Data egress fees are charges that a cloud or hosting provider bills for data leaving its network, whether to the internet, to your own offices or data center, to another cloud, or in some cases between its own regions. They are usually priced per gigabyte, often in volume tiers, and they are one of the least predictable parts of a cloud bill because they depend on how applications move data rather than on how many resources you run.
At a glance
- Most large cloud providers charge for data going out and do not charge for data coming in.
- Rates vary by provider, destination (internet, another region, another zone) and connection type.
- Egress grows with usage patterns such as serving content, replicating data, backups to another site and running across several clouds.
- Private interconnects, content delivery networks and architecture changes can often reduce it.
- It matters most when you plan a migration, an exit or a multi-cloud design.
What problem it solves
For buyers, the term names a cost that is easy to miss. A cloud estimate built from compute and storage prices can look attractive, and then the first real bill includes a sizeable data transfer line. Egress also affects strategy: if moving a large data set out of a provider is expensive, it becomes harder to switch providers, split work across clouds or bring workloads back on-premises. Understanding egress up front lets you design for it, negotiate it, and put a realistic figure on any future move.
How it works
What is metered. Providers measure bytes leaving their network and bill them at a per-gigabyte rate. Common categories are transfer to the internet, transfer between regions, transfer between availability zones, and transfer over dedicated interconnects. Exact categories and whether each is charged differ by provider and service.
Tiers and allowances. Many providers include a small free allowance each month and lower the per-gigabyte rate as volume rises. Large customers can sometimes negotiate rates as part of a cloud commitment agreement.
Paths that change the price. Traffic leaving over a private connection, often through cloud connectivity services, is often billed at a lower rate than internet egress, with port and circuit charges added. Serving content through a content delivery network (CDN) can shift where the traffic is billed and how much is pulled from the origin. Some providers price egress very differently, including flat or bundled transfer, so compare total cost rather than headline compute prices.
Design choices. Keeping chatty services in the same zone or region, compressing and caching data, avoiding unnecessary replication, and processing data where it lives can all reduce egress. Teams practising FinOps track it as its own cost line so they can see which applications drive it.
For connecting your sites or data center to cloud providers privately, see our cloud connect solution page.
When it matters for buyers
- Planning a migration. Estimate ongoing egress for the new architecture, not just compute and storage.
- Considering an exit or cloud repatriation. Moving large data sets out can be costly unless a waiver applies, so ask early.
- Running across several clouds. Data that moves between providers is often billed as egress by the sending side.
- Serving users or customers from the cloud. Media, downloads, APIs and backups to another location all generate outbound traffic.
- Negotiating a commitment. Egress rates, allowances and exit terms can sometimes be part of the deal.
Questions to ask vendors
- What do you charge for data leaving to the internet, to another region, between zones and over a private interconnect?
- Is there a free monthly allowance, and how do volume tiers work?
- Do you offer any waiver or credit for egress when we move data off your platform, and what are the conditions?
- Can egress rates be negotiated as part of a committed spend agreement?
- Which of our planned services generate cross-zone or cross-region traffic?
- How can we see egress broken down by application or account?
How it differs from bandwidth
Bandwidth on a network circuit, such as a dedicated internet or IP transit service, is usually sold as a rate (for example, megabits or gigabits per second) for a monthly fee, sometimes with burstable billing above a commitment. Data egress fees are usually metered by volume: you pay for each gigabyte that leaves, however fast it moves. A cloud workload can therefore have plenty of bandwidth and still run up a large egress bill, and moving traffic onto a fixed-price circuit does not always remove the cloud provider’s own transfer charge.