Forgotten cloud sandboxes are a predictable byproduct of moving fast. The cloud makes it trivial to stand up an isolated account or project to try something, which is exactly the agility teams want, but the same ease means the teardown is the step that gets skipped. A proof of concept proves its point and the engineer moves on; a demo environment outlives the demo; a training account is left for the next cohort that never comes. The resources inside, instances, databases, clusters, storage, keep billing because nothing connects their continued existence to a continued purpose. And because sandboxes are deliberately separate from production, they often sit outside the tagging, budgets and monitoring that would otherwise flag them, so the cost compounds in a blind spot until someone goes looking.
This article is part of our complete guide to cloud rightsizing and waste elimination, the cluster pillar it links up to. Forgotten sandboxes are a close cousin of zombie infrastructure: finding what everyone forgot, with the same root cause, no link between a resource and a live reason to keep it, scaled up to whole accounts.
Sandboxes are easy to create and easy to skip tearing down. They keep billing outside production governance, so the cost grows in a blind spot until an audit finds it. The fix is expiry by default.
Why forgotten sandboxes cost more than they look
The cost of a forgotten sandbox is rarely just the obvious idle instance. A proof of concept often provisions the full shape of a real system to test it properly: managed databases that bill whether or not they hold live data, a Kubernetes cluster with a control plane charge running with no workloads, snapshots and storage left behind, reserved IPs and load balancers wired up for the demo. Each of these keeps billing on the same hourly or capacity model it would in production, so an abandoned sandbox can carry the cost profile of a small real environment for months. Worse, because sandboxes sit outside production governance, the resources inside are frequently over-provisioned, left at default sizes, and never cleaned up, the exact pattern described in cleaning up dev and test environments. The total across a large organization is usually a surprise, because no single sandbox looks expensive enough to notice.
| Sandbox type | What keeps billing | Why it survives |
|---|---|---|
| Proof of concept | Full system shape: databases, clusters, storage | Point proven, teardown skipped |
| Demo environment | Running stack kept "just in case" | Outlives the demo it was built for |
| Training account | Per-seat resources left provisioned | Held for a next cohort that does not come |
| Personal test project | Whatever the engineer spun up | Owner moved on, no expiry set |
How much is sitting in sandboxes nobody owns anymore?
Our cloud cost audit finds every sandbox and proof-of-concept account across your organization, quantifies what each is costing, and proves the reclaim against a clean baseline on AWS, Azure, GCP and OCI. On the performance model, you pay only from realized savings. No savings, no fee.
Book a cloud cost audit →How to find the forgotten sandboxes
Finding forgotten sandboxes means looking at the account and project level, not just the resource level, because the unit of waste here is the whole environment. Inventory every account, subscription and project in the organization and classify each by purpose, flagging the ones that are non-production experiments. For each, the questions are simple: when was it created, who owns it now, when did it last see real activity, and is there a current reason for it to exist. Accounts with no recent meaningful activity and no identifiable live owner are the forgotten ones, and the cost inside them is your reclaim. Where ownership has gone stale, the process in tackling untagged and unowned resources applies at account scale, tracing each sandbox back to a responsible team before anything is shut down. This account-level inventory is part of any thorough audit of a cloud environment for waste.
Shut them down, and make expiry the default
Once a sandbox is confirmed forgotten and traced to an owner, shutting it down is usually straightforward, since by definition nothing live depends on it, though it is worth capturing anything of lasting value, a dataset or a configuration, before the environment is torn down. The more important move is preventing the next generation, because forgotten sandboxes regrow as fast as you clean them up unless the creation pattern changes. The durable fix is expiry by default: every sandbox is created with a time-to-live and an owner, automatically flagged for renewal or teardown when it expires, so the absence of a deliberate decision results in shutdown rather than indefinite billing. That inversion, where a sandbox must be justified to survive rather than justified to remove, is the same principle that keeps any category of waste from rebuilding, as set out in stopping cloud waste from coming back. Verify current pricing for the managed services typically left in sandboxes against each provider's documentation as of May 2026 when sizing the reclaim, since those rates change.
The Cloud Waste Audit Framework includes the account-level inventory and the sandbox expiry policy we use to find forgotten environments and stop new ones outliving their purpose.
The short version
Forgotten cloud sandboxes are accounts and projects created for a proof of concept, demo, training or test, then left running after the reason expired. They cost more than they look because a sandbox often carries the full shape of a real system, billing for databases, clusters and storage outside production governance, so the total grows in a blind spot. Find them by inventorying every account and project, flagging non-production experiments with no recent activity and no live owner; shut down the confirmed ones after capturing anything of value; and make expiry the default so every sandbox carries a time-to-live and must be justified to survive. When you want every forgotten sandbox found, quantified and reclaimed with the saving proven down, that is part of what our rightsizing and waste elimination service delivers.