Glossary

Egress Fees

An egress fee is a charge for moving data out of a provider's network — billed per gigabyte read back over the internet, separately from what storing the data costs.

The charge attaches to the direction of travel, not to the work done. Reading a file costs the same per gigabyte whether it is read once or read as part of a full evacuation of the account.

How egress is billed

Egress is metered per gigabyte leaving the provider's network to the public internet, aggregated across services and summed over a calendar month. It is a separate line from storage capacity and from request charges, and it is tiered: the rate falls as monthly volume rises, so the marginal cost of the last terabyte is lower than the first.

Three properties follow from that structure. The bill is driven by read volume rather than stored volume, so two accounts holding identical data can have very different transfer costs. The tiering is monthly and does not carry over, so the same annual volume costs more when spread evenly than when concentrated. And because the meter sits on the network boundary, the charge is indifferent to whether the data is a routine download or a migration away from the provider.

Published rates

Internet egress from North America and Europe, at list, as published by the two largest providers:

The 100 GB monthly allowance is aggregated across all services and regions in both cases. Transfer between regions is charged separately and more cheaply — on the order of one to two cents per gigabyte — and transfer between services inside a single region is generally free.

What a restore costs

The tiering makes the arithmetic non-obvious, so it is worth working through. Reading 100 TB (102,400 GB) back from AWS S3 in US East at list rates:

That figure is for one complete read of the data. It is charged again on the next full read, and it is unrelated to how long the data has been stored.

Movements that are not charged

Not every read crosses a billing boundary. Data moving between services within the same region typically does not attract an egress charge, nor does data moving between buckets in the same region. Inbound transfer is not charged at all. Delivery through a content delivery network owned by the same provider is usually billed at a lower rate than direct internet egress.

This matters for how the charge accumulates. An application that reads heavily but stays inside one region can generate very little egress, while a modest amount of external delivery — backup verification, analytics on an external platform, serving media to end users — generates a great deal.

Switching charges under the EU Data Act

Regulation (EU) 2023/2854 treats charges for leaving a provider separately from ordinary transfer pricing. Under the transitional rule now in force, a provider may levy switching charges only up to the costs it actually incurs that are directly linked to the switching process. From 12 January 2027, switching charges are prohibited outright.

The prohibition covers charges for the act of switching. It does not remove ordinary in-life data transfer pricing, which continues to apply to normal reads. Since 2024 the major providers have also offered to waive egress for customers leaving entirely, subject to conditions on notice and on the migration being complete rather than a continuing hybrid arrangement.

Egress and self-hosted storage

Egress fees are a property of the commercial model rather than of the technology. Storage running on infrastructure the organisation owns has no per-gigabyte read charge by construction: the cost of reading is the cost of a network already paid for. Scality states this directly for RING active archive deployments — “No hidden data access fees”.

The practical consequence is that the comparison between hosted and self-hosted storage changes with read volume. At low read rates the two are close. At high read rates, or where a full copy of the data may need to be moved, the transfer line rather than the capacity line is what separates them.