AWS DataSync pricing, 2026

The online data-transfer service that moves files and objects into, out of, and across AWS. The Basic and Enhanced mode per-GB rates, the per-task-execution fee, the request and storage charges DataSync does not include, and a worked 50 TB scenario showing where DataSync beats Snowball Edge and where it does not.

AWS DataSync is the managed online data-transfer service. It moves data between on-premise storage and AWS, between AWS storage services, and between AWS and other public clouds, with automatic encryption, integrity verification, and incremental transfer. Its pricing model is refreshingly simple: a flat per-GB fee on data transferred, with no separate compute charge for the managed transfer infrastructure. This page is the 2026 pricing reference for DataSync specifically, including the two transfer modes and the downstream charges that a DataSync-only cost estimate misses.

Pricing verification

The Basic mode rate ($0.0125 per GB), the Enhanced mode rate ($0.015 per GB), and the Enhanced mode task-execution fee ($0.55 per execution) were checked directly against the AWS DataSync pricing page in June 2026. AWS notes that the per-GB rate is the same across Regions for the base transfer fee. Source and destination service charges (S3 requests, storage, cross-Region transfer) are separate and are not part of the DataSync fee. Verified June 2026.

The DataSync pricing model

DataSync charges a flat fee per gigabyte of data actually transferred between the source and destination locations defined in a task. There are two modes. Basic mode is the original transfer mode and the cheaper per-GB rate; Enhanced mode is the higher-throughput mode for very large Amazon S3 transfers and adds a small per-task-execution fee.

AWS DataSync transfer pricing (AWS DataSync pricing page, June 2026)

Mode	Per-GB transferred	Per-task-execution fee	Best for
Basic	$0.0125	None	Most file and object transfers; up to ~25M objects per task
Enhanced	$0.015	$0.55	Very large S3-to-S3 transfers; tens of millions to billions of objects

There is no separate charge for the DataSync managed transfer infrastructure and no per-agent fee. For an on-premise source you deploy a DataSync agent (a VM or EC2 instance) that you run on your own infrastructure; the agent itself carries only the underlying compute cost where you run it, not a DataSync licence fee.

The charges DataSync does not include

The flat per-GB fee is only the DataSync line. A realistic DataSync cost model has to add the charges that land on the source and destination services, which on high-object-count transfers can exceed the DataSync fee itself.

Charges that sit alongside the DataSync per-GB fee

Charge	Where it applies	Why it matters
S3 request charges (PUT, COPY, LIST, GET, HEAD)	Source and destination S3	Dominates on many-small-files transfers
Destination storage	S3, EFS, FSx	The migrated data's ongoing storage cost
Cross-Region / internet data transfer out	Where source and target differ in Region or cloud	Standard egress rates apply on top
PrivateLink endpoint	If the agent uses a VPC endpoint	Per-hour and per-GB endpoint cost
CloudWatch monitoring	Optional task logging and metrics	Small but non-zero on long-running tasks

The small-files trap

On a transfer of many small objects the S3 request charges can rival or exceed the DataSync per-GB fee. A transfer of 50 million 1 KB objects moves only about 50 GB of data (roughly $0.63 in DataSync Basic mode) but issues 50 million PUT requests on the destination, costing roughly $250 at S3 PUT rates. Always model request charges separately from the per-GB DataSync fee on high-object-count transfers.

Worked 50 TB migration scenario

A representative cost build for moving 50 TB from an on-premise NFS file server to Amazon S3 over a 1 Gbps connection, using DataSync Basic mode for the full transfer (no Snowball Edge), with a moderate object count of roughly 5 million files.

Worked DataSync cost build, 50 TB on-premise NFS to S3, Basic mode

Cost line	Low estimate	Typical estimate	High estimate
DataSync transfer fee (50 TB at $0.0125/GB)	$640	$640	$640
S3 PUT requests (~5M objects at $0.005/1,000)	$25	$25	$25
Source-side internet egress (provider rate)	$0	$2,560	$4,600
DataSync agent EC2 or on-prem VM (1 month)	$0	$70	$150
CloudWatch task logging	$0	$5	$20
DataSync-path total (excluding S3 storage)	$665	$3,300	$5,435

The DataSync fee itself ($640) is small; the cost case turns on the source-side egress rate. Where the on-premise provider charges nothing for outbound traffic (owned data centre, flat-rate transit), DataSync is the clear winner under roughly 50 TB. Where the source charges metered egress, the Snowball Edge physical path (a single $1,800 job for up to 100 TB) becomes competitive above roughly 25 to 40 TB. DataSync and Snowball Edge are complements, not rivals: the standard pattern uses Snowball for the bulk seed and DataSync for the incremental deltas during cutover.

DataSync as the incremental complement to Snowball

The highest-value use of DataSync in a large migration is not the bulk move; it is the incremental replication that keeps the destination current while the bulk seed is in transit and during the final cutover window. The bulk dataset ships physically on Snowball Edge; DataSync then runs scheduled tasks to copy only the files that changed or were added after the seed snapshot, so that cutover requires moving only a small final delta.

For a 12-week migration of a 200 TB file estate with roughly 5 TB of weekly change, DataSync moves about 60 TB of deltas over the programme at a Basic mode fee of roughly $768, on top of a one-off Snowball Edge job at $3,200 for the bulk seed. The combined physical-plus-incremental pattern is materially cheaper than moving the full 200 TB online and far faster than waiting for the deltas to clear over a constrained link at cutover.

How to reduce AWS DataSync cost

Use Basic mode unless the transfer genuinely needs Enhanced mode's very-high object-count throughput. Basic is cheaper per GB and has no per-execution fee.
Model S3 request charges separately. On many-small-files transfers the request cost can exceed the per-GB fee; aggregating small files before transfer cuts it.
Use DataSync for deltas, not bulk. For large one-off seeds above roughly 25 to 40 TB, Snowball Edge is usually cheaper; reserve DataSync for the incremental tail.
Run the agent close to the source to maximise throughput and minimise the transfer window (which keeps agent compute cost down).
Schedule incremental tasks rather than continuous transfer. DataSync charges only on changed data, so well-spaced scheduled tasks avoid re-scanning cost overhead.
Use Direct Connect or a VPC endpoint where high volumes justify it, to control the egress component that sits alongside the DataSync fee.

AWS DataSync is one of the simplest cost lines in any migration: a flat per-GB fee with no hidden compute charge. The discipline is not about the DataSync fee itself, which is small, but about the request charges and source-side egress that sit alongside it, and about using DataSync for the job it is best at, the incremental replication that complements a Snowball Edge bulk seed rather than competing with it.

Q&A

Frequently asked

Q. What does AWS DataSync cost?

A. AWS DataSync charges a flat per-GB fee on data transferred. Basic mode costs $0.0125 per GB. Enhanced mode costs $0.015 per GB plus a $0.55 fee per task execution. For a one-off 50 TB transfer in Basic mode the DataSync fee is roughly $640 (50 TB at $0.0125/GB). DataSync has no separate compute or agent charge; you pay only the per-GB fee plus, in Enhanced mode, the per-execution fee. Standard charges for the source and destination services (S3 requests, cross-Region transfer, storage) apply on top and are not part of the DataSync price.

Q. What is the difference between Basic and Enhanced mode?

A. Basic mode is the original DataSync transfer mode at $0.0125 per GB, suitable for most file and object transfers and for source or destination locations with up to roughly 25 million objects per task. Enhanced mode, at $0.015 per GB plus $0.55 per task execution, is designed for very large Amazon S3 transfers (tens of millions to billions of objects) with higher parallelism and no per-task object-count ceiling. For small or one-off transfers Basic mode is cheaper; for large recurring S3-to-S3 replication at very high object counts Enhanced mode is the supported path.

Q. Does DataSync charge for the data it reads, or only what it writes?

A. DataSync charges per GB of data actually transferred between the source and destination locations. For an incremental transfer (a scheduled task that only moves changed or new files), you pay the per-GB fee only on the changed data, not on the full dataset scanned. This is what makes DataSync economical for ongoing replication during a migration cutover window: the bulk seed can go via Snowball Edge, and DataSync moves only the deltas that accumulate.

Q. What costs does DataSync not include?

A. The DataSync per-GB fee does not include: destination storage cost (for example, S3 storage of the migrated objects), request charges on the source and destination (S3 LIST, HEAD, GET, PUT and COPY operations, which add up on high-object-count transfers), cross-Region or internet data transfer out charges where applicable, PrivateLink endpoint cost if used, and CloudWatch monitoring. On a transfer of many small files the S3 request charges can rival or exceed the DataSync fee itself, so they belong in any DataSync cost model.

Q. When should I use DataSync instead of Snowball Edge?

A. DataSync is the online path: it wins when the dataset is small enough to move over the available bandwidth in an acceptable window, or when you need ongoing incremental replication rather than a one-off bulk move. Snowball Edge is the physical path: it wins on large one-off bulk transfers (typically above 25 to 40 TB depending on the source egress rate) or where bandwidth is the constraint. The common production pattern is both together: Snowball Edge for the initial bulk seed, DataSync for the incremental deltas that accumulate during the migration window and final cutover.

Q. Can DataSync move data between clouds?

A. Yes. DataSync supports transfers between on-premise storage (NFS, SMB, HDFS, self-managed object storage), AWS storage services (S3, EFS, FSx), and other public clouds (it can read from Google Cloud Storage and Azure Blob Storage as object-storage sources). Cross-cloud transfers still pay the DataSync per-GB fee plus the source provider's egress charge and the destination's ingest and request charges. For very large cross-cloud bulk moves, a physical appliance on each side (Snowball Edge plus the destination cloud's import device) is usually cheaper than online DataSync.