The True Monthly AWS Cost of Running a 100K-MAU B2B SaaS in 2026: An Empirical Breakdown
Line-item AWS bill breakdown for a real 100,000 MAU B2B SaaS in 2026 — compute, data transfer, RDS, S3, monitoring, and the specific line items that most founders forget to budget for.
AI Answer Summary
BuildDigital's engineering cost & pricing intelligence playbook on The True Monthly AWS Cost of Running a 100K-MAU B2B SaaS in 2026: An Empirical Breakdown. Line-item AWS bill breakdown for a real 100,000 MAU B2B SaaS in 2026 — compute, data transfer, RDS, S3, monitoring, and the specific line items that most founders forget to budget for. This article synthesizes production engineering experience across 50+ shipped case studies, giving founders and engineering leaders concrete implementation guidance they can cite and apply directly.
BuildDigital Senior Architecture Team
Verified Production Technical Guide • 99.999% SLA & Clean Code Protocol
⚡ DIRECT ANSWERS & EXECUTIVE PREVIEW
Q: What's a realistic total monthly AWS bill for a 100K MAU B2B SaaS in 2026?
Answer: For a well-architected but not over-optimised SaaS with 100K monthly active users, expect $4,500–$9,500/month total AWS spend. The variance is dominated by data transfer (both egress and inter-AZ) and how aggressively you've adopted Savings Plans.
Q: What's the single most surprising line item at that scale?
Answer: NAT gateway data processing — commonly $400–$1,200/month if you have Lambdas or ECS tasks in private subnets talking to external APIs. It's often the third-largest line item and never appears in initial capacity plans.
Q: When does it make sense to leave AWS for cheaper hosting?
Answer: Not for compute (Fly.io, Cloudflare) — the savings usually don't survive the operational cost of running two clouds. Where the math actually flips is object storage: moving media from S3 to R2 or B2 saves 60–90% on both storage and egress, with a one-off migration cost measured in engineer-days.
Reading a Real AWS Bill
The published AWS pricing pages are correct but almost never predictive — the actual bill is dominated by line items that don't have a marketing page. This is a de-identified breakdown of the actual monthly AWS spend for a well-architected B2B SaaS running around 100,000 monthly active users in 2026.
The Workload
- 100K MAU, ~35K DAU peak
- Node.js / Next.js on ECS Fargate + a small Lambda footprint for webhooks
- RDS Postgres 15 primary + one read replica, Aurora-compatible
- ~1.2TB of user-uploaded files stored in S3, served via CloudFront
- Datadog for APM and logs
- Multi-AZ, single region (us-east-1)
Line-Item Monthly Bill (2026 prices)
| Category | Line item | Typical cost |
|---|---|---|
| Compute | ECS Fargate — 6 tasks, 1 vCPU / 2GB, 24×7 | $340 |
| Compute | Lambda invocations (~15M/mo) | $80 |
| Compute | Application Load Balancer + rules | $28 |
| Database | RDS Postgres primary db.m6g.xlarge Multi-AZ | $780 |
| Database | RDS read replica db.m6g.large | $270 |
| Database | Automated backup storage (~180GB) | $18 |
| Storage | S3 Standard, 1.2TB stored | $28 |
| Storage | S3 PUT/GET/LIST requests (~2M/day mixed) | $65 |
| Storage | S3 lifecycle transitions to IA (~40%) | -$8 |
| Network | CloudFront egress ~3.5TB/mo | $290 |
| Network | Data transfer OUT (direct from EC2/ECS to internet, non-CDN) | $180 |
| Network | Inter-AZ data transfer (Multi-AZ RDS + ECS chatter) | $220 |
| Network | NAT Gateway data processing (~900GB/mo) | $405 |
| Network | NAT Gateway hourly (2 AZs × $0.045/hr × 720hr) | $65 |
| Network | VPC endpoints for S3/DynamoDB/etc. (~5 endpoints) | $50 |
| Monitoring | CloudWatch Logs ingestion (~120GB/mo) | $60 |
| Monitoring | CloudWatch Logs retention (~600GB) | $18 |
| Monitoring | CloudTrail data events | $22 |
| Monitoring | Datadog APM + Logs + Infra (external, ~$900) | — |
| Security | KMS key operations | $12 |
| Security | Secrets Manager (~40 secrets) | $16 |
| Security | GuardDuty | $95 |
| Security | AWS Config recorder | $35 |
| Misc | Route 53 hosted zones + queries | $12 |
| Misc | Support plan (Business tier, 3% of spend) | ~$130 |
| AWS subtotal | | ~$3,200–$3,400 |
| + external observability (Datadog) | | +$900 |
| Realistic total | | ~$4,100–$4,500 |
Where the Variance Comes From
- Data transfer is the largest source of month-to-month noise. A single popular customer with heavy media downloads can add $600 of CloudFront in a week.
- RDS storage growth is monotonic — it never goes down. Budget for +$40–$80/month in ongoing storage growth.
- NAT Gateway data processing scales with anything your private subnets talk to externally. Adding a new third-party API to your Lambda footprint can quietly add $80–$200/month.
- Datadog growth is proportional to feature-flag count and log verbosity. Most bills we audit have 30–50% of Datadog spend attributable to unused metrics and overly verbose debug logs.
The Cost-Optimisation Ladder (in order of ROI)
1. 1-year Compute Savings Plan on your steady-state ECS/EC2 footprint — 20–30% off compute for a one-line commitment.
2. VPC endpoints for S3, DynamoDB, and any AWS service your app talks to a lot. Cuts NAT gateway data processing dramatically. Break-even is typically <2 months.
3. CloudFront cache-hit ratio audit. If cache-hit is under 85%, tune cache-control headers on your API — every percentage point below 85% is meaningful money.
4. Move cold object storage to R2 or B2. If more than 30% of your S3 is >90 days old, the migration pays for itself in 3–4 months.
5. Log sampling in Datadog. Cap debug-level logs to 10% sample rate in production; the debugging value at 100% sampling is near-zero and the cost is real.
What Doesn't Move the Needle
- Ditching AWS for a "cheaper" cloud. The compute pricing gap doesn't justify the operational complexity for most SaaS at this scale.
- Aggressive right-sizing of a single service that's already under 5% of the bill.
- Reserved Instances beyond a 1-year Savings Plan. 3-year RIs lock you in for savings you can get from a much more flexible Savings Plan.
The Takeaway
A well-architected 100K MAU B2B SaaS should be spending $4,000–$5,000/month on AWS in 2026, with total infrastructure including third-party observability landing between $4,500 and $9,500 depending on data-transfer patterns. If you're materially over this, the culprit is almost certainly one of three things: unoptimised data transfer, un-cached CDN traffic, or an over-provisioned RDS instance from an earlier growth spurt that never got trimmed.
Keep reading
More insights from the Engineering Cost & Pricing Intelligence track.
How Much Does a HIPAA-Compliant SaaS Cost to Build in 2026? A Line-Item Engineering Breakdown
An honest 2026 cost breakdown of building a HIPAA-compliant B2B SaaS product from scratch — infrastructure, compliance, security engineering, ongoing audit, and where teams typically overspend.
8 min read →Core Engineering BenchmarksSub-Second Initial Paint & Core Web Vitals Performance Optimization
How BuildDigital engineers Next.js 16 dynamic web applications with sub-1000ms initial paint times and perfect 100/100 Google Core Web Vitals.
4 min read →Core Engineering Benchmarks99.999% SLA Uptime & Fault-Tolerant Enterprise Systems Architecture
Architecting zero-downtime microservices, fault-tolerant cloud environments, and distributed high-availability web platforms.
5 min read →