awss3storagecost-optimization

S3 Glacier Instant Retrieval cost: cheap archive with fast access

S3 Glacier Instant Retrieval stores data at about $0.004 per GB, roughly a sixth of S3 Standard, while still giving millisecond access. Here is what it costs, where the retrieval fees bite, and when it beats Standard-IA and the slower Glacier tiers.

The C3X Team··6 min read

Quick answer

S3 Glacier Instant Retrieval costs about $0.004 per GB-month in us-east-1, roughly a sixth of S3 Standard ($0.023) and a third of Standard-IA ($0.0125), while still returning objects in milliseconds like Standard. The catch is a $0.03 per GB retrieval fee, a 128 KB minimum billable object size, and a 90-day minimum storage duration. It is the right class for archives you rarely read but must read instantly when you do, such as compliance records, medical images, or old assets served on demand. If you access data more than roughly once a quarter, Standard-IA is usually cheaper once retrieval fees are counted.

S3 Glacier Instant Retrieval sits in an unusual spot: it is priced like cold archive storage but behaves like hot storage. You pay about $0.004 per GB-month in us-east-1, close to the Glacier Flexible Retrieval rate, yet objects come back in milliseconds with a normal GET, not a multi-hour restore job. That combination makes it the go-to class for data you almost never touch but must have instantly when you do.

How the pricing works

Storage classStorage /GB-moRetrieval /GBAccess
S3 Standard$0.023noneInstant
Standard-IA$0.0125$0.01Instant
Glacier Instant Retrieval$0.004$0.03Instant (ms)
Glacier Flexible Retrieval$0.0036$0.01Minutes to hours
Glacier Deep Archive$0.00099$0.02Hours

Storage is the cheap part. The fees that decide whether the class saves money are the retrieval charge of $0.03 per GB and the per-request GET cost, which for Glacier Instant Retrieval is about $0.01 per 1,000 GETs, much higher than the $0.0004 per 1,000 on Standard. There is also a 90-day minimum storage duration: delete or move an object before 90 days and you still pay for the remaining days. And every object is billed as at least 128 KB, so millions of tiny files store far more expensively than their real size suggests.

When it beats Standard-IA

The break-even is all about access frequency. Storing 100 TB in Glacier Instant Retrieval costs about $400 per month versus $1,250 in Standard-IA, a $850 saving. But if you retrieve that whole 100 TB in a month, the $0.03 per GB retrieval fee adds roughly $3,000, wiping out the saving many times over. As a rule of thumb, if you read less than about 5 to 10 percent of the data per month, Glacier Instant Retrieval wins. Above that, the cheaper $0.01 retrieval on Standard-IA makes it the better home.

Where it fits in a lifecycle

Most teams reach Glacier Instant Retrieval through a lifecycle policy rather than by writing there directly. A common pattern keeps fresh data in Standard, transitions to Standard-IA after 30 days, then to Glacier Instant Retrieval after 90 days for objects that must stay instantly readable but are seldom touched. Data with no instant-access requirement should skip straight to Glacier Flexible Retrieval or Deep Archive, which store for $0.0036 and $0.00099 per GB. See Glacier versus Deep Archive and the storage class comparison for the full ladder, and lifecycle policies for how to automate the transitions.

The small-object trap

The 128 KB minimum billable size is the most common reason Glacier Instant Retrieval disappoints. Archiving 50 million objects that average 20 KB each stores as if they were 128 KB, so you pay for 6.4 TB instead of 1 TB, a 6x markup that can make the cheap class more expensive than Standard for that data. Intelligent-Tiering handles this more gracefully because objects under 128 KB are never charged monitoring or moved to archive tiers. For huge counts of tiny files, aggregate them or lean on Intelligent-Tiering instead.

Deciding with numbers, not guesses

Glacier Instant Retrieval is a genuinely good deal for rarely-read, instantly-needed archives, but only once you model the retrieval and request fees against your real access pattern. The storage saving is easy to see; the retrieval cost is the part that surprises people on the invoice. Estimate both before you set the lifecycle rule, and price your storage footprint against the resource catalog so the archive tier is a deliberate choice rather than a default that turns out to cost more than the tier it replaced.

FAQ

How much does S3 Glacier Instant Retrieval cost?

About $0.004 per GB-month in us-east-1 for storage, roughly a sixth of S3 Standard ($0.023) and a third of Standard-IA ($0.0125). On top of storage you pay a $0.03 per GB retrieval fee, higher per-request GET costs of about $0.01 per 1,000 requests, a 128 KB minimum billable object size, and a 90-day minimum storage duration.

When is Glacier Instant Retrieval cheaper than Standard-IA?

When you read less than roughly 5 to 10 percent of the data per month. Storage is much cheaper ($0.004 versus $0.0125 per GB), but the $0.03 per GB retrieval fee is triple the $0.01 on Standard-IA. If you frequently read the data, the higher retrieval cost overwhelms the storage saving and Standard-IA becomes the cheaper home.

Does Glacier Instant Retrieval have slow restore times?

No. Despite the Glacier name, Instant Retrieval returns objects in milliseconds with a normal GET, just like S3 Standard. It is Glacier Flexible Retrieval and Deep Archive that require an asynchronous restore taking minutes to hours. Instant Retrieval trades that instant access for a higher $0.03 per GB retrieval fee than the slower tiers.

What is the 128 KB minimum on Glacier Instant Retrieval?

Every object is billed as if it were at least 128 KB. Archiving millions of small files, say 20 KB each, means you pay for 128 KB per object, a large markup. For huge counts of tiny objects, aggregate them into larger files or use Intelligent-Tiering, which does not charge or archive objects under 128 KB.

How does C3X help with Glacier Instant Retrieval cost?

C3X prices S3 buckets and lifecycle configuration from Terraform before deploy, so the storage class, transition timing, and expected access pattern are visible in the pull request. That lets teams model retrieval and request fees against real usage and catch cases where a cold tier would cost more than the tier it replaced, rather than discovering it on the invoice.

What to do next

Model your archive tiers before you set the lifecycle rule. C3X prices S3 storage classes from Terraform against a live resource catalog. Start with the quickstart.

Try C3X on your own Terraform

Free and open source. No API key required. One command to install, one command to estimate.