Cloud Cost Optimization: How We Cut a Client's AWS Bill by 35%
Real techniques and architectural changes that reduced a FinTech startup's monthly AWS spend from $48K to $31K without sacrificing performance.
Cloud costs spiral quietly. What starts as a manageable $5K/month infrastructure bill becomes $48K before anyone notices. This is the story of how we helped a FinTech client reclaim $204K in annual AWS spend.
The Situation
When they came to us, the client was spending $48,000/month on AWS infrastructure supporting a transaction processing platform with ~200K daily active users. Not unreasonable on the surface — but a thorough audit revealed significant waste.
The Audit Findings
After three days of analysis using AWS Cost Explorer, Compute Optimizer, and Trusted Advisor:
Over-provisioned EC2 instances — 40% of compute resources were running at < 10% CPU utilization. The team had provisioned for peak Black Friday traffic and never right-sized afterwards.
Data Transfer costs — $8K/month in cross-region data transfer that could be eliminated with proper VPC endpoint configuration and CloudFront caching.
RDS multi-AZ for dev/staging — Running production-grade database setups in non-production environments. Pure waste.
Unused Elastic IPs and snapshots — $1,200/month in forgotten resources from deprecated services.
S3 storage class mismanagement — 2TB of data in S3 Standard that was accessed < once per month. Should be in Glacier or Intelligent-Tiering.
The Optimization Plan
Phase 1: Quick Wins (Week 1-2)
- Right-size EC2 instances using Compute Optimizer recommendations
- Delete 847 unused EBS snapshots and 12 unattached Elastic IPs
- Move cold S3 data to Intelligent-Tiering
Monthly saving: $6,800
Phase 2: Architecture Changes (Weeks 3-6)
- Migrate stateless workloads from EC2 to Lambda + Fargate (pay per use)
- Implement CloudFront for all static assets and API caching
- Configure VPC endpoints to eliminate cross-region transfer costs
- Downgrade dev/staging databases to Single-AZ
Monthly saving: $7,400
Phase 3: Reserved Capacity (Week 7+)
- Purchase 1-year Reserved Instances for baseline stable workloads
- Savings Plans for Lambda and Fargate usage
Monthly saving: $2,600
The Results
| Metric | Before | After | Change |
|---|---|---|---|
| Monthly AWS cost | $48,000 | $31,200 | -35% |
| P99 API latency | 280ms | 190ms | -32% |
| Deployment frequency | Weekly | Daily | 7x |
The architecture changes didn’t just reduce costs — they improved performance and enabled faster deployments.
The Key Lesson
Cloud cost optimization is not a one-time project. The client now has:
- Monthly cost review cadence
- Budget alerts at 80% and 100% of targets
- Resource tagging policy so every cost is attributable
- Compute Optimizer recommendations reviewed quarterly
Treat your cloud bill like a product. Own it, understand it, and optimize it continuously.
If your AWS bill has been growing faster than your user base, get in touch. A one-week audit typically uncovers 20–40% savings.