Your AI bill is growing.
Who’s driving it?
The Problem
You only see the total AI invoice
Most teams see a single bill from their AI provider each month. They know they're spending real money on AI, but not which customers, features, or pricing tiers are driving that cost.
You're spending $12,000/month on OpenAI but have no idea:
Which specific customers are the most expensive to serve.
Whether free-tier users are burning more credits than paid users.
Which AI features (chatbot, email writer, analytics, etc.) are driving 80% of your cost.
How AI costs compare to the revenue you collect from each account.
What you see today
OpenAI Invoice · April 2026
$12,173.44
Total tokens used: 847,231,000
No breakdown by customer, feature, or tier available.
What PerUnit shows you
| Customer | AI Cost | Margin |
|---|---|---|
| acme-corp.io | $4,234 | +50% |
| startupxyz.com | $3,891 | −46% |
| enterprise.co | $2,108 | +83% |
| free-user-4821 | $1,940 | −100% |
Your AI dashboard shows totals.
PerUnit shows answers.
Your AI provider dashboards tell you how much you spent overall. PerUnit tells you who spent it, which feature drove it, and whether that customer is profitable.
| Capability | PerUnit | AI ProviderOpenAI / Anthropic | ObservabilityHelicone / LangSmith |
|---|---|---|---|
| Cost per customer | |||
| Cost by pricing tier (Free / Pro / Enterprise) | |||
| Cost by AI feature (chatbot, email writer, etc.) | |||
| Revenue vs. AI cost margin (Stripe integration) | |||
| Margin per account | |||
| Direct AI provider API sync (no pipelines) | |||
| Cost alerts by customer | |||
| No data engineering team required |
What PerUnit shows you
Turn AI invoices into unit economics
PerUnit gives you a real-time view of AI spend per customer, per feature, and per pricing tier so you can make clear, defensible product and pricing decisions.
Cost per customer
See exactly how much each account costs you in AI usage, broken down by workspace or tenant.
Cost by pricing tier
Understand if free, Pro, or Enterprise plans are actually profitable once AI costs are included.
Cost by feature
Attribute spend to each AI feature so you know what to gate, bundle, or deprecate.
Revenue vs. AI cost
Connect Stripe revenue to AI spend so you can see margins by customer, tier, and feature.
| Customer | Tier | Top Feature | AI Cost | Revenue | Margin |
|---|---|---|---|---|---|
| acme-corp.io | Enterprise | AI Analyst | $4,234 | $8,400 | +50% |
| startupxyz.com | Pro | Chat Assistant | $3,891 | $2,100 | −46% |
| enterprise.co | Enterprise | Email Writer | $2,108 | $12,000 | +83% |
| growthco.io | Pro | Chat Assistant | $1,640 | $2,800 | +41% |
| free-user-4821 | Free | Chat Assistant | $940 | $0 | −100% |
Showing 5 of 142 customers · Sorted by AI cost · Updated 2 minutes ago
What's leaking in your AI bill?
Plug in four numbers. See how much of your AI spend is going to free users, underpriced heavy users, and unattributed requests — and what you could recover.
Free tier drain
$1,500.00/mo on users paying $0. Capping top consumers recovers ~$525.00/mo.
Underpriced heavy users
Heavy users still within margin at $49.00/mo. No repricing needed yet.
Invisible spend
~30% of your bill ($1,500.00) is unattributed. A 20% optimization recovers ~$300.00/mo.
Total recoverable with PerUnit
$825.00/ month
17% of your AI bill
PerUnit cost
$99/mo
Payback
4 days
Annual net
$8,712.00(733% ROI)
Want the real numbers, not estimates? PerUnit pulls them straight from OpenAI, Anthropic, and Google.
Get early accessEstimates use Pareto-style assumptions (top 20% of paid users drive ~80% of paid AI cost) and conservative recovery rates (35% on free tier, 60% on repricing, 20% on invisible). Your real breakdown is in your provider data — that's what PerUnit shows.
Up and running in days,
not months
Most teams see their first unit economics view within a week of signing up. No data team, no complex pipelines.
Connect your stack
Plug in OpenAI, Anthropic, Google, and Stripe in under an hour. No data warehouse required to get started—just drop in the SDK or point us at your existing logs.
OpenAI · Anthropic · Google · Stripe
Track per customer
Every AI request is tagged by customer ID, feature name, and pricing tier. Cost attribution happens automatically—no custom metering infrastructure needed.
Real-time · Per-request · Automatic
Make confident decisions
See exact margins by customer and feature. Know which plans are unprofitable, which features to gate, and where to optimize before your next board meeting.
Margins · Cost alerts · Pricing insights
Average onboarding time: under 1 week. We work with your team to map data sources and validate attribution before you go live.
Instrument AI usage in minutes
Connect your AI providers and Stripe directly, or instrument individual requests with strongly-typed SDKs—no custom metering platform required.
import { perunit } from "@perunit/sdk";
// Called whenever an AI request is made
await perunit.trackUsage({
customerId: "acct_123",
feature: "chat-assistant",
tier: "pro",
provider: "openai",
model: "gpt-4.1",
tokensInput: 824,
tokensOutput: 1932,
metadata: {
workspaceId: "workspace_45",
messageId: "msg_987",
},
});- Drop-in SDKs
- Client and server SDKs designed for modern AI stacks, from Next.js to background workers.
- Plug into billing
- Connect Stripe to tie AI spend directly to revenue and customer accounts — no manual mapping.
- Pre-built integrations
- Bring in usage from OpenAI, Anthropic, and more without managing brittle custom scripts.
- Security & privacy
- PII-aware data models, envelope-encrypted API keys, and clear data ownership boundaries.
Numbers your finance team can trust
PerUnit connects your AI providers and billing data to give you an always-on view of AI economics—so product, finance, and leadership are finally looking at the same numbers.
- 95%
- Spend attributed
- vs. 0% with your AI provider dashboard alone
- < 1 week
- Time to clarity
- Average from signup to first cost-per-customer view
- Up to 40%
- Cost savings identified
- After surfacing top cost drivers by customer and feature
Based on data from early access teams. Individual results vary by stack complexity and AI spend profile.
Data Security
Your data stays yours.
Always.
PerUnit connects to your AI providers and Stripe account. We take that access seriously. Here is exactly what we collect, what we don’t, and how we protect it.
No prompt or message content
PerUnit only ingests token counts, model names, cost values, and the identifiers you provide (customer ID, feature, tier). Your AI prompts and completions are never sent to us.
You own your data
Your cost and revenue data belongs to you. Export it at any time in CSV or JSON. If you cancel, we delete it within 30 days — no questions asked.
Encrypted in transit and at rest
All data is transmitted over TLS 1.2+ and stored encrypted at rest. API keys for OpenAI, Anthropic, Google, and Stripe are stored with envelope encryption and never logged.
Minimal data footprint
We collect only what is necessary to compute cost attribution: token usage, timestamps, and the customer/feature/tier identifiers you define. Nothing more.
Questions about data handling? founders@perunit.ai
What early access teams discovered
The answers were in the data.
They just couldn’t see them.
These are real patterns from early access teams — shared anonymously with their permission.
“Free tier was consuming 60% of AI budget”
A product team discovered their free users—less than 20% of accounts—were responsible for the majority of their monthly OpenAI spend. They repriced their free plan within two weeks.
Repriced free plan · Recovered margin
“One feature drove 80% of total AI cost”
A product team found a single AI feature—their document summarizer—was responsible for 4 out of every 5 dollars spent on Anthropic, despite being used by only 12% of customers.
Feature gated behind paid tier · Cost halved
“Top 10 accounts cost 3x more than all others combined”
A fintech startup found their heaviest enterprise accounts generated enormous AI costs—but also 80% of revenue. That data justified moving their AI-heavy features exclusively to an Enterprise tier.
New Enterprise plan launched · Margins recovered
Understand your AI costs
before you sign up for anything
Six focused calculators to estimate spend, check margins, and compare models. No account required.
AI Cost Calculator
Monthly OpenAI, Anthropic, or Gemini spend by model and request volume. Single-call, agent, and budget modes.
OpenAI vs Anthropic vs Gemini
Compare every major model side by side at your token volume. Use-case presets to find the cheapest fit.
Token Counter & Cost Estimator
Paste any text to count tokens and instantly see costs across every major OpenAI, Anthropic, and Google model.
AI Margin Calculator
Check AI cost per customer, AI as % of revenue, and how your margins compare to typical SaaS benchmarks.
Context Window Cost
Long context costs more. Per-request cost for 50K, 100K, or 200K-token prompts across all providers.
Image Generation Cost
DALL·E 3, gpt-image-1, and Imagen-style per-image estimates. Monthly and yearly from images per day.
Frequently Asked Questions
Can’t find the answer you’re looking for? Don’t hesitate to email us.
Get early access to PerUnit
We're onboarding a small group of teams who need clarity on their AI unit economics.
$199 $99/mo · Early access offer · No long-term commitment
We'll follow up with a short onboarding call to map your AI provider and Stripe data.