Optimizing Storage Costs with S3 Intelligent-Tiering - Automatic Tiering Based on Access Patterns
Learn how storage tiers are transitioned automatically based on access patterns, how to choose between Intelligent-Tiering and lifecycle policies, and how to use the archive tiers.
How Automatic Tiering Works
S3 Intelligent-Tiering automatically monitors the access patterns of objects and moves them to the most cost-efficient access tier. Three tiers are enabled automatically: from the default Frequent Access tier, objects not accessed for 30 days move to the Infrequent Access tier, and objects not accessed for 90 days then move to the Archive Instant Access tier. All three of these tiers offer low latency and high throughput, and objects in them can be retrieved in milliseconds. In addition, you can enable through opt-in the Archive Access tier (moves after a minimum of 90 days without access) and the Deep Archive Access tier (moves after a minimum of 180 days without access). These tiers can store data at storage rates equivalent to Glacier Flexible Retrieval and Glacier Deep Archive, respectively. Objects in the Infrequent Access tier and the Archive Instant Access tier automatically return to the Frequent Access tier when accessed with GetObject or similar requests, but the two opt-in tiers assume asynchronous access and require explicit restoration with a RestoreObject request.
Choosing Between Lifecycle Policies and Intelligent-Tiering
Lifecycle policies transition storage classes with fixed rules based on elapsed days, so they suit data whose access patterns can be clearly predicted. Intelligent-Tiering, on the other hand, tiers data dynamically based on actual access patterns, which makes it optimal for data whose access frequency fluctuates or is hard to predict. Intelligent-Tiering incurs a monthly per-object monitoring and automation charge, but objects smaller than 128 KB are not subject to this charge. However, in exchange for being exempt, they are also not eligible for automatic tiering and are always kept in the Frequent Access tier at that tier's rates, so in buckets where small objects make up the majority, you can hardly gain any savings. The cost benefit is greatest for workloads with large objects and irregular access patterns.
Archive Tiers and Monitoring
When you enable the optional Archive Access tier and Deep Archive Access tier of S3 Intelligent-Tiering, objects that have not been accessed for a long time automatically move to Glacier-equivalent storage. The number of days without access before the move is a minimum of 90 days for the Archive Access tier and a minimum of 180 days for the Deep Archive Access tier, and both can be extended up to a maximum of 730 days. Enabling the Archive Access tier at 90 days bypasses the automatic Archive Instant Access tier and moves objects directly to the Archive Access tier, so it is a choice that gives up millisecond access to lower storage charges. Objects in these two tiers cannot be retrieved with GET alone and require an asynchronous procedure that starts restoration with a RestoreObject request. Standard retrieval takes 3 to 5 hours for the Archive Access tier and within 12 hours for the Deep Archive Access tier, and when restoration is started from S3 Batch Operations, restoration processing begins within minutes and within 9 hours, respectively. Restore requests can be made at up to 1,000 requests per second for a single account in a single Region. Monitor the number of objects and the amount of storage in each tier with CloudWatch metrics to evaluate the effect of tiering quantitatively. Use S3 Storage Lens to visualize the adoption of Intelligent-Tiering across the entire account.
Intelligent-Tiering Pricing
(As of August 2026, US East (N. Virginia) / Asia Pacific (Tokyo)) The Frequent Access tier of Intelligent-Tiering has the same price as S3 Standard: 0.023 USD per 1 GB per month (N. Virginia, first 50 TB) / 0.025 USD (Tokyo). The Infrequent Access tier is 0.0125 USD / 0.0138 USD, the automatic Archive Instant Access tier is 0.004 USD / 0.005 USD, the opt-in Archive Access tier is 0.0036 USD / 0.0045 USD, and the Deep Archive Access tier is 0.00099 USD / 0.002 USD (all per 1 GB per month). The unit price falls with each step down in tier. On top of this, a monitoring and automation charge of 0.0025 USD per 1,000 objects per month (the same in both Regions) applies. Objects smaller than 128 KB are exempt from this charge, but they are not tiered automatically either and are always kept in the Frequent Access tier. In buckets with a large number of objects and a small size per object, savings from tiering are hard to achieve, so you check the distribution of object sizes in advance.
Data That Suits It and Data That Does Not
Intelligent-Tiering delivers the greatest benefit for data whose access frequency is hard to predict. It automatically sorts data whose usage patterns cannot be read, and for which designing lifecycle rules manually is difficult, into the optimal tier. On the other hand, for data whose access patterns are clearly known, specifying the tier with fixed lifecycle rules is simpler and leaves no waste. Also, when there is a large number of tiny objects smaller than 128 KB, they incur no monitoring charge but are also not eligible for automatic tiering and always incur Frequent Access tier charges, so you use it after assessing these characteristics. What matters is to judge, according to the nature of the data, whether automation or manual design pays off.
Using the Archive Access Tiers
Intelligent-Tiering has a feature that automatically moves data not accessed for a long time to lower-cost tiers intended for archiving. When you enable it, frequently used data stays in tiers that can be accessed immediately, while data that has not been touched for a long time is moved to much cheaper tiers. However, data that has moved to an archive tier cannot be retrieved with GET alone and requires waiting several hours with a restore via RestoreObject in between, so it does not suit data that needs immediate access. Whether to enable the archive tiers is decided based on whether the waiting time can be tolerated, including the effort of building the application side around restoration. The balance between immediacy of access and storage cost is designed to match how the data is used.
Adoption and Migration
Intelligent-Tiering can be applied per bucket or per prefix, and you specify it as the storage class for newly stored objects. Existing data can also be moved to Intelligent-Tiering with lifecycle rules. After adoption, keep in mind that monitoring each object 128 KB or larger incurs a small charge. Objects smaller than 128 KB are exempt from this charge but are not tiered either, so for data consisting mostly of small files the effect is hard to obtain, and you identify which data to target. Starting with large volumes of data whose access patterns are unknown and widening the scope while confirming the effect is a manageable way to proceed with adoption.
Combining with Other Optimization Methods
Intelligent-Tiering brings even greater benefits when combined with other cost optimization methods. While the movement between access tiers is automated, lifecycle rules that delete unneeded data after a set period are configured separately, adding the decision to stop storing data at all. Tools that visualize storage usage show which data is in which tier and how much it costs, which helps find room for optimization. By not leaving everything to automatic tiering and operating it together with the overall data retention policy, you can keep storage costs low on an ongoing basis. A design that combines multiple methods is the key.
Summary
S3 Intelligent-Tiering is a service that optimizes storage costs through automatic tiering based on access patterns, with no additional charges for moving objects between tiers or retrieving them. Three tiers, Frequent, Infrequent, and Archive Instant Access, work automatically, and the 5-tier structure that adds the opt-in Archive Access and Deep Archive Access achieves cost savings of up to 95%. The three automatic tiers return objects to the Frequent Access tier automatically when accessed, so they require no operational effort, but the two opt-in tiers require explicit restoration with RestoreObject. It is optimal for data whose access patterns are hard to predict, and choosing between it and lifecycle rules is important.
References (Official AWS Resources)
The primary sources for this page are the official AWS website and documentation. Check the official pages below for the latest specifications and pricing.
If this page and the official documentation disagree, treat the official documentation as authoritative.