7 Zombie Cloud Resources Quietly Draining Your Bill (and How to Kill Them)

Published by Rushikesh Dhule, LessBill · October 27, 2026 · Updated November 16, 2026


Every cloud account has a graveyard. A test environment someone spun up in March. A volume left behind when an instance was terminated. A load balancer for a feature that was cancelled two sprints ago. Nobody is using any of it, but all of it is still on the bill.

These are zombie resources: things that are dead to your team but very much alive to your cloud provider's billing system. Each one looks small on its own, which is exactly why they survive. A few dollars here, thirty dollars there. Add them up across a year, a few teams and a few regions, and you are paying for a small server farm that does nothing.

Here are the seven most common zombies we see, what they cost, and how to find them before the next invoice does.

1. Unattached disk volumes

When you terminate a virtual machine, its extra disks often stay behind. On AWS, any EBS volume that was not set to "delete on termination" keeps billing in the "available" state. On Azure, managed disks stay put after a VM is deleted unless you remove them too. GCP persistent disks behave the same way.

A gp3 volume on AWS costs about $0.08 per GB per month in us-east-1. That sounds tiny until you find a forgotten 500 GB volume: $40 a month, $480 a year, holding data nobody remembers.

How to find them: In AWS, filter EBS volumes by state = available. In Azure, filter disks where "Disk state" is Unattached. In GCP, look for disks with no "In use by" value. Snapshot anything you are unsure about, then delete the volume.

2. Old snapshots that never expire

Snapshots are cheap insurance, so teams take lots of them and then never clean them up. Automated backup scripts are the worst offenders: they create a snapshot every night and delete nothing.

The fix is a retention policy, not a manual cleanup. Use AWS Data Lifecycle Manager or AWS Backup, Azure Backup policies, or GCP snapshot schedules with a retention period. Decide how far back you genuinely need to restore (often 7 to 30 days for most workloads) and let the policy do the deleting.

3. Idle public IP addresses

This one catches a lot of teams on AWS. Since February 2024, AWS charges $0.005 per hour for every public IPv4 address, whether it is attached to something or not. That is roughly $3.60 a month per address.

Small, but it multiplies. An account with 30 public IPs spread across instances, load balancers, NAT gateways and leftover Elastic IPs is paying over $1,200 a year just for addresses.

What to do: Release Elastic IPs that are not associated with anything. For internal services, drop public IPs entirely and reach them through a load balancer, VPN or Systems Manager Session Manager. Where you can, use IPv6, which has no address charge on AWS.

4. Load balancers with no traffic

Load balancers are easy to create and easy to forget. Each one has an hourly base charge whether it serves a million requests or zero. An Application Load Balancer in us-east-1 runs about $0.0225 per hour, which is around $16 a month before traffic charges, plus the public IP charges above.

How to find them: Look for load balancers with no registered healthy targets, or with near-zero request counts over the last 30 days in CloudWatch, Azure Monitor or Cloud Monitoring. If nothing has hit it in a month, it is a zombie.

5. NAT gateways in forgotten VPCs

NAT gateways are one of the most expensive "idle" resources in AWS networking. In us-east-1 they cost $0.045 per hour (about $32.85 a month) just for existing, plus $0.045 for every GB they process.

Dev and staging VPCs are the usual suspects. A three-AZ setup copied from production means three NAT gateways, close to $100 a month, for an environment someone uses twice a week.

What to do: Delete NAT gateways in VPCs with no running workloads. In non-production environments, one NAT gateway is usually enough for all AZs. And add free Gateway VPC Endpoints for S3 and DynamoDB so that traffic stops flowing through NAT at all.

6. Non-production environments that run 24/7

Dev, QA and staging environments are usually used during working hours, maybe 50 hours a week. But they often run all 168 hours. That means roughly 70% of what you pay for them buys nothing.

The same applies to Kubernetes clusters. An EKS control plane alone costs $0.10 per hour, about $73 a month, before a single node is added. Teams that create a cluster per feature branch or per developer can rack up several of these without noticing.

What to do: Schedule non-production instances and databases to stop outside working hours using AWS Instance Scheduler, Azure Automation start/stop, or GCP instance schedules. Scale dev node groups to zero overnight. Tear down short-lived clusters automatically when the branch is merged.

7. Orphaned storage buckets and logs

Object storage is cheap per GB, so nobody questions it. But buckets full of old build artifacts, debug logs and one-off data exports grow quietly for years. Log retention is a related trap: CloudWatch log groups, for example, are set to "never expire" by default.

What to do: Add lifecycle rules that move old objects to cheaper storage classes (S3 Glacier, Azure Cool/Archive, GCP Coldline/Archive) and delete them after a set period. Set a retention period on every log group. Turn on incomplete multipart upload cleanup, which is easy to forget and surprisingly common in large buckets.

How to stop zombies from coming back

Cleaning up once feels great. Three months later, the graveyard is back. The real fix is making zombies hard to create and easy to spot:

  • Tag everything at creation. At minimum: owner, environment and project. An untagged resource with no owner is the first candidate for deletion.
  • Enforce tags with policy. Use AWS tag policies or Service Control Policies, Azure Policy, or GCP organization policies so untagged resources can't be created in the first place.
  • Set expiry dates on experiments. A simple "expires-on" tag plus a weekly script that lists anything past its date catches most test leftovers.
  • Review idle resources every week, not every quarter. A zombie found in week one costs a few dollars. Found at the quarterly review, it has already cost twelve times that.
  • Set budget alerts. AWS Budgets, Azure Cost Management and GCP budgets can all warn you when spend drifts above normal, often before you would ever see it on the invoice.

A 30-minute zombie sweep

Three commands find most of the list above on AWS: unattached disks, idle Elastic IPs and NAT gateways.

aws ec2 describe-volumes --filters Name=status,Values=available --query "Volumes[].[VolumeId,Size,CreateTime]" --output table

aws ec2 describe-addresses --query "Addresses[?AssociationId==null].[PublicIp,AllocationId]" --output table

aws ec2 describe-nat-gateways --filter Name=state,Values=available --query "NatGateways[].[NatGatewayId,VpcId]" --output table

What one of each costs if it is doing nothing:

ZombieHow it is pricedPer month
500 GB unattached gp3 disk500 x $0.08$40.00
Idle public IPv4 address$0.005 x 730 hours$3.65
Load balancer with no traffic$0.0225 x 730 hours$16.43
NAT gateway in an unused VPC$0.045 x 730 hours$32.85

One of each is about $93 a month, or $1,115 a year, and real accounts rarely have just one. Prices are from AWS's pages for EBS, public IPv4, load balancers and NAT gateways; the load balancer and NAT gateway also hold public IPs that are billed on top.

Before deleting: snapshot a disk if you are unsure (a snapshot is cheaper than the volume), and check a NAT gateway's traffic metrics for a week, because some are used only by a monthly job.

Related guides: AWS Public IPv4 Charges: What Each Address Costs and How to Cut It and S3, EBS and Snapshot Costs: Where Storage Money Leaks.

Let LessBill do the hunting

Doing all of this by hand across three clouds, multiple accounts and every region is tedious, and it is the kind of job that quietly stops happening after the first month.

LessBill connects to your AWS, Azure and GCP accounts with read-only access and scans 60+ resource types for exactly these problems: unattached disks, idle IPs, empty load balancers, forgotten NAT gateways, always-on dev environments and more. Every finding comes with an estimated monthly saving and a clear next step, so you can clean up in minutes instead of spending an afternoon in three different consoles.

LessBill is free during beta. Run your first scan and see what is hiding in your bill.

Prices quoted are AWS us-east-1 list prices as of October 2026 and vary by region. Always confirm with your provider's pricing pages before budgeting.

Sources and further reading

Comments

Popular posts from this blog

Stop Paying On-Demand Prices: Savings Plans, Reserved Instances and Spot Explained

When Does Azure SQL Serverless Beat Provisioned? The Maths