Archive storage tiers are priced far below standard storage, and the difference is recovered elsewhere. Understanding where determines whether the tier saves money or costs it.
The low rate applies only to sitting still
Archive pricing covers holding data. Reading it back carries a separate charge per unit retrieved, and that charge can exceed months of storage cost for the same data.
There is often a request charge as well, which matters for archives made of many small objects rather than a few large ones.
The tier therefore suits data that is stored for compliance or contingency and genuinely not read, which is a narrower category than it first appears.
Retrieval takes time by design
The cheapest tiers do not make data available immediately. A retrieval request is queued and completed within a window that may be hours.
Faster retrieval options exist at higher cost, which reintroduces the trade the tier was chosen to avoid.
This makes archive tiers unsuitable for disaster recovery where a recovery time objective is measured in minutes, even though they are often proposed for exactly that.
Minimum retention periods apply
Archive tiers typically bill a minimum duration regardless of how long data is actually kept, commonly measured in months.
Deleting or moving data before that period elapses incurs a charge for the remainder, which surprises teams that use lifecycle rules to move data aggressively.
Data that turns over quickly therefore costs more in an archive tier than it would have cost in standard storage.
Transitions are billed as operations
Moving objects between tiers is charged per object, so a lifecycle policy applied to a large number of small files can produce a substantial one-off cost.
The saving on storage may take a long time to repay that transition cost, particularly when the objects are small.
Aggregating small files into larger archives before transition addresses both the transition cost and the per-request retrieval charge.
Testing the restore is the part that gets skipped
An archive that has never been read back is an assumption rather than a backup, and the assumption fails at the worst moment.
Periodic partial restores confirm that the data is intact, that credentials still work and that the retrieval process is understood by whoever will run it.
Budgeting for those test retrievals is part of the true cost of the tier, and omitting them is how organisations discover the charges during an incident instead.