100 TB of compliance data sitting in S3 Standard costs about $2,300 a month. The same data in Glacier Deep Archive costs about $99. Same bytes, same durability, a 23x difference, because almost nobody actually reads compliance archives.
100 TB of compliance data sitting in S3 Standard costs about $2,300 a month, based on AWS's published rate of $0.023 per GB. Move that same data to Glacier Deep Archive, priced at $0.00099 per GB, and the bill drops to about $99 a month. Same bytes, same durability guarantees, a 23x difference. The only thing that changed is which tier it lives in. That gap is not a rounding error, it's roughly $26,500 a year sitting on the table for one dataset that almost nobody actually reads.
Most organizations still store the overwhelming majority of their data in the most expensive tier available, not because it needs to be there, but because nobody built a policy to move it. Data tiering is the fix, and it is one of the highest-leverage, lowest-risk cost optimizations available in cloud storage today.
Every major storage provider offers multiple storage classes priced according to how quickly and how often you expect to access the data. The categories are consistent across providers even though the exact names differ:
Hot storage is built for data accessed frequently and needed instantly: active application data, current-quarter analytics, anything a user might request right now. It carries the highest per-gigabyte price and the lowest latency.
Warm or cool storage suits data accessed occasionally, perhaps monthly, where a small retrieval fee and modest latency are acceptable trade-offs for meaningfully lower storage cost.
Cold storage is for data accessed rarely, quarterly or less, where higher retrieval fees and delays of minutes to hours are fine because the data mostly just needs to exist and be durable.
Archive storage is the cheapest tier by far, designed for data accessed once or twice a year at most, such as compliance records or long-term backups, with retrieval delays that can run from several hours to nearly two days depending on the provider and retrieval speed selected.
The pricing spread across tiers is dramatic and consistent across every major provider, based on published 2026 rates:
| Provider | Hot Tier | Warm / Cold Tier | Archive Tier |
|---|---|---|---|
| AWS S3 | Standard: $0.023/GB | Glacier Instant Retrieval: $0.004/GB | Glacier Deep Archive: $0.00099/GB |
| Azure Blob | Hot: $0.018/GB | Cool: $0.010/GB, Cold: $0.0045/GB | Archive: $0.00099/GB |
| Google Cloud Storage | Standard: $0.020/GB | Nearline / Coldline | Archive: ~$0.0012/GB |
AWS S3 Standard costs $0.023 per GB per month, while Glacier Deep Archive costs $0.00099, a spread of more than 23x. Azure's Hot tier runs $0.018 per GB against $0.00099 for Archive, similarly wide. Google Cloud Storage's Standard tier starts at $0.020 per GB with Archive priced around $0.0012. Across every provider, the pattern holds: archive tiers cost roughly 1 to 2 percent of what hot storage costs per gigabyte.
The cheapest tier on paper is not automatically the cheapest tier in practice. Cold and archive tiers charge separately for getting data back out, and those fees can erase the storage savings fast if you retrieve more often than the tier is designed for. This is closely related to the hidden costs your cloud provider isn't telling you about.
Pulling data out of Glacier-class storage runs roughly $0.01 to $0.05 per GB depending on retrieval speed, plus separate per-request charges. A bulk restore of a large archive can turn a nearly-free monthly storage bill into a five-figure one-time charge if it isn't planned for.
AWS bills Standard-Infrequent Access data for a minimum of 30 days even if you delete it after 20. Glacier Deep Archive enforces a 180-day minimum, billing the full period regardless of when data is actually removed. Moving data to a deep archive tier only makes sense if it's genuinely going to sit there for months.
Azure enforces a 128 KiB minimum billable size on Cool, Cold, and Archive tiers, and Google Cloud applies a similar 128 KB minimum on Nearline, Coldline, and Archive. A tiering policy that moves millions of small files into a cold tier without accounting for this can end up costing more than expected.
Standard archive retrieval on AWS Glacier Deep Archive takes up to 12 hours; Azure and Google Cloud archive tiers carry comparable delays. If a compliance audit or legal hold needs data back within the hour, the cheapest tier is the wrong tier, regardless of the per-gigabyte price.
Applying AWS's published per-gigabyte rates to a single 100 TB compliance archive shows exactly what's at stake:
| Storage Class | Monthly Cost | Annual Cost |
|---|---|---|
| S3 Standard | $2,300 | $27,600 |
| S3 Glacier Instant Retrieval | $400 | $4,800 |
| S3 Glacier Deep Archive | $99 | $1,188 |
Moving from Standard to Glacier Deep Archive saves roughly $26,400 a year on this dataset alone. Even the middle option, Glacier Instant Retrieval, which still allows millisecond access, cuts the bill by more than 80 percent. For data genuinely accessed once or twice a year, Deep Archive's 12-hour retrieval window is rarely a practical problem, and the savings are close to total.
Before moving anything, identify what's actually being accessed and how often. Cloud provider analytics tools, S3 Storage Lens on AWS, Azure Storage analytics, and similar tools on other platforms, show real access frequency per object or bucket, which is a far better starting point than guessing based on data age alone.
AWS Intelligent-Tiering, and equivalent automated tiering products on other platforms, move data between tiers automatically based on observed access, without manual lifecycle rules. Industry data suggests automated tiering can cut storage costs 30 to 50 percent with no change to application behavior, at the cost of a small per-object monitoring fee that typically pays for itself many times over.
For data with a known access pattern, application logs that matter for 90 days and rarely after, compliance records that need to exist for seven years but are read almost never, explicit lifecycle rules that transition data on a schedule are more predictable and often cheaper than automated tiering's monitoring overhead.
Before moving a dataset to Glacier Deep Archive or its equivalent, estimate how often it will realistically need to be retrieved and at what volume. If the honest answer is more than once or twice a year, or if any single retrieval could involve pulling back a large fraction of the archive at once, a warmer tier with faster, cheaper retrieval is very likely the better economic choice despite the higher monthly storage rate.
Abandoned multipart uploads and orphaned snapshot data accumulate storage charges indefinitely if nothing cleans them up. A lifecycle rule that aborts incomplete uploads after seven days and removes orphaned data on a schedule is a small, one-time setup step that prevents a slow, invisible cost leak.
The cheapest byte is the one sitting in the right tier, not the coldest tier available. Match the storage class to how the data actually gets used, and the savings take care of themselves.
Compare AWS, Google Cloud, Azure, and alternatives like Backblaze B2 Discover how much you could save in seconds
