Data Egress Fees: The Cloud Cost Nobody Budgets For
- 3 hours ago
- 7 min read
Data egress fees are the charges public cloud providers apply when you move data out of their platform, to the internet, to another cloud, or even between regions inside the same provider. They're metered by the gigabyte, they compound silently as your data volumes grow, and they're now the single most-cited reason mid-market companies are pulling workloads off public cloud in 2026.
Most cloud budgets get built around compute and storage, the line items the sales engineer walks you through. Egress sits in a different section of the invoice, often invisible until the workload scales. This guide covers what egress fees actually cost across the major providers, where they hide in your bill, and how to model the impact before it changes your architecture.
What data egress fees are
In cloud pricing, "egress" means data leaving the provider's network. "Ingress", data coming in, is almost always free. That asymmetry is intentional: making it cheap to put data in and expensive to take it out is one of the most effective lock-in mechanisms in the cloud business model.
Egress charges typically apply to:
Internet egress, data sent from the cloud to anywhere outside the provider's network (your users, your data center, another cloud)
Inter-region egress, data moved between geographic regions inside the same provider (US East to EU West, for example)
Inter-availability-zone egress, in some providers, data moved between availability zones inside the same region
Cross-cloud transfers, data moved between cloud providers, which usually shows up as internet egress
The first one is what people think of as "egress." The other three are where the surprises live.
What egress fees actually cost in 2026
Pricing varies by provider, by region, and by volume tier. As working numbers for a US-based mid-market workload:
AWS internet egress: roughly $0.09 per GB for the first 10 TB per month, with tiered discounts down to roughly $0.05 per GB above 150 TB. The first 100 GB per month is free.
Azure internet egress: roughly $0.087 per GB for the first 10 TB, tiered down similarly. First 100 GB free.
Google Cloud internet egress: roughly $0.12 per GB for the first 1 TB, with steeper tiered discounts.
Inter-region transfers (AWS): roughly $0.02 per GB
Inter-AZ transfers (AWS): roughly $0.01 per GB in each direction (so $0.02 round trip)
Those per-GB numbers look small. The trap is what they multiply to.
A single TB of internet egress per day on AWS standard rates lands at roughly $2,700 per month, for one application moving a modest amount of data to its users. Scale that across a mid-market company with 10 applications averaging 500 GB of daily egress, and you're at $15,000+ per month in egress alone, before compute, storage, or premium support.
There has been some regulatory pressure. Under the EU Data Act, hyperscalers have started waiving egress fees, specifically when customers exit a cloud platform entirely. That helps if you're leaving, but it doesn't help anyone running normal operations who's paying egress every month they stay.
Why egress fees exist (the REAL answer)
The standard explanation from cloud providers is that egress fees cover network infrastructure costs. There's some truth to that, moving data costs money, but the pricing math doesn't really hold up. Wholesale internet bandwidth in 2026 runs at a tiny fraction of what providers charge for egress. The actual gap between cost and price is roughly two orders of magnitude.
The honest answer is that egress fees are a deliberate friction against leaving the platform. They make multi-cloud architectures expensive, make data lake exports painful, and make migration projects look like a worse idea than they are. Regulators have started taking notice, competition authorities in both the UK and EU have flagged egress fees as potentially anti-competitive, but practical relief for normal cloud customers remains limited.
This isn't a moral judgment. It's a planning input: egress fees are a feature of the public cloud business model, not a bug to be engineered around at scale. Architectures that ignore them will not survive contact with the invoice.
Where egress hides in your bill
Cloud bills are notoriously dense. A few places egress charges typically lurk:
CloudFront / CDN charges (AWS) — separate egress pricing for content delivery, often where customer-facing data egress concentrates
Cross-region replication — automated backups or DR replication between regions accumulate inter-region transfer costs that compound 24/7
Inter-AZ chatter from microservices — applications that spread services across availability zones can rack up significant inter-AZ transfer that wasn't visible at design time
Logging and monitoring exports — sending logs to an external SIEM, an external observability platform, or a security tool means continuous egress
Backup egress — backing up cloud workloads to off-cloud destinations (a smart compliance practice) generates ongoing egress
The single highest-impact diagnostic move you can make this quarter: ask your cloud team to pull egress charges by service for the last 12 months, broken out from compute and storage. Most teams have never looked at that view, and the answer usually surprises everyone in the room.
How to model egress before scaling
A simple framework that catches most surprises:
Estimate steady-state egress per workload. Average daily GB transferred, multiplied by 30, multiplied by the provider's tier rate. Be honest, most teams underestimate this by 2–3x.
Add inter-region and replication overhead. If you have multi-region failover, include the continuous replication egress. This is often half the egress bill at companies that don't track it.
Add the "what if it works" multiplier. If the workload grows 5x, does egress grow 5x with it? Usually yes. Egress is one of the line items least likely to benefit from economies of scale.
Compare against destination alternatives. Same data living in a colocation deployment has zero ongoing egress cost. We unpack the broader trade-offs in our guide to cloud repatriation, including the destination economics in colocation pricing.
If steady-state egress is more than about 15% of total cloud spend on a workload, that workload is a candidate for architectural review or repatriation. Above 25%, it's almost always cheaper somewhere else.
How to reduce data egress fees
You don't have to leave public cloud to take a meaningful chunk out of egress. A few levers, in rough order of impact:
Co-locate data and compute — if you can move workloads to the same region (or same AZ) as the data they're processing, you eliminate the transfer
Use the provider's CDN for customer-facing traffic — bulk content delivery via CDN is typically cheaper than direct internet egress
Compress aggressively — egress is metered by raw bytes; compression directly cuts the bill
Reserve and commit — negotiate committed-use discounts for predictable egress; this is often a separate negotiation from compute discounts and most customers don't ask
Re-architect chatty microservices — applications that bounce data across AZs repeatedly can often be redesigned to consolidate that traffic
Move data-heavy workloads — for the workloads where egress is structurally heavy (large analytics, big data lake outputs, backup destinations), colocation or private cloud usually has dramatically better economics
This is also where an independent technology advisor changes the math. We work across cloud, colocation, private cloud, and hybrid providers — and we benchmark real contract pricing, including egress, across the market. We're compensated by the providers we place, at the same rate regardless of which is selected. That means the recommendation is built around your workload's economics, not anyone's quota.
The repatriation connection
Egress is rarely the sole reason a company repatriates a workload, but it's almost always one of the top three. The workloads that get pulled out of public cloud first are the ones where the egress bill compounded faster than anyone planned for, data lakes, backup destinations, video and media workflows, AI training pipelines, and steady-state high-volume serving.
If you're already running that pattern math in your head while reading this, that's the signal. Most repatriation projects with strong ROI started with someone looking at the egress line item and asking, "wait, that's how much?"
Frequently asked questions
What are data egress fees in cloud computing?
Data egress fees are charges that cloud providers apply when data moves out of their platform, to the internet, to another cloud, or between geographic regions. Data moving into the cloud (ingress) is almost always free. The asymmetric pricing creates friction against leaving the provider.
How much does AWS egress cost?
AWS charges roughly $0.09 per GB for internet egress on the first 10 TB per month, with tiered discounts down to about $0.05 per GB at very high volumes. The first 100 GB per month is free. Inter-region transfers run roughly $0.02 per GB; inter-availability-zone runs about $0.01 per GB.
Why are cloud egress fees so high?
The honest answer is that egress fees create economic friction against moving workloads off the platform. The actual cost of bandwidth to providers is a small fraction of what they charge. Regulators in the EU and UK have flagged the practice as potentially anti-competitive.
Can I avoid data egress fees entirely?
Not entirely, if you're using a public cloud. But you can substantially reduce them through architecture (co-locating data and compute, using CDN for customer egress, compression), through negotiation (committed-use discounts), and through workload placement (moving data-heavy workloads to colocation or private cloud where egress is effectively zero).
Are egress fees ever free?r
Under the EU Data Act, the major hyperscalers have introduced free egress when a customer is fully exiting the platform. Day-to-day operational egress remains charged.
The bottom line
Data egress fees are the line item that quietly defines whether public cloud math works for your workload at scale. They're invisible in most planning conversations and decisive in most budget conversations 12 months later. The discipline isn't avoiding them — it's modeling them honestly before scaling, and matching workloads to the destinations where their economics actually work.
Want a real read on whether egress is driving your cloud bill? Book a discovery call and we'll audit your cloud spend by service, model your steady-state egress, and benchmark the alternatives, at no cost to you.
Comments