This cleanup playbook lists fifty concrete things to delete today, the resources that quietly bill every month while delivering nothing. The reason a delete list works is that a large share of cloud waste is not oversized infrastructure that needs careful rightsizing but simply abandoned infrastructure that needs removing: a disk detached six months ago, a snapshot from a migration nobody finished, an IP reserved and never used. Each item here is safe to remove after one quick check, and together they recover real money in an afternoon. Treat this as the fast pass you run before the deeper optimization work begins.
This article is part of our complete guide to cloud rightsizing and waste elimination, the cluster pillar it links up to. It is the action-list companion to how to audit a cloud environment for waste, and what you delete here should feed into a continuous waste detection process so it never piles up again.
Check it is genuinely unused, snapshot it if it holds data, and prefer stop-then-delete with a short grace period over immediate deletion. Fast does not mean reckless. Every item below is safe with one check first.
Compute: 12 things to delete
Start with compute, because stopped and forgotten machines are the most common abandoned resource. Delete: stopped instances no longer needed, instances tagged for a finished project, the oldest unused machine images and their backing storage, unused launch templates and configurations, instances in a region you no longer operate in, duplicate instances from a failed deployment, test instances older than your longest sprint, instances with zero network activity for thirty days, scale groups stuck at a minimum nobody set, scheduled instances for events long past, instances running a service since migrated elsewhere, and personal experiment boxes whose owner has left. Confirm each is truly idle using the method in how to find idle cloud resources across providers before removing it.
Storage: 14 things to delete
Storage is where abandoned spend accumulates fastest, because deleting a server rarely deletes its disks. Delete: volumes detached from any instance, snapshots older than your retention policy, snapshots of resources that no longer exist, redundant backups beyond what the policy requires, incomplete multipart uploads, old machine image copies, empty or near-empty buckets and shares billed for provisioned capacity, duplicate datasets copied for a one-off analysis, test data refreshed from production and never cleared, logs past their useful life, exports nobody downloaded, temporary files that outlived their job, versioned object history beyond what you need, and orphaned managed-disk artifacts. These overlap directly with storage waste: snapshots, orphaned disks, and old backups; snapshot anything uncertain before deleting.
| Category | Items to delete | Check first |
|---|---|---|
| Compute | 12 (stopped instances, old images, unused templates) | Confirm idle 30+ days |
| Storage | 14 (detached volumes, stale snapshots, old backups) | Snapshot if it holds data |
| Network | 12 (idle IPs, load balancers, gateways) | Confirm no live attachment |
| Data and other | 12 (dormant DBs, unused endpoints, old keys) | Verify no active reads |
Want the whole estate swept, not just the obvious wins?
Our cloud cost audit runs this delete pass across AWS, Azure, GCP and OCI, finds the items you cannot see from the console, and leaves behind the detection process that stops them returning. On the performance model, you pay only from realized savings. No savings, no fee.
Book a cloud cost audit →Network: 12 things to delete
Network resources bill silently because they often cost money only when idle. Delete: reserved IP addresses not attached to anything, which on most clouds are charged precisely because they are unused, load balancers with no healthy targets, NAT gateways serving no traffic, unused VPN connections and gateways, orphaned network interfaces, idle private endpoints, security groups and route tables tied to deleted resources, unused peering connections, DNS records pointing at nothing, certificates for retired domains, and bandwidth allocations nobody draws on. The idle-IP and idle-load-balancer cases are detailed in how to eliminate idle load balancers and IPs. Confirm there is no live attachment before removing each.
Data and the rest: 12 things to delete
Finally, the managed services and miscellaneous resources that escape a console sweep. Delete: dormant database instances with no connections, read replicas nobody queries, idle data-warehouse clusters left running, unused message queues and topics, stale container images in registries, orphaned Kubernetes persistent volumes, unused API endpoints and gateways, expired or rotated keys still provisioned, monitoring on resources that no longer exist, scheduled jobs for retired pipelines, sandbox accounts nobody uses, and trial subscriptions converted to paid and forgotten. The dormant database and warehouse cases connect to how to rightsize databases without hurting performance. Verify no active reads before deleting anything that holds data.
The Cloud Waste Audit Framework turns this delete list into a repeatable sweep, with the safety checks for each resource type and the queries that surface every item across a multi-cloud estate.
Delete today, then stop it returning
A delete pass is the fastest cloud saving available, but it is a one-time event unless you make it stick. Whatever you remove today will reaccumulate as new resources are created and abandoned, so feed this list into a recurring scan and the ownership work in how to tackle untagged and unowned resources so the cleanup becomes a habit rather than a heroic afternoon. That is the move from Cut to Lock in our See, Cut, Lock, Run method. Resource types, idle-charge rules, and what is billed when unused differ across AWS, Azure, GCP and OCI and change, so verify the current billing behavior for each item in the relevant provider's documentation before relying on it, as of May 2026.
The short version
A lot of cloud waste does not need analysis, it needs deleting: roughly fifty specific items across compute, storage, network, and data that bill every month while doing nothing. Run the delete pass today with one safety check per item, then feed the list into a recurring scan so it never piles up again. When you want the whole estate swept and kept clean, that is part of what our rightsizing and waste elimination service delivers.