Unit Economics

Per-User Cost Model

Real unit economics at every scale

9 Enterprises • March 26, 2026

This model tracks actual API consumption, infrastructure, and voice costs from production usage and projects them forward at 10x, 100x, and 1,000x user scale. Margins improve dramatically with volume — model tiering, caching, and shared infrastructure compress per-user cost from ~$115/month at 1 user down to $8-20/month at 1,000.

Scale Breakdown

Cost per user at each scale tier

Tier 1 — Current
Single User (Jasson)
Estimated Total
$80–150
/ month / power user
Line ItemBasisCost / Month
Claude API — Opus (terminal)~500K tokens/day$50–80
Claude API — Haiku (OC/hub)~50K tokens/day$5–10
Claude API — Agents & sprintsVariable$15–30
Twilio — VoiceCalls + number$5–15
ElevenLabs — TTSUsage-based$5–10
HostingMac (current) or VPS$0–4
Total$80–150

Full-power, no optimization. Founder-mode usage with Opus for all heavy lifting.

Tier 2
10 Users
Estimated Total
$30–60
/ user / month
DriverMechanismImpact
Model tieringHaiku for 80% of queries-60% API cost
Shared infrastructureVPS cost split 10 ways-$0.40/user
Voice poolingShared Twilio number-$1–3/user

Key lever: Routing 80% of queries to Haiku at $0.0008/1K tokens vs. Opus at $0.015/1K — an 18x cost reduction per token on routine tasks.

Tier 3
100 Users
Estimated Total
$15–35
/ user / month
DriverMechanismImpact
Multi-tenant infraSingle cluster, per-user isolation-80% infra/user
API cachingSemantic cache for repeated queries-25–40% tokens
Anthropic volumeUsage-based discounts begin-10–20% API

Key lever: Semantic caching eliminates redundant API calls. Common queries (guideline lookups, standard Q&A) hit cache instead of the model.

Tier 4
1,000 Users
Estimated Total
$8–20
/ user / month
DriverMechanismImpact
Volume API pricingNegotiated enterprise rate-20–30% API
Aggressive cachingCache hit rate 50–70% at scale-50% tokens
Infra amortizationFixed costs divided by 1K<$0.10/user

Key lever: At 1,000 users, Anthropic enterprise pricing + 60% cache hit rate brings effective per-token cost below $0.005/1K on average across model tiers.

Product-Specific Cost Breakdown

Each product has a distinct cost profile based on query volume and model usage

AI Underwriter
$0.01–0.03
per query
$83–222/lender/month at full utilization (8 hrs/day, 22 days/mo). RAG + Haiku primary, Sonnet for edge cases.
Jules
$3–5
per user / month
Haiku for all responses + Twilio SMS/voice. Low token count per interaction. Scales nearly flat.
AiNFL GM
$0
per user
Static site. Ad-supported. No per-user API cost. Server cost shared across all traffic.

Competitor Pricing Comparison

Where 9 Enterprise sits in the market

Product Category Price / Month Notes
Devin AI software engineer $500 500 ACUs included. Usage-based above that.
ChatGPT Pro AI assistant $200 Unlimited o1 + advanced features. Consumer-grade.
Cursor AI code editor $20–40 Dev tool only. No workflow automation.
GitHub Copilot AI code completion $19 Single-purpose. IDE-only context.
9 Enterprise (at scale) AI operating system $15–60 Full workflow automation, voice, comms, domain AI. Multi-product.

Unit Economics at a Glance

7x
Cost Reduction
1 to 1,000 users
$3–5
Jules floor
per user/mo
$0.02
AI Underwriter
per query avg

Model tiering, semantic caching, and shared infrastructure are the three levers that make this business viable at scale.

Prepared by 9 Enterprises • March 26, 2026