If you've ever looked at your cloud bill and noticed that doubling your user base didn't double your infrastructure cost, you've already observed the most important principle in SaaS economics: sublinear scaling.
This article explains why it happens, quantifies the effect, and shows how to use it for pricing strategy. The numbers are drawn from our cost modelling work across real SaaS and AI product deployments.
The Sublinear Scaling Principle
Infrastructure cost doesn't grow linearly with users. It follows a power law:
infra_multiplier = pow(user_ratio, 0.6)
Where user_ratio is your current user count divided by your starting user count. The exponent 0.6 means infrastructure grows at roughly 60% of the rate you might naively expect.
Why? Because shared resources serve many users simultaneously:
- Databases handle concurrent queries from thousands of users on the same hardware.
- Caches (Redis, CDN) serve the same content to many users without re-computing.
- Load balancers distribute traffic across a pool of servers that doesn't need to scale 1:1.
- Application servers share memory, connection pools and background workers.
The result: each additional user costs less to serve than the last.
Worked Example: 500 to 100,000 Users
Let's start with a medium-complexity SaaS product paying $4,000/month in infrastructure to serve 500 users. Now let's grow to 100,000 users — a 200x increase.
What Linear Scaling Would Predict
If infrastructure scaled linearly (1:1), the cost at 100,000 users would be:
$4,000 × 200 = $800,000/month
What Sublinear Scaling Actually Gives You
With the 0.6 power law:
$4,000 × pow(200, 0.6) = $4,000 × 27.3 = ~$109,200/month
200x users, but only ~27x infrastructure cost. That's a 7.3x efficiency gain over linear scaling. Your cost per user dropped from $8.00 to $1.09.
The Journey from 500 to 100,000 Users
| Users | User Ratio | Infra Multiplier | Monthly Infra | Cost Per User |
|---|---|---|---|---|
| 500 | 1x | 1.0x | $4,000 | $8.00 |
| 1,000 | 2x | 1.52x | $6,063 | $6.06 |
| 2,500 | 5x | 2.63x | $10,520 | $4.21 |
| 5,000 | 10x | 3.98x | $15,920 | $3.18 |
| 10,000 | 20x | 6.03x | $24,120 | $2.41 |
| 25,000 | 50x | 10.46x | $41,840 | $1.67 |
| 50,000 | 100x | 15.85x | $63,400 | $1.27 |
| 100,000 | 200x | 27.30x | $109,200 | $1.09 |
Cost per user dropped from $8.00 to $1.09 — an 86% reduction. This is the compounding benefit of sublinear scaling.
The 40% Floor
Sublinear scaling doesn't mean infrastructure cost shrinks to near-zero. There's a floor: infrastructure never drops below 40% of what linear scaling would predict.
Some resources genuinely scale with users: per-user storage, session state, notification delivery, email sends. Even with perfect caching and shared compute, each user has a minimum resource footprint. Our modelling puts this floor at 40% of the linear rate.
In practice, the floor matters most at very large scale. Here's how it plays out:
| User Ratio | pow(ratio, 0.6) | Linear Cost | Sublinear as % of Linear |
|---|---|---|---|
| 2x | 1.52x | 2.0x | 76% |
| 10x | 3.98x | 10.0x | 40% |
| 50x | 10.46x | 50.0x | 21% |
| 200x | 27.30x | 200.0x | 14% |
| 1,000x | 63.10x | 1,000x | 6% |
At 10x users, the 0.6 power gives you 40% of linear cost — right at the floor. Beyond that, the formula would predict even lower ratios, but in practice you'd see costs level off near the 40% mark. The model applies max(pow(ratio, 0.6), 0.4 × ratio) to capture this.
Cost-Per-User Curve Over 36 Months
What does this look like as your product grows month by month? Let's model a medium-complexity SaaS product ($4,000/month infra base) growing at 8% monthly from 500 users:
| Month | Users | Monthly Infra | Cost Per User | Savings vs Linear |
|---|---|---|---|---|
| 0 | 500 | $4,000 | $8.00 | — |
| 6 | 793 | $5,460 | $6.88 | 14% |
| 12 | 1,259 | $7,260 | $5.77 | 28% |
| 18 | 1,998 | $9,500 | $4.76 | 40% |
| 24 | 3,172 | $12,280 | $3.87 | 52% |
| 30 | 5,034 | $15,720 | $3.12 | 61% |
| 36 | 7,990 | $19,980 | $2.50 | 69% |
By month 36, you're saving 69% compared to what you'd pay under linear scaling. That's real money — the difference between $19,980/month and $63,920/month.
Why This Matters for Pricing Strategy
Understanding the cost-per-user curve has direct implications for how you price your product:
1. Your Margins Improve Automatically
If you charge a flat $15/user/month and your cost per user drops from $8 to $2.50, your gross margin per user goes from 47% to 83%. This happens without any pricing change — growth itself improves your economics.
2. You Can Offer Volume Discounts Profitably
Enterprise customers often demand lower per-seat pricing. Sublinear scaling means you can offer 30–40% discounts to large accounts and still maintain healthy margins, because serving them costs disproportionately less.
3. Lower Tiers Become Viable Acquisition Tools
A free or low-cost tier that seems unprofitable at 1,000 users may become profitable at 50,000 because the incremental cost per user is so low. The scaling curve turns your pricing ladder into a growth engine.
4. You Have a Structural Advantage Over AI-Heavy Competitors
Products with high token costs scale linearly — their cost per user never drops. If your SaaS product competes with an AI-native alternative, your improving unit economics become a moat as both products grow.
Don't set prices based on today's cost per user. Model the curve over 12–36 months. Your current margins are your worst margins — they only improve from here.
Key Takeaways
- Infrastructure grows at the 0.6 power of users — doubling users costs ~52% more, not 100%.
- 200x users = ~27x infrastructure cost — the efficiency gain compounds dramatically at scale.
- The 40% floor sets a lower bound on per-user infrastructure cost, but still represents massive savings vs linear scaling.
- Cost per user drops continuously — from $8.00 to $2.50 over 36 months at 8% monthly growth in our model.
- Price for the curve, not the snapshot — today's margins are your worst margins.
Model Your Infrastructure Costs
Use our interactive Cost Analyser to project your infrastructure costs across different growth rates and complexity tiers.
Open the Cost Analyser →