half of revenue, a 50% gross margin leaves nothing. As a reference for software without per-call model costs, Damodaran's January 2026 data puts gross margin for listed US system and application software companies at 71.7%. Per-user API cost is what pulls an AI product below that, and this calculator shows by how much.

What cost optimization strategies work best for AI products?

Four strategies in order of typical impact: (1) Model routing: use cheap models (Claude Haiku 5.5, GPT-6 Luna, Gemini 3.5 Flash-Lite) for simple, routine queries and expensive models (Claude Sonnet 5.5, GPT-5.5) only for complex tasks requiring higher capability; the saving depends on how many of your requests the cheap model can handle. (2) Semantic caching: store and reuse responses for similar queries using embedding similarity, which pays when your users ask similar things. (3) Prompt optimization: reduce system prompt length, minimize few-shot examples, and use structured output formats to reduce both input and output tokens. (4) Provider discounts: use batch processing for requests that can wait and prompt caching for long repeated prefixes, both of which the major providers price below their standard rates.

Why is AI product margin different from traditional SaaS margin?

Traditional SaaS has near-zero marginal cost per user: once the software is built and hosted, each additional user costs fractions of a penny in compute and bandwidth. AI products have meaningful marginal cost per user because every AI-powered interaction consumes API tokens that have a real, linear cost. This fundamentally changes the economics: a traditional SaaS product at $10/month with $0.10/user marginal cost has 99% contribution margin, while an AI product at $10/month with $3/user API cost has 70% contribution margin. This 29-percentage-point difference compounds across the P&L, affecting how much you can spend on customer acquisition, how quickly you reach profitability, and your ultimate valuation multiple.

How do token costs trend over time and should I factor this into planning?

Prices fall fastest for a fixed level of capability, and much more slowly at the frontier. The original GPT-4 launched in 2023 at $30/$60 per million tokens (input/output); in 2024 the smaller GPT-4o-mini listed at $0.15/$0.60, 1/200th of that input price, and the legacy GPT-4o was $2.50/$10. Flagships fell less: Claude Opus 5.5 lists at $4/$20 and GPT-5.5 at $5/$30, 6 to 7.5 times cheaper than the original GPT-4 on input and 2 to 3 times on output after three and a half years, while budget tiers such as Gemini 3.5 Flash-Lite ($0.30/$2.50) sit far below. For planning, model today's list prices and treat a future cut as upside rather than as part of the margin plan: a product that needs a price cut to be profitable is not profitable yet. The Model Price Drop Stress Test shows what a given cut would do to your margin.

How does payment processing affect AI product margins?

Card processing charges a percentage plus a fixed fee per transaction, and the fixed part hits low prices hardest. At Stripe's standard US card rate of 2.9% + $0.30 (it differs by country, for example 1.5% + EUR 0.25 for standard EEA cards), a $10/month subscription pays $0.59, 5.9% of revenue, and a $5/month subscription pays $0.445, 8.9%. For AI products whose margins are already squeezed by API costs, that can be the difference between viable and not. Annual billing helps most: one transaction instead of twelve pays the fixed fee once a year.

Do AI product margins improve at scale?

In this calculator's linear model, per-user margin stays constant because all modeled costs are per-user. In practice, per-user costs can fall at scale if you negotiate model or payment-processing rates below list, or if hosting gets cheaper per user on larger plans; none of that is guaranteed, so plan on list prices. Additionally, fixed costs not modeled here (base hosting, monitoring subscriptions, SaaS tools) are amortized across more users, improving overall company-level margins even if per-user variable margins are flat.

Is this tool free and private?

Yes. All calculations run entirely in your browser. No data is sent anywhere. No signup or account required.

Is this professional business or financial advice?

No. Outputs are planning estimates based on the prices you enter. Actual margins depend on usage patterns, negotiated rates, caching effectiveness, and pricing strategy details not fully captured by any calculator.

The same numbers as a table

All articles

More in AI product economics

Understand the costs, margins, and pricing of building AI-powered products.

The AI product economics guide

Use it from code

The same calculation as a JavaScript module you can import. It runs where you import it, with no request, key or rate limit.

import { compute } from "https://aibizhub.io/engines/ai-product-margin-calculator.js";

Input and output contract, TypeScript types and the guide for agents.