Cloud backup costs at AWS, Azure and Google Cloud are mostly the cost of stored copies: disk snapshots, backups in a backup service and data replicated for disaster recovery are billed per gigabyte and month, plus, depending on the provider, a fee per protected instance or replicating server, copies to other regions and restores. Backup cost therefore grows with retention and change rate, not with load, and keeps running after the protected resource is deleted.
How it works
Three layers produce charges: disk snapshots, managed backup services with policies and vaults, and replication that keeps an environment ready for a disaster.
Snapshots
On AWS the first EBS snapshot of a volume is full and every later one incremental: it stores only blocks changed or written since the previous snapshot, and the charge follows the size of that data, not of the volume. Archiving converts a snapshot into a full one in the EBS Snapshots Archive tier, which costs less per GB-month, requires 90 days of retention and charges retrieval per GB.
Azure managed disk snapshots are full or incremental. Incremental ones sit on Standard HDD whatever the source disk, on ZRS where available and LRS otherwise, and are billed on used size only.
Google Cloud's standard and archive snapshots are incremental, stored redundantly in one or more regions and billed on their total size; archive snapshots cost less and take longer to restore. Unlike instant snapshots, they survive the deletion of their disk.
Backup services
AWS Backup charges storage per GB-month in warm and cold storage, restores per GB and copies to other Regions per GB transferred. Cold storage exists only for some resource types, EBS and EFS among them, with a 90-day minimum. For copies of fully managed resources the source account pays the transfer and the destination account always pays the storage.
Azure Backup splits two meters: a monthly fee per protected instance by used size (up to 50 GB, up to 500 GB, then per started 500 GB) and backup storage per GB by tier and redundancy. The Standard tier comes with LRS, ZRS or GRS; the Archive tier for long-term backups has a 180-day minimum and a retrieval fee. Restores from the Standard tier carry no transaction or egress charges.
Google Cloud's Backup and DR Service charges backup vault storage per GiB, a management fee per GiB of protected source capacity (for VMs the VM's size, not the storage used) and transfer between regions. According to the pricing page, SKUs prefixed BackupDR: Storage, BackupDR: Management and BackupDR: Networking have been billed since 2024-12-02; they were free while the backup vault was in Preview.
Replication for disaster recovery
AWS Elastic Disaster Recovery charges per replicating source server and hour, plus resources in the customer's own account: an equally sized staging disk per source disk, replication servers on EC2 and EBS snapshots for recovery points; full-size servers run only during drills and recovery. Azure Site Recovery charges per protected instance and month, each instance free for its first 31 days; replica storage, snapshots, transactions and egress are charged during that period too, compute only on failover. On Google Cloud the disaster copy comes from backup vaults in another region or a multi-region, with transfer or upload charges per GiB.
| AWS | Azure | Google Cloud | |
|---|---|---|---|
| Snapshot | EBS snapshot, incremental, per GB-month | Incremental managed disk snapshot on used size | Standard and archive snapshot by stored size |
| Backup service | AWS Backup: GB-month, restore per GB | Azure Backup: instance per month plus storage per GB | Backup and DR Service: storage and management per GiB |
| Disaster recovery | Elastic Disaster Recovery: server per hour plus staging | Site Recovery: instance per month from day 32 plus replicas | Vault in another region, transfer per GiB |
What it means for your cost
Backup storage grows with retention times change rate. Copies to a second region add data transfer, priced as Data transfer and egress describes.
Billing data keeps the layers apart: snapshots sit with the disks, on AWS under the Amazon EC2 service (usage type EBS:SnapshotUsage, shown as EC2-Other in Cost Explorer), on Azure under managed disks and on Google Cloud under Compute Engine; backup services appear as their own service with separate line items for storage, instances or management; disaster recovery shows as a fee plus the compute and storage lines of the staging environment. For incremental snapshots Azure's daily usage report shows used size divided by the days of the month as the quantity.
Because copies outlive their source, orphaned copies are a savings opportunity (see Idle and orphaned resources). The decisions are how long each class of data is kept, what moves to an archive tier with a minimum retention, which copies need a second region, and whether disaster recovery relies on restores or on continuous replication, which costs every month. Storage types and classes are priced in Block, object and file storage and Storage tiers and lifecycle, database backups in Managed databases. Azure also sells backup storage as reserved capacity, a commitment of the kind covered in Reserved capacity for databases, storage and serverless.
Where it goes wrong
- Snapshots without disks. Symptom: snapshot storage stays although the volumes are gone. Cause: snapshots are resources of their own and bill until they are deleted; Google states explicitly that standard and archive snapshots outlive their disk.
- Protection stopped, data kept. Symptom: Azure Backup bills an instance for a VM that no longer exists. Cause: while data stays in the vault, instance fee and storage keep running; Google bills such data under its
BackupDR: Management - In GCP - Long-Term RetentionSKU. - Early deletion from an archive tier. Symptom: a charge for storage already deleted. Cause: AWS Backup cold storage and EBS Snapshots Archive require 90 days and Azure Backup's Archive tier 180 days, and the remainder is charged pro rata.
- A cross-region copy in the wrong budget. Symptom: transfer in the production account, storage in the backup account, and no budget sees both. Cause: for fully managed resources AWS Backup assigns transfer to the source account and storage to the destination account.
- Staging that keeps running. Symptom: EC2 and EBS lines in the recovery region with no application. Cause: Elastic Disaster Recovery keeps replication servers and staging disks running for as long as servers replicate.
Example
List prices in USD, read on 2026-10-16. AWS in Europe (Frankfurt): EBS snapshots USD 0.054 per GB-month, EBS Snapshots Archive USD 0.0135; Elastic Disaster Recovery USD 0.028 per source server and hour. Azure in Germany West Central: Azure Backup USD 5 a month for an instance up to 50 GB, USD 10 up to 500 GB and USD 10 for each started 500 GB above that, Standard tier backup storage USD 0.03 per GB with LRS and USD 0.06 with GRS; Site Recovery to Azure USD 25 per instance and month after 31 free days. Google Cloud in Frankfurt (europe-west3): backup vault storage USD 0.052 per GiB and month.
Sources
Read on 2026-10-16. How Amazon EBS snapshots work, Amazon EBS pricing, Understand Amazon EBS billing information, AWS Backup pricing, AWS Elastic Disaster Recovery pricing, Create an incremental snapshot for managed disks, Azure Backup pricing, Azure Site Recovery pricing, About archive and standard disk snapshots, Backup and DR Service pricing.