AWS Cost Optimization: Stop Burning Money

I've seen startups burning $50k/month on AWS when they could be paying$ 15k. The waste is usually hiding in plain sight - you just need to know where to look.

Where The Money Goes

1. Right-Size Your EC2 Instances

Most instances are oversized. Check actual utilization:

# Get CPU utilization for last 2 weeks
aws cloudwatch get-metric-statistics \
  --namespace AWS/EC2 \
  --metric-name CPUUtilization \
  --dimensions Name=InstanceId,Value=i-1234567890 \
  --start-time $(date -v-14d +%Y-%m-%dT%H:%M:%S) \
  --end-time $(date +%Y-%m-%dT%H:%M:%S) \
  --period 3600 \
  --statistics Average

If average CPU is under 20%, you can probably downsize.

2. Use Spot Instances for Non-Critical Workloads

Spot instances are 60-90% cheaper. Use them for:

Dev/staging environments
Batch processing
CI/CD runners
Stateless workers

// AWS CDK example
const spotFleet = new ec2.CfnSpotFleet(this, 'SpotFleet', {
  spotFleetRequestConfigData: {
    iamFleetRole: fleetRole.roleArn,
    targetCapacity: 10,
    launchSpecifications: [{
      instanceType: 't3.medium',
      imageId: ami.imageId,
      spotPrice: '0.02',  // Max you're willing to pay
    }],
  },
});

3. Reserved Instances & Savings Plans

If you know youll use EC2 for a year, commit to it:

Savings Plans are more flexible than Reserved Instances - they apply across instance families.

4. NAT Gateway Costs Are Sneaky

NAT Gateways charge $0.045 per GB. If your Lambda functions are hitting external APIs, that adds up fast.

Solutions:

Does Lambda really need to be in VPC?
Use VPC Endpoints for AWS services (S3, DynamoDB, etc.)
Consolidate outbound traffic

// VPC Endpoint for S3 - no NAT needed
const s3Endpoint = vpc.addGatewayEndpoint('S3Endpoint', {
  service: ec2.GatewayVpcEndpointAwsService.S3,
});

5. S3 Lifecycle Policies

Data just sits there forever, costing money:

{
  "Rules": [
    {
      "ID": "Move to IA after 30 days",
      "Status": "Enabled",
      "Transitions": [
        {
          "Days": 30,
          "StorageClass": "STANDARD_IA"
        },
        {
          "Days": 90,
          "StorageClass": "GLACIER"
        }
      ],
      "Expiration": {
        "Days": 365
      }
    }
  ]
}

6. Clean Up Orphaned Resources

Resources that get forgotten:

# Find unattached EBS volumes
aws ec2 describe-volumes \
  --filters Name=status,Values=available \
  --query 'Volumes[*].[VolumeId,Size,CreateTime]'

# Find unused Elastic IPs
aws ec2 describe-addresses \
  --query 'Addresses[?AssociationId==`null`].[PublicIp,AllocationId]'

# Old snapshots
aws ec2 describe-snapshots --owner-ids self \
  --query 'Snapshots[?StartTime<`2024-01-01`].[SnapshotId,StartTime,VolumeSize]'

7. RDS Optimization

Dev/staging: Single-AZ, smaller instance, stop outside work hours
Production: Consider Aurora Serverless for variable workloads

# Stop RDS instance (saves money when not in use)
aws rds stop-db-instance --db-instance-identifier dev-database

# Automatically stops after 7 days, need to restart

Quick Wins Checklist

[ ] Enable Cost Explorer and set budget alerts
[ ] Right-size EC2 instances (check CloudWatch)
[ ] Use Spot for dev/staging/batch
[ ] Buy Savings Plans for predictable workloads
[ ] Add S3 lifecycle policies
[ ] Use VPC Endpoints instead of NAT for AWS services
[ ] Delete orphaned EBS volumes and old snapshots
[ ] Stop dev/staging databases outside work hours
[ ] Enable S3 Intelligent-Tiering for uncertain access patterns

Tools That Help

AWS Cost Explorer - Built-in, see where money goes
AWS Compute Optimizer - Right-sizing recommendations
Infracost - See cost impact in PR reviews
Spot.io - Automated spot instance management