The pricing structure itself adds complexity. Egress charges generally depend on how much data moved, where it went, and which geographic region it originated from. Traffic staying within a single cloud ecosystem may be discounted or included at no additional charge; traffic heading out to the public internet or to a competing provider is typically charged at a higher rate. There is no universal number to quote here as providers differ considerably in their rates, regional variations, free allowances, and how different services interact with one another.
When cloud storage first became mainstream, a lot of early workloads were uncomplicated. Companies moved backups offsite, set up disaster recovery, and largely left the data sitting there. Data mostly went in. Fees for data going out were easy to overlook because not much was going out.
So why does leaving cost money?
Every major cloud provider publishes pricing documentation covering storage and network transfers. Reading that documentation before committing to an architecture is worth the time, and looking beyond the storage price per gigabyte is especially important — that single number gives you an incomplete view of what the service will actually cost once data starts moving. The relevant variables include outbound network transfer rates, regional pricing differences, inter-service networking fees, data retrieval charges, available cost management tools, and whatever migration policies the provider offers.
But moving data between cloud platforms can materially increase transfer costs. If leaving one provider costs significantly more than staying, that price differential can function as an incentive to stay put, even when moving would otherwise make business sense. The term for this dynamic is vendor lock-in, meaning a situation where switching costs become high enough to effectively trap a customer. Some providers have introduced programs that reduce or waive certain migration-related egress fees, but these are not universal, and the terms vary.
Why this catches businesses by surprise
Backup strategies matter more than many teams realize. Incremental backups and block-level replication, both common in modern backup tooling, transfer only the data that has changed since the last backup rather than the entire dataset. Over time, the difference in outbound volume can be substantial.
Here is where I think egress pricing deserves more scrutiny than it usually gets. Many enterprises deliberately spread workloads across multiple cloud providers — not out of confusion, but because it offers resilience, reduces dependence on any single vendor, and lets them match specific workloads to whichever platform handles them best. That is a sound architecture strategy.
The charging logic is not arbitrary, even if it can feel that way. Running a global cloud network requires high-capacity networking equipment, sustained connections with internet service providers, and serious investment in backbone infrastructure that spans continents. That is real overhead, and it does not scale the same way storage does. Storage costs are relatively predictable once you know your volume. Network usage, on the other hand, varies dramatically from one customer to the next — a static archive barely touches the network; a media company streaming high-resolution video hammers it constantly. Because that variation is so extreme, most providers treat network traffic as a separately metered service rather than bundling it into storage pricing.
The vendor lock-in problem nobody talks about loudly enough
By Gary Bernstein
Visibility is a precondition for all of this. Billing dashboards and monitoring tools that surface where outbound traffic is actually occurring give teams the information they need to make decisions. Without that visibility, egress costs accumulate quietly. Data lifecycle policies, which move infrequently accessed data into archival storage tiers, reduce how often cold data gets touched and transferred.
The most effective interventions tend to be architectural rather than contractual. Keeping your applications and storage in the same cloud region eliminates a category of unnecessary transfers before they start. If data does not have to cross a regional boundary to reach the system that needs it, it does not generate a cross-region fee.
What does reducing egress costs actually look like?
Reasonable, maybe. But the practical effect is that a straightforward action such as restoring a backup, downloading files, delivering website assets to end users, replicating data to a second cloud provider — each of those generates outbound traffic, and outbound traffic generates fees. Video streaming, software distribution, AI and machine learning model training that pulls from remote datasets, operating a hybrid environment where some systems live on-premises and some in the cloud: all of it touches egress.
This is a genuinely uncomfortable tension. There is no clean answer. The providers are covering real infrastructure costs; the enterprises are trying to make rational architectural choices; the pricing structure creates friction between those two things that does not dissolve cleanly.
That is what egress is. Data egress, to use the technical term, means data leaving a cloud provider’s network — travelling outward to your users, your systems, or another provider entirely. Its opposite, ingress, is data moving into the cloud. And here is the asymmetry that catches so many organizations off guard: charges are more commonly applied when data leaves a provider’s infrastructure than when it enters. Most major cloud providers allow customers to upload data without charging transfer fees. Getting the data out is another matter.
For organizations delivering content to end users, content delivery networks, systems that cache copies of files closer to the people requesting them which can significantly reduce the volume of data travelling from origin storage each time a user downloads or streams something. The CDN serves the cached copy; the origin storage is only hit when something is new or has expired from the cache.
Before you sign anything
Modern operations are different in kind, not just scale. Businesses now synchronize data across multiple cloud services, replicate backups across geographic regions, serve global user bases with media and software, and increasingly train machine learning models that pull large volumes of data repeatedly through a pipeline. Each of those patterns generates significant outbound traffic. An organization that designed its cost model around early, simpler workloads may find that model no longer holds and not because they are doing anything wrong, but because the shape of what they are building has changed.
You signed up for cloud storage because the monthly price per gigabyte looked reasonable. Then the bill arrived and there was a line item you had not planned for, one that had nothing to do with how much data you were storing. It had to do with how much data had left.
A dataset stored for years may cost almost nothing to hold. The moment you need it somewhere else, the meter starts in a different column.
