Amazon EFS Pricing and Cost Optimization

Amazon Elastic File System solves a problem that is genuinely hard, shared file storage that many servers, containers, and functions can mount at once, that grows and shrinks on its own, and that nobody has to patch or provision. It is also, per gigabyte, one of the most expensive places on AWS to keep a file. EFS Standard costs about $0.30 per gigabyte-month, roughly thirteen times S3 Standard and nearly four times an EBS gp3 volume. That premium is the price of convenience, and for the right data it is worth every cent.

The trouble starts when the wrong data pays it. Most EFS file systems hold a small core of actively used files and a long tail of data nobody has opened in months, all billed at the same Standard rate because nothing ever told the service otherwise. EFS cost optimization is largely the work of separating those two populations, and because AWS built tiering into the service, most of it is a lifecycle policy rather than a migration. This guide covers what EFS actually bills for, what each storage class costs in 2026, the charges that hide outside the storage line, and the levers that typically cut the bill by half or more.

Key Takeaways

• EFS Standard at about $0.30 per GB-month is roughly 13 times S3 Standard and 4 times EBS gp3; the premium buys a shared, elastic, fully managed file system.
• Infrequent Access ($0.016) and Archive ($0.008) cut storage cost by 95 to 97 percent, and lifecycle management moves cold files there automatically.
• EFS bills on four dimensions: storage class, throughput mode, data access on cold tiers, and cross-AZ or cross-region transfer. Optimizing storage alone misses half the bill.
• One Zone file systems cost about 47 percent less than Regional, at the cost of single-AZ durability; they suit dev, test, and re-creatable data.
• File metadata always stays in Standard regardless of where the data tiers, so a fully tiered file system still shows Standard storage on the bill.

How EFS Pricing Works

EFS is billed on the average storage used through the month, measured in GB-month, plus three variable dimensions. The complete rate card, including regional pricing, is on the official Amazon EFS pricing page, and the dimensions are: storage, by class and by whether the file system is Regional (multi-AZ) or One Zone; throughput, either Elastic (billed per GB read and written) or Provisioned (billed per MB/s-month above the included baseline); data access, a per-GB fee when reading or writing data that lives in Infrequent Access or Archive; and data transfer, for access across Availability Zones or Regions. There is no minimum fee, and new AWS accounts receive 5 GB of Standard Regional storage free each month for the first year.

The four-dimension model is what catches teams out. A file system that looks cheap on storage can double its bill through provisioned throughput it never uses, or through cross-AZ traffic from compute mounted in the wrong zone. Read the bill by dimension, not just by total.

EFS Storage Classes and What They Cost

Rates below are for US East (N. Virginia) in 2026 and vary slightly by region.

Two things stand out. Infrequent Access is 95 percent cheaper than Standard and Archive 97 percent cheaper, so the tiering decision dominates everything else. And One Zone cuts every rate by roughly 47 percent in exchange for storing data in a single Availability Zone, which makes it the right choice for development, testing, and data you can regenerate, and the wrong choice for anything you cannot afford to lose access to during a zone outage.

Throughput: The Charge That Doubles Bills

Throughput is where EFS bills diverge most from expectations, because two modes bill in opposite ways.

Elastic Throughput is the default and the right choice for almost everyone. Performance scales automatically with the workload, and you pay only for data moved: about $0.03 per GB read and $0.06 per GB written. For variable or unpredictable I/O it is both the simplest and usually the cheapest option, because idle file systems cost nothing in throughput.

Provisioned Throughput lets you buy a guaranteed level of MB/s independent of how much data you store, at about $6 per MB/s-month above the baseline your storage already includes. It exists for workloads that need consistently high throughput on small file systems. It is also the most common source of EFS overspend, because throughput provisioned for a peak and never revisited bills every hour at full rate. A file system provisioned at 100 MB/s that uses 10 costs about $600 a month for headroom it never touches. Unless you have measured a sustained need, stay on Elastic.

The Hidden Costs Most Teams Miss

• Cross-AZ access. EC2 instances or containers mounting EFS from a different Availability Zone than the mount target pay about $0.01 per GB in transfer. Multi-AZ compute reading a busy file system can generate this continuously. Create a mount target in every AZ your compute runs in.
• Data access fees on cold tiers. Reading from IA or Archive adds $0.01 per GB on top of throughput. Data that is tiered too aggressively and then read often can cost more than it did in Standard, which is why the lifecycle window matters.
• Metadata stays in Standard. File names, ownership, and directory structure are always billed at the Standard rate regardless of where the file data lives. A file system with millions of small files can show a surprisingly large Standard line even when nearly all data has tiered.
• Forgotten file systems. Because EFS has no minimum fee, an abandoned file system costs only its storage, which is exactly why it goes unnoticed. Audit for file systems with no recent mounts.
• Backups and replication. AWS Backup for EFS and cross-region replication each bill separately for the stored copy and the transfer, and are easy to leave running long after the need has passed.

Six Levers That Cut EFS Costs

1. Turn on lifecycle management

EFS lifecycle management automatically moves files that have not been accessed for a set period, selectable from one day up to a year, into Infrequent Access, and optionally on into Archive after a longer window. It also moves files back to Standard on first access if you enable that transition. For most file systems this single setting is the entire optimization, because the long tail of untouched files is where the money is. Start with a 30-day window to IA and 90 days to Archive, then tighten once you have seen the access fees.

2. Stay on Elastic Throughput unless you have measured a need

Move any file system on Provisioned Throughput back to Elastic unless CloudWatch shows sustained throughput above the baseline. If a workload genuinely needs provisioned capacity, size it to the measured p95 rather than the peak, and revisit it quarterly.

3. Use One Zone for data that can tolerate a zone outage

Development, test, CI, scratch, and re-creatable datasets rarely justify multi-AZ durability. Creating them as One Zone file systems saves roughly 47 percent across storage, throughput, and access charges. Keep production and irreplaceable data Regional.

4. Fix cross-AZ mounts

Create a mount target in each Availability Zone where compute runs and ensure clients mount through their local target. This removes the per-GB cross-AZ charge entirely and usually improves latency as well.

5. Put the right data in EFS in the first place

EFS exists for data that needs to be shared across many clients through a file interface. Data accessed by one instance belongs on EBS at a quarter of the price; objects, backups, and archives belong in S3 at a fraction of that. Auditing what lives on EFS and moving what does not need a shared file system is often the largest single saving, larger even than tiering. Our guide to AWS storage pricing across EC2, S3, EBS, and RDS covers the comparison, and our S3 cost optimization guide covers where the archived data should go.

6. Delete what nobody mounts

List file systems, check the last mount and I/O activity in CloudWatch, and remove or snapshot the ones nobody has touched. Combine this with a review of AWS Backup plans and replication configurations that outlived their purpose.

A Worked Example

Consider a 2 TB Regional file system on Standard storage with Elastic Throughput, serving a fleet of application servers. At $0.30 per GB-month it costs about $600 in storage plus throughput and any cross-AZ traffic.

Analysis shows that about 300 GB is actively read and written while roughly 1.7 TB has not been opened in over 30 days, most of it in over 90. Enabling lifecycle management with a 30-day IA transition and a 90-day Archive transition leaves 300 GB in Standard (about $90), moves around 400 GB to Infrequent Access (about $6.40), and around 1.3 TB to Archive (about $10.40). Storage drops from roughly $600 to roughly $107 a month, an 82 percent reduction, before throughput. Add a small allowance for IA and Archive access fees on the occasional reads, and the file system still costs less than a fifth of what it did, with no change to the applications mounting it.

That is consistent with what most teams see in practice: lifecycle tiering alone typically cuts EFS bills by 60 percent or more, and adding One Zone where appropriate pushes total savings toward 75 percent.

When EFS Optimization Backfires

• Tiering hot data. Files that are read regularly incur a $0.01 per GB access fee plus throughput every time, so an aggressive lifecycle window can cost more than Standard. Use access metrics to set the window, not a guess.
• Millions of tiny files. Metadata stays in Standard, so a file system of small files gains little from tiering and may be better served by aggregating files or moving them to S3.
• One Zone for anything critical. A single-AZ outage makes the file system unavailable until the zone recovers. The 47 percent saving is not worth it for production data you cannot regenerate.
• Latency-sensitive first reads. Reading from Archive has higher first-byte latency than Standard, so keep anything latency-sensitive out of the coldest tier.

The Bottom Line

EFS is expensive per gigabyte because it does something genuinely valuable, and it becomes cheap the moment you stop paying the premium for data that does not need it. Turn on lifecycle management, stay on Elastic Throughput, use One Zone where a zone outage is tolerable, fix cross-AZ mounts, and make sure the data on EFS actually needs a shared file system. Then, as with any storage line, keep watching it, because new files land in Standard every day. Model your own numbers in the AWS Pricing Calculator, and for how EFS fits into a wider cost practice, see our guides to AWS cost optimization and Kubernetes cost optimization, where EFS often appears as persistent storage for EKS.