GCP Cost Optimization: Top 10 Leaks (2026)

Google Cloud bills often hide waste in a few predictable places. Here are the top 10 GCP cost optimization levers we see in mid-market audits. For a full framework see our FinOps audit guide and cloud cost optimization checklist.

1. Over-provisioned VMs

Right-size Compute Engine instances using utilization metrics. Downsize or switch to E2 or preemptible where workload allows.

2. Persistent disks

Unattached or oversized disks. Delete unattached disks; move cold data to Standard or balance performance/cost with pd-balanced/pd-ssd.

3. Committed use discounts (CUDs)

Underused or misaligned commitments. Align CUDs to steady-state usage and term length (1-year vs 3-year).

4. Sustained use discounts

Automatic but only apply to on-demand. Ensure you're not double-paying with unnecessary committed use where sustained use already helps.

5. Egress and network

Cross-region and internet egress. Use Cloud CDN, keep traffic in-region where possible, and see reduce egress costs (CDN/WAF).

6. BigQuery

Slot usage, on-demand query patterns, and storage. Use BI Engine or slot reservations for predictable workloads; archive old partitions to cold storage.

7. Pub/Sub and message volume

Message volume and retention. Tune retention and subscription counts to actual needs.

8. Cloud Storage

Bucket class and lifecycle. Move cold data to Nearline/Coldline/Archive and set lifecycle rules.

9. Idle or forgotten resources

Old snapshots, unused IPs, abandoned VMs in dev projects. Regular cleanup and project-level tagging.

10. Kubernetes (GKE)

Node pool sizing, autoscaling, and cluster footprint. Right-size node pools and use node auto-repair to avoid over-provisioning.