After migrating 50+ applications from EC2 to serverless, here's what we learned: The choice between EC2 and serverless isn't just about cost—it's about workload patterns, utilization, and hidden expenses.
The Bottom Line: When Each Option Wins
EC2 Wins When:
- Consistent high utilization (>30%)
- Long-running processes (>15 min)
- Predictable, steady traffic
- Need custom OS/kernel access
$74/month — High traffic example (t3.large)
Serverless Wins When:
- Low/sporadic utilization (<15%)
- Short-lived requests (<15 min)
- Variable, unpredictable traffic
- Event-driven workloads
$42/month — 10M requests/month example
Real-World Cost Breakdown
Here's the actual cost data from our AWS migration case study for a healthcare platform:
| Scenario | EC2 Monthly Cost | Serverless Monthly Cost | Savings |
|---|---|---|---|
| Low Traffic API (100K requests/month) | $7.49 (t3.micro) | $2.10 | 72% savings |
| Medium Traffic API (1M requests/month) | $14.98 (t3.small) | $8.50 | 43% savings |
| High Traffic API (10M requests/month) | $29.95 (t3.medium) | $42.00 | EC2 cheaper |
| Very High Traffic (100M requests/month) | $74.88 (t3.large) | $420.00 | EC2 much cheaper |
| Cron Jobs (Runs 1hr/day) | $7.49 (always running) | $0.35 | 95% savings |
The 15% Rule: If your EC2 utilization is below 15%, serverless will almost always be cheaper. We've seen this pattern across dozens of migrations. Use our Lambda cost calculator to verify.
Hidden Costs Analysis
EC2 Hidden Costs:
- Data Transfer: $0.09/GB outbound (same for both)
- EBS Storage: $0.10/GB-month for gp3 volumes
- Load Balancer: $16.20/month + $0.008/LCU-hour
- Auto Scaling: No cost, but complexity overhead
- Operational Costs: Server management, patching, monitoring
Serverless Hidden Costs:
- API Gateway: $3.50 per million requests (if needed)
- CloudWatch Logs: $0.50/GB ingested
- NAT Gateway: $45/month + $0.045/GB (if VPC needed)
- Provisioned Concurrency: $0.000041667/GB-second
Decision Framework
Use this framework to make the right choice for your workload. For complete architectural guidance, check our serverless vs traditional architecture guide.
Choose EC2 When:
- Traffic is predictable and consistent
- Utilization exceeds 30% of instance capacity
- Processes run longer than 15 minutes
- You need Reserved Instances for 1-3 year commitment savings (up to 72%)
- Application requires persistent connections or stateful processing
Choose Serverless When:
- Traffic is sporadic or unpredictable
- Workloads are event-driven
- Each request completes in under 15 minutes
- You want zero infrastructure management
- Development speed matters more than marginal cost differences
Real Migration Case Study
We migrated a healthcare platform from EC2 to serverless. Here's what happened:
- Before: 3x t3.medium instances ($90/month) running 24/7
- After: Lambda + API Gateway ($38/month) for actual usage
- Savings: 58% monthly cost reduction
- Bonus: Auto-scaling, zero maintenance, better reliability
The platform handled 5M requests/month with 95% of those requests happening during business hours. EC2 instances sat idle 70% of the time—perfect serverless candidate.
Read the full story in our 85% AWS cost reduction case study.
Optimization Tips
For EC2:
- Use Reserved Instances for predictable workloads (save 30-72%)
- Right-size instances based on CloudWatch metrics
- Use Spot Instances for fault-tolerant workloads (save 90%)
- Leverage Auto Scaling to match demand
For Serverless:
- Optimize Lambda memory for best price/performance ratio
- Use ARM (Graviton2) for 20% better price-performance
- Batch requests when possible to reduce invocations
- Monitor and eliminate cold starts for latency-sensitive apps