S3 Glacier Pricing & Cost Guide (storage, retrieval, Deep Archive)

Archive storage is cheap for data you rarely read. But total cost depends on how often you restore and how your data is packaged (large objects vs many small objects).

What to model (the full archive cost picture)

  • Stored GB-month: average stored GB over the month
  • Retrieval GB/month: how much data you restore and read back
  • Retrieval requests/month: how many objects you retrieve
  • Minimum duration / early deletion: if you delete/overwrite before the minimum duration
  • Transitions: how often objects move into archive (and whether you churn objects between tiers)

How to estimate without copying price tables

You do not need a full price matrix to build a good budget. Build a model with stable inputs and validate with billing after one cycle: average stored GB-month, restored GB/month, and objects restored/month (requests).

A fast estimate (baseline + peak)

Use AWS S3 Glacier / Deep Archive Cost Calculator for a storage + retrieval model. If your workload frequently deletes or replaces objects, add minimum-duration effects to your estimate.

  • Baseline: typical month restores and object counts.
  • Peak: audits, backfills, reprocessing, incident-driven restores.

Common pitfalls

  • Using "total archive size" as the retrieval input (retrieval is about restored volume, not stored volume).
  • Ignoring object counts when archives contain many small files.
  • Short-lived data in tiers with minimum duration.
  • Frequent rehydration workflows that turn archive into a warm tier by accident.
  • Assuming all restores look like the average (backfills are usually the outlier).

How to validate the estimate

  • Validate retrieval GB/month and retrieval request counts from restore jobs or access patterns.
  • Watch for small-object amplification: many objects means many retrieval requests.
  • Check minimum duration rules if data is short-lived or overwritten frequently.
  • After one billing cycle, reconcile usage types in CUR/Cost Explorer against your baseline and peak scenarios.

How to choose an archival class (latency vs retrieval frequency)

"Archive" is not one thing. The right class depends on how often you restore and how fast you need the data. If restores are part of a weekly workflow, you may want a warmer tier or a cached analysis copy instead of treating restores as a routine operation.

  • Latency requirement: seconds/minutes/hours to first byte changes your operational workflow (and what a "failed restore" means).
  • Restore volume: large backfills are usually the peak month; model that peak explicitly.
  • Object count: packaging many small files into fewer objects can reduce request-like fees and make restores operationally simpler.
  • Data lifetime: if you delete or overwrite frequently, minimum duration can erase savings.

Glacier vs Deep Archive (quick comparison)

  • Deep Archive: lowest storage price, longest retrieval times, highest penalty for frequent restores.
  • Glacier: higher storage price, faster retrieval, better fit for occasional restores.
  • Cost trade-off: frequent restores usually favor Glacier or a warmer tier even if storage is higher.

Related tools

Archive cost Estimate retrieval Retrieval time Retrieval pricing Transfer cost Reduce archive costs Storage costs

Sources

Related guides

Estimate Glacier/Deep Archive retrieval volume (GB and requests)
How to estimate archival retrieval costs: model GB restored per month and the number of objects retrieved (requests), plus common drivers like restores, rehydration, and analytics.
Glacier/Deep Archive cost optimization (reduce restores and requests)
A practical playbook to reduce archival storage costs: reduce restores, reduce small-object request volume, and avoid minimum duration penalties. Includes validation steps and related tools.
S3 Glacier retrieval pricing per GB and per request
A practical breakdown of Glacier retrieval pricing: cost per GB retrieved plus request fees, with guidance for small-object amplification and tier selection.
S3 pricing: a practical model for storage, requests, egress, and replication
A practical S3 pricing guide: what to include (GB-month, requests, egress, replication) and how to estimate the key inputs without copying price tables.
Estimate Secrets Manager API calls per month (GetSecretValue volume)
A practical workflow to estimate Secrets Manager API request volume (especially GetSecretValue): measure and scale when possible, model from runtime churn when not, and validate with CloudTrail so your budget survives peaks.
S3 Glacier retrieval time: how long restores take by tier
A practical guide to S3 Glacier retrieval time: how restore tiers map to latency, what drives delays, and how to plan workflows without surprises.

Related calculators

RPS to Monthly Requests Calculator
Estimate monthly request volume from RPS, hours/day, and utilization.
API Request Cost Calculator
Estimate request-based charges from monthly requests and $ per million.
CDN Request Cost Calculator
Estimate CDN request fees from monthly requests and $ per 10k/1M pricing.

FAQ

What are the main cost components for archival storage?
Storage (GB-month) plus retrieval (GB retrieved and retrieval requests). Some classes also have minimum storage duration and early deletion fees.
Is Amazon Glacier the same as S3 Glacier?
Amazon Glacier is now delivered as S3 Glacier storage classes. Pricing is still based on storage, retrieval, and minimum duration rules, but the service is accessed through S3.
Why do small objects increase cost?
Because retrieval is billed on both GB and requests. Many small-object restores can create high request counts even if total GB is modest.
What else should I include besides storage and retrieval?
For a realistic model, include transition/restore events (if applicable), minimum duration effects, and the operational workflow (how often you rehydrate and how long you keep restored copies).
Last updated: 2026-02-23