
When a cloud bill is higher than expected, the cause is usually not the virtual machines themselves but the items nobody watches: data egress to the internet, forgotten snapshots and backups, disks not attached to any machine, idle public IP addresses, oversized machines running around the clock, and log ingestion charges. Group your cost analysis by service and tag, set up budget alerts, then close these items one by one.
Symptoms: the bill grows but nobody created anything new
- The monthly bill rises a little every month although the number of servers has not changed.
- “Bandwidth”, “Data Transfer Out” or similar network items on the bill are unexpectedly large.
- Disks and snapshots of a virtual machine you thought was deleted still appear on the bill.
- An Azure virtual machine shown as “Stopped” in the portal keeps generating compute charges.
- Monitoring services such as Log Analytics or CloudWatch make up an unexpected share of the bill.
- Your AWS bill contains a public IPv4 address item you have not seen before.
Root cause: every resource you do not delete is billed
A cloud bill is calculated on the resources that exist and run, not on what you think you are using. The problem is that some resources silently stay behind after an operation. According to Microsoft’s documentation, when you delete a virtual machine in Azure, the attached disks are not deleted by default to prevent data loss, and you continue to pay for these unattached disks.
The second common trap is how a machine is stopped. According to Azure’s documentation on VM states, a machine shut down from inside the guest operating system goes into the “Stopped” (Stopped Allocated) state and is billed because the hardware is still allocated. Compute charges stop only when the machine is in the “Stopped (Deallocated)” state, and even then disks and networking resources continue to incur charges.
The third item is data transfer. On Azure’s bandwidth pricing page, inbound data transfer is free while outbound data to the internet is charged in tiers, and transfer between regions is charged separately. On AWS, data out to the internet is also priced in tiers. Large file downloads, copying backups out of the cloud or cross-region replication grow this item quickly.
Finally, some charges change over time. AWS announced that from 1 February 2024 it charges for all public IPv4 addresses, whether attached to a service or not. According to Azure Monitor documentation, the most significant charges in Log Analytics are typically data ingestion and retention. Because prices change, we do not quote amounts here; check the provider’s pricing page for current figures.
Step-by-step fix
- Break the bill down by service: in Azure Cost Management use Cost analysis, or in AWS Cost Explorer group costs by service, resource group and tag. Start with the three largest items.
- List unattached disks: in Azure, managed disks whose ManagedBy field is empty are not attached to any machine; in AWS, EBS volumes in the available state are the same.
- Review old snapshots and backups: sort them by creation date and identify which retention rule each one belongs to.
- Find idle public IP addresses: Azure public IPs not attached to any network interface and AWS Elastic IP addresses without an association.
- Check the power state of virtual machines: in Azure, only machines in the “VM deallocated” state stop incurring compute charges.
- Document deletion decisions, tag the resources and act only after the owner has approved.
# Azure PowerShell: unattached managed disks
Get-AzDisk | Where-Object { $_.ManagedBy -eq $null } | Select-Object Name, ResourceGroupName, DiskSizeGB
# Azure: snapshots (by creation date)
Get-AzSnapshot | Sort-Object TimeCreated | Select-Object Name, ResourceGroupName, TimeCreated, DiskSizeGB
# Azure: public IPs not attached to any configuration
Get-AzPublicIpAddress | Where-Object { $_.IpConfiguration -eq $null } | Select-Object Name, ResourceGroupName, IpAddress
# Azure: power state of virtual machines
Get-AzVM -Status | Select-Object Name, ResourceGroupName, PowerState
# AWS CLI: unattached EBS volumes, your own snapshots, unassociated Elastic IPs
aws ec2 describe-volumes --filters Name=status,Values=available
aws ec2 describe-snapshots --owner-ids self --query "Snapshots[].[SnapshotId,StartTime,VolumeSize]" --output table
aws ec2 describe-addresses --query "Addresses[?AssociationId==null]"These commands only list resources. Deleting a disk or a snapshot cannot be undone; Microsoft’s own script is also meant to be run in list-only mode first. Before deleting, confirm that the content exists in another backup and that no restore plan depends on that resource.
How do you reduce egress costs?
First find where the traffic goes: network flow logs or a per-service cost breakdown usually point to a specific storage account or backup job. Using a CDN for frequently downloaded static content, keeping data in the region where it is needed, transferring backups out of the cloud incrementally instead of as a daily full copy, and switching off unnecessary cross-region replication are the most effective steps. A rise in egress is often noticed only at the end of the month; because budgets can be filtered by service, a separate budget for the networking item gives an early warning.
Oversized machines that are always on
Tools such as Azure Advisor and AWS Compute Optimizer use utilisation data to show under-used machines. Set up a schedule that deallocates test and development machines automatically outside working hours. For production workloads that must run continuously, consider reservations or savings plans (Azure Reservations, Savings Plans, AWS Savings Plans); they give a discount in return for a commitment, but get the sizing right first.
Permanent prevention
- Set up budgets and alerts. According to Microsoft, Azure budgets send notifications at actual or forecasted cost thresholds but do not stop resources; for automatic action you need to attach an action group. On AWS, AWS Budgets does the same job.
- Make owner, project and environment tags mandatory on every resource; an untagged resource is a cost without an owner.
- Define lifecycle rules for snapshots and backups (Azure Backup retention policies, storage lifecycle management, Amazon Data Lifecycle Manager).
- When deleting a virtual machine, choose to delete its disks and public IP as well, or use a post-deletion checklist.
- Review the scope of log collection: lower unnecessary verbosity and set an appropriate retention period for each table.
- Review the bill monthly, broken down by service and tag.
When should you call an expert?
If you cannot find which resource the cost comes from, if the source of the egress item is unclear, or if decisions such as downsizing production machines or buying reservations carry downtime and commitment risks, expert support saves time. For billing problems like these, which we encounter in the field at businesses in Denizli, Workas carries out cost analysis, cleans up idle resources and sets up tagging and budget alerts in a planned way as part of its cloud and server hosting service; we work Monday to Saturday, 08:00–18:00, and respond to emergencies with priority.
Frequently Asked Questions
Does shutting down an Azure VM from the operating system stop the charges?
No. A machine shut down from inside the OS stays in the “Stopped” state and is billed. To stop compute charges it must be put into the “Stopped (Deallocated)” state from the portal or the CLI.
Are the disks deleted when I delete the virtual machine?
Not by default in Azure. Attached disks are kept to prevent data loss and continue to be billed; they must be deleted separately.
Does a budget alert stop spending automatically?
Azure budgets only send notifications and do not affect resources. If you want an automatic action, you need to attach an action group to the budget.
Is inbound data transfer charged too?
According to Azure’s bandwidth page, inbound data transfer is free; charges apply to outbound data to the internet and to transfer between regions. Check the provider’s pricing page for current amounts.


