What will your AI product cost to build, run and sell?
Describe your idea or pick a starting point. You get the monthly AI and running costs, a build budget range, the price you need to charge for a healthy margin, and where to save. Every number is editable.
- Claude API prices as of June 2026
- Every assumption editable
- Estimates, not quotes
1. Describe your product
One or two sentences is enough. Who uses it, what the AI does, and roughly how many people.
Or start from a common product
One click fills in typical starting assumptions.
Usage and AI inference
How many people use it, how often, and how much text goes in and out of the model each time.
30,000 interactions a month in total
About 1,500 English words
About 300 English words
Model
Model prices used (Anthropic API, June 2026)
| Model | Input | Output | Cache read |
|---|---|---|---|
| Claude Opus 5 | $5 | $25 | $0.50 |
| Claude Opus 5.5 | $4 | $20 | $0.20 |
| Claude Sonnet 5 | $2 | $10 | $0.20 |
| Claude Haiku 4.5 | $1 | $5 | $0.10 |
| Claude Fable 5.1 | $10 | $50 | $0.25 |
US dollars per million tokens on Anthropic's own API. 5-minute cache writes cost 1.25x the input price. Cache reads cost 0.1x input on most models, and less on Claude Opus 5.5 and Claude Fable 5.1 as shown. The Batch API takes 50% off every token.
Prices change. Check the provider's pricing page before you rely on them: anthropic.com/pricing
Cost optimizations
Search over documents (RAG)
AI cost per month
$212
AI cost per user
$0.2122
AI cost per interaction
$0.00707
Tokens per month
79.2M
$51.81 saved by the options below
Other running costs
Everything besides the AI model that you pay for each month.
The defaults are a typical small-scale starting point, not real prices. Replace them with your own quotes.
Features that cost money per use
Infrastructure (per month)
per month
per month
Turn on document search (section 2) to count this.
per month
per month
Email and payments
People (per month)
per month
per hour
per month
per hour
Running costs besides AI
$1,324
Usage-based
$4
Fixed each month
$870
Payment fees
$450
Build cost (one time)
What it takes to design, build, test and launch it. Hours are rough estimates.
What you need to build
The core product (app, AI integration, prompts, deployment) is 120 h. Each feature adds the estimated hours shown.
Hourly rate
per hour
Quick set
These buttons are only shortcuts, not market rates. Set the rate you have actually been quoted.
Design, testing and risk
One-time amount from your quotes
Hours are rough planning estimates for a small team using common frameworks and managed services. Your project can differ a lot. Get a real quote before you budget.
Low
$11,000
Likely
$16,500
High
$24,750
Low assumes no surprises and no contingency. High allows for scope growth (1.5x the likely total).
Pricing and unit economics
What you charge, what you keep, and when it pays back.
Monthly revenue
$15,000
per month
Gross margin
93%
After cost to serve
Monthly profit
$13,464
After all running costs
Contribution per user
$14.33
Revenue minus usage costs
Break-even users
61
Active users to cover fixed costs
Build cost paid back in
1.2 months
Recovering $16,500 from monthly profit
Suggested price for a 70% gross margin
$2.17 to $3.45 per user / month
At the current number of users. The lowest price that covers all running costs is $1.12 per user / month. Compare with what your customers pay for alternatives today.
A heavy user (5x average usage)
Usage cost of one heavy user
$1.08
What they pay you (after fees)
$14.55
No warnings: the margin meets your target and a heavy user still pays for themselves.
Scenarios: launch, growth, scale
The same product at different sizes. Change the multipliers to match your plan.
| Item | Launch | Growth | Scale |
|---|---|---|---|
| Users (x today) | x | x | x |
| Fixed costs (x today)Servers, support and maintenance grow too | x | x | x |
| Active users | 1,000 | 10,000 | 50,000 |
| AI inference and embeddings | $212 | $2,122 | $10,610 |
| Voice, messages, images, storage, email | $4 | $40 | $200 |
| Infrastructure | $120 | $240 | $480 |
| Support and maintenance | $750 | $1,500 | $3,000 |
| Payment fees | $450 | $4,500 | $22,500 |
| Total monthly cost | $1,536 | $8,402 | $36,790 |
| Monthly revenue | $15,000 | $150,000 | $750,000 |
| Gross margin | 93% | 95% | 95% |
| Monthly profit | $13,464 | $141,598 | $713,211 |
Usage costs grow in line with users. Prices per token and per unit are assumed not to change; volume discounts would improve the larger scenarios.
Where you can save
Calculated from your numbers. Each saving assumes quality stays acceptable, so test before you switch.
Route simple requests to Claude Haiku 4.5
Save $53.05 / monthSending 50% of requests (greetings, short lookups, classification) to Claude Haiku 4.5 saves about $53.05 a month. Check answer quality on real traffic first.
Ask for shorter answers
Save $33 / monthOutput is most of your AI bill. Cutting average answer length by 25% saves about $33 a month.
Batch the work that can wait
Save $26.52 / monthIf 25% of your work can wait a few hours (nightly summaries, bulk generation), the Batch API's 50% discount saves about $26.52 a month.
Your biggest cost is not AI
$750 / monthSupport and maintenance is $750 a month, 49% of running costs. That is where negotiating, automating or better tooling pays off most.
Savings overlap, so they do not simply add up. Measure on real requests before changing models or prompts.
How the calculator works
AI cost = interactions x tokens x price per token, plus extra tokens for retries (overhead). Interactions = active users x interactions per user.
Cacheable input is billed at the cache read price, except the share marked as cache misses, which is billed as a 5-minute cache write (1.25x input). Batched traffic gets 50% off every token. Routed traffic is billed at the cheaper model's prices.
Usage-based costs (voice, messages, images, storage, email, embeddings) grow with users. Hosting, databases, monitoring, sign-in, support and maintenance are fixed each month and grow with the fixed-cost multiplier in the scenarios.
Gross margin = (revenue - cost to serve) / revenue. Cost to serve is every running cost except maintenance engineering, which is counted in monthly profit instead.
Break-even users = fixed monthly costs / contribution per user, where contribution = revenue per user - usage costs per user - payment fees.
The suggested price makes gross margin equal your target at the current number of users. The top of the range aims 10 points higher.
Build hours are rough planning estimates. Low = 80% of the cost before contingency. High = 1.5x the likely total.
This is a planning tool, not a quote. Real costs depend on your design, vendors, taxes and exchange rates. Check pricing pages and get quotes before you commit.
Want a real quote?
Talk to TIBLOGICS about your product. We will check these assumptions with you and give you a real build plan and price.