S3 Transfer Acceleration cost: what the speed premium actually buys

Quick answer

S3 Transfer Acceleration adds 0.04 dollars per GB for transfers routed through US, Europe, or Japan edge locations and 0.08 dollars per GB through other edges, on top of standard S3 request and data transfer charges. Uploads, which are normally free, therefore become 0.04 to 0.08 dollars per GB. AWS bills the acceleration premium only when the transfer is actually faster than the direct path. For a team uploading 10 TB per month from distant offices, the premium is about 400 to 800 dollars, so it is worth it when the speed gain removes a real bottleneck, not as a default setting.

Uploading to S3 is free. That single fact shapes a lot of architecture: ingest as much as you like, pay only for storage and requests. Transfer Acceleration breaks that assumption by charging for uploads, and it does so at a rate that can exceed what you pay to download the same bytes. Understanding when that trade is good requires knowing both what it costs and what it actually does.

What it does

Without acceleration, a client in Singapore uploading to a bucket in us-east-1 opens a TCP connection across the public internet to Virginia. Long round-trip times limit throughput because TCP congestion control ramps slowly and any packet loss triggers a backoff. With acceleration enabled, the client connects to a nearby CloudFront edge location instead, and AWS carries the data from that edge to the bucket over its own backbone, where the path is optimized and loss is low.

The gain is therefore a function of distance and network quality. A client already in the same region as the bucket sees no improvement at all. A client 200 ms away on a lossy link can see two to five times the throughput.

The pricing

Transfer	Standard S3	With acceleration
Upload (in)	Free	0.04 per GB via US/EU/JP edge
Upload (in), other edges	Free	0.08 per GB
Download to internet	0.09 per GB	0.09 plus 0.04 per GB
PUT requests	0.005 per 1,000	0.005 per 1,000

AWS applies a fairness rule: if a transfer is not faster over the accelerated endpoint than it would have been directly, the acceleration charge is not applied for that transfer. This protects you from paying for nothing when the client is already close to the bucket, but it does not protect you from paying a premium for a marginal gain.

Worked examples

A media company ingests 10 TB per month of raw footage from field offices in Asia Pacific and South America into a us-east-1 bucket. Without acceleration, ingest costs nothing in transfer; with it, the premium at the non-US edge rate of 0.08 dollars per GB is about 820 dollars per month. If that acceleration turns a 14 hour overnight upload into a 4 hour one and lets editors start work the same day, 820 dollars is easily justified. If the upload already completed overnight either way, it buys nothing.

A SaaS product accepting 2 TB per month of customer file uploads from a globally distributed user base faces a different calculus. At a blended 0.05 dollars per GB the premium is about 100 dollars, and the benefit is a visibly faster upload progress bar for distant users, which is a product quality improvement. Here the cost is small enough that the user-experience argument wins.

The alternatives

Multipart upload with parallel parts is free and often captures most of the benefit. Splitting a 5 GB file into 100 parts of 50 MB and uploading 16 at a time works around single-connection TCP limits without any per-GB premium. Many clients get 3 to 5 times better throughput from parallelism alone. The mechanics and the abort-cleanup trap are covered in S3 multipart upload cost.

A second alternative is placing the bucket closer to the uploaders. If 80 percent of ingest comes from Asia Pacific, a bucket in ap-southeast-1 with cross-region replication to us-east-1 for processing may cost less in total than accelerating every upload, because replication runs at 0.02 dollars per GB between regions rather than 0.08 at the edge. A third is AWS DataSync or a physical transfer appliance for large one-time migrations, where the per-GB economics differ entirely.

When to enable it

Enable Transfer Acceleration when uploaders are consistently more than about 1,000 km from the bucket, when individual transfers are large enough that throughput matters, when you have already exhausted multipart parallelism, and when the time saved has a business value you can name. Do not enable it as a default on every bucket, and do not enable it for downloads where CloudFront is the better answer because it also caches. Test it first: AWS publishes a speed comparison tool that measures accelerated against direct throughput from a browser, and the result will tell you in minutes whether the premium buys anything for your users. Price your storage and transfer against the resource catalog before switching it on estate-wide.

FAQ

How much does S3 Transfer Acceleration cost?

0.04 dollars per GB for transfers routed through edge locations in the United States, Europe, or Japan, and 0.08 dollars per GB through other edge locations, charged on top of standard S3 request and data transfer fees. Because uploads to S3 are normally free, enabling acceleration turns ingest into a metered charge of 0.04 to 0.08 dollars per GB.

Do I pay if acceleration does not help?

No. AWS applies a fairness rule: if a transfer over the accelerated endpoint is not actually faster than it would have been over the direct path, the acceleration charge is not applied to that transfer. This protects against paying when the client is already near the bucket, but it does not protect against paying a full premium for a marginal improvement.

How much faster is Transfer Acceleration?

It depends entirely on distance and network quality. A client in the same region as the bucket sees no improvement. A client 200 ms away on a lossy link can see two to five times the throughput, because connecting to a nearby edge and riding the AWS backbone avoids the slow TCP ramp and congestion backoff that long, lossy internet paths impose.

Is multipart upload a cheaper alternative?

Often, yes. Multipart upload with parallel parts carries no per-GB premium and frequently captures most of the benefit. Splitting a 5 GB file into 100 parts of 50 MB and uploading 16 concurrently works around single-connection TCP limits and commonly yields 3 to 5 times better throughput on its own, which should be exhausted before paying for acceleration.

When should I not use Transfer Acceleration?

When uploaders are near the bucket, when transfers are small enough that throughput is irrelevant, when multipart parallelism has not been tried, or when the workload is downloads rather than uploads. For downloads, a CDN is the better answer because it also caches. Placing a bucket closer to the uploaders with cross-region replication at 0.02 dollars per GB can also beat accelerating every upload.

How does C3X help with S3 transfer cost?

C3X prices S3 configuration from Terraform before deployment, including buckets, storage classes, replication rules, and the transfer paths that determine whether data moves for free or at a premium. Because acceleration converts free ingest into a metered charge, seeing that change priced in the pull request is where the decision deserves a second look.

What to do next

See what a storage setting costs before you enable it. C3X prices Terraform against a live resource catalog. Start with the quickstart.