FinOps Foundation estimates 32% of cloud spend is wasted on idle resources. Stopped instances, orphaned volumes, unused IPs, forgotten load balancers. This guide shows how to find and eliminate waste systematically.
💡 Quick start: TCOIQ gives instant AI-powered results in 60 seconds. Built by Wekams. Free at tcoiq.com.
Stopped instances still attached to volumes (paying storage, not compute), unattached EBS volumes/Managed Disks, unused Elastic IPs ($0.005/hr = $43/year per unused IP), old snapshots not associated with active volumes, empty load balancers with no targets, idle NAT Gateways, over-provisioned RDS running at 5% utilisation, dev environments running 24/7 but used 8hrs/day.
AWS: use Cost Explorer Resource Optimization or Compute Optimizer. Azure: use Azure Advisor Cost recommendations. GCP: use GCP Recommender. Or query directly: AWS CLI to list stopped instances >30 days, unattached EBS volumes, unassociated Elastic IPs. Tag every resource with owner — makes cleanup easy.
Dev/test running 24/7 when used 8hrs/day wastes 67% of compute. AWS Instance Scheduler (free CloudFormation template): shut down at 6pm, restart at 8am. Azure DevTest Labs: built-in auto-shutdown. GCP instance schedules: built-in. Saving: 67% of dev/test compute — $6,700/month on a $10,000/month dev environment.
Snapshots accumulate with no cleanup. Best practices: delete snapshots older than 90 days for non-production volumes, create snapshot lifecycle policies (AWS DLM, Azure Backup, GCP Snapshot Schedules) to expire old snapshots automatically. Maintain 7 daily, 4 weekly, 3 monthly — delete the rest. Snapshot cleanup saves $1,000-$5,000/month for large environments.
Monthly process: automated idle resource report (TCOIQ or native tools), assign cleanup tasks to resource owners (via tags), set 30-day deadline for cleanup or escalation, track waste eliminated vs total cloud spend as KPI. Over 12 months, reduces cloud spend growth by 15-25%.
AI-powered results in 60 seconds. No consultant needed. Free plan available.
Find Cloud Waste →