<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=1705902170274878&amp;ev=PageView&amp;noscript=1">

Key takeaways

  • Each account gets 100 GB of free egress a month, aggregated across services and regions. China and GovCloud are excluded.
  •  Internet egress then starts at US$0.09 per GB and tiers down to US$0.05 above 150 TB.
  • There is no separate bandwidth meter on AWS. What is called AWS bandwidth cost is the same tiered transfer charge.
  • Cross-zone, cross-region and NAT gateway traffic run on separate meters. None of it counts towards the internet egress tiers.
  • Private interconnection carries a lower per-GB rate than public internet egress. One insurer cut internet egress cost by around 30 per cent.

AWS meters outbound traffic only, so data transfer into AWS is free. Services in the path, such as NAT gateways, can still charge for processing in both directions.

Egress rarely appears as its own line on an AWS invoice. It sits inside a category called EC2-Other. Most teams only spot it once the bill has already grown. The charge itself is simple enough: data moving into AWS is free, and AWS meters data moving out per gigabyte.

What complicates it is that AWS bills outbound traffic through several separate meters, each carrying its own rate. This guide sets out what AWS charges for data egress, how Amazon S3 fits in, and the practical levers for bringing the total down.

What are AWS egress costs?

AWS bills you when data leaves its network and charges nothing when data arrives. Egress is that outbound direction: traffic from an AWS service to the public internet, to your own data centre, or to another provider. Ingress carries no data transfer charge, which is why AWS ingress cost does not appear as a transfer line item. Services in the path can still bill for processing it.

AWS meters the charge per gigabyte and aggregates it across services. Traffic leaving Amazon EC2, Amazon S3, Amazon RDS, and Lambda feeds one tiered meter rather than four independent ones.

Two things make this harder to predict than it sounds. Rates differ by region, so identical workloads in different places produce different bills. And several categories of internal traffic carry their own charges, which means data can cost money without ever leaving AWS.

Types of data transfers that incur egress fees on AWS

There are transfer patterns that account for most egress fees on AWS. AWS meters each one differently, and the distinctions matter when you are tracing where a bill came from.

Transfer type

Charged?

Notes

Out to the public internet

Yes, tiered

First 100 GB per month free, then from US$0.09 per GB

Out to your own data centre

Yes

Standard internet egress rates apply unless a private link carries it

Between Availability Zones (AZs)

Yes

US$0.01 per GB in each direction, so US$0.02 for a round trip

Between AWS regions

Yes

US$0.01 to US$0.02 per GB between US regions. Rates from Asia Pacific and South America run materially higher

Within one AZ over private addresses

No

Free between instances using private IP addresses

Through a content delivery network (CDN)

Yes, separate rate

Amazon CloudFront runs on its own tier, with a terabyte free each month

The cross-zone charge is the one teams miss most often. A multi-AZ design that queries a database in another zone pays for every response, and that traffic never touches the internet. Azure and GCP meter the same pattern at different rates.

AWS data transfer and bandwidth pricing

There is no separate bandwidth meter on AWS: AWS bandwidth cost, AWS network pricing and egress fees all describe the same tiered per-gigabyte charge on outbound data transfer.

The structure has two parts. AWS aggregates outbound internet traffic across every service into one tiered meter, so a month's usage from S3, EC2 and Lambda combines before AWS applies a rate. Separate meters then stack on top, and they do not count towards those tiers.

Separate meter

Rate

Cross-AZ transfer

US$0.01 per GB each direction

Cross-region transfer

US$0.01 to US$0.02 per GB

NAT gateway processing

US$0.045 per GB

Amazon CloudFront

US$0.085 per GB for the first 10 TB, a terabyte free monthly

Network address translation (NAT) gateway processing is the surcharge that surprises people. Traffic from a private subnet to the internet pays the gateway rate and the egress rate together.

Separate meters stack up quietly. Model what your outbound traffic actually costs.

 

Amazon S3 egress costs per GB

Amazon S3 uses the same tiered rates as the rest of AWS. No separate S3 egress cost table exists. S3 traffic to the internet feeds the same aggregated meter as everything else, starting at US$0.09 per GB once the free allowance runs out.

What differs is which paths are free:

  • S3 to CloudFront: Free. This is the reason CDN delivery is the standard pattern for content served at volume.
  • S3 to EC2 in the same region: Free.
  • S3 to EC2 in a different region: Cross-region rates apply.
  • S3 Transfer Acceleration: Adds a surcharge per GB on top of standard rates.

The practical consequence is that S3 egress cost is usually an architecture question rather than a pricing question. Serving the same object through CloudFront instead of directly from the bucket changes what you pay without changing the object.

How much does AWS charge for data egress?

AWS prices internet egress in descending tiers, so the per-gigabyte rate falls as monthly volume rises. These are the standard rates for United States and European regions, verified against AWS EC2 on-demand pricing in July 2026.

Monthly volume

Rate (USD per GB)

First 100 GB

Free

First 10 TB

0.09

Next 40 TB (10 to 50 TB)

0.085

Next 100 TB (50 to 150 TB)

0.07

Over 150 TB

0.05


Tiers accumulate across the billing month. Each gigabyte carries the rate of the tier it falls into, not one flat rate across the whole total.
Regional variation is significant. Asia Pacific (Singapore) starts at US$0.12 per GB, a third higher than the United States and Europe. Several regions in South America and Africa also run above the standard rate.

Factors influencing AWS egress costs

Four variables determine what you actually pay, and volume is only the most obvious one.

  • Volume: Charges scale with gigabytes transferred, though the tiering means the marginal rate falls as you grow.
  • Destination: Traffic to the public internet costs most. Cross-region costs less, cross-zone less again, and same-zone private traffic is free.
  • Region: Placement is a cost decision as well as a latency one, since the same workload costs materially more in some regions than others.
  • Transfer method: A dedicated private connection carries a lower per-gigabyte rate than the public internet for the same data. At sustained volume, this is usually the largest single lever available.

Transfer method is the largest lever. See what a private path into AWS looks like.

 

How to reduce AWS egress costs

Five approaches account for most of the savings available, roughly in order of how much they typically return.

  • Put a CDN in front of origin traffic: Serving through CloudFront rather than directly from S3 or EC2 moves traffic onto a different meter. The S3 to CloudFront path itself costs nothing.
  • Keep chatty traffic inside one Availability Zone: Placing application instances, read replicas, and caches in the same zone removes the per-gigabyte cross-zone charge entirely.
  • Use private interconnection for sustained volume: A private link carries a lower per-gigabyte rate than internet egress. Platforms such as Tata Communications IZO™+ Multi Cloud Connect route traffic over dedicated cloud on-ramps instead of internet paths, cutting egress exposure and making performance more predictable.
  • Compress what you send: Gzip or Brotli on web assets reduces transferred bytes directly, and the savings apply to every request.
  • Attribute the cost before optimising it: This is the first step in reducing cloud egress costs on any provider. AWS reports egress inside EC2-Other rather than as its own line.

Impact of egress costs on cloud budgets

Egress distorts cloud budgets because it stays invisible until it is large. It runs on consumption with no commitment mechanism, so you cannot reserve or discount it the way you can compute. It also grows with usage rather than with headcount or capacity planning.

The effect concentrates in particular workloads. Media delivery, cross-provider analytics pipelines, backup and disaster recovery replication, and anything serving large files to end users generate outbound volume well beyond their compute footprint. This also shapes architecture decisions that look unrelated to networking. Teams choose regions, split workloads across providers, or repatriate data based on transfer costs rather than technical merit. Treating egress as a design constraint from the start avoids rebuilding later.

Why egress costs increase in multi-cloud environments

Everything above assumes one provider. Most enterprises run several, and traffic between them is billed as internet egress at both ends.

  • Cloud-to-cloud movement: Data leaving AWS for Azure or Google Cloud pays AWS internet egress rates. There is no discounted inter-provider tier.
  • Replication: Keeping a dataset synchronised across providers means paying egress on every change, continuously rather than once.
  • Backup: Backing up to a second provider is sound practice and expensive. The full dataset crosses a metered boundary on each cycle.
  • Analytics: Pipelines that read from one provider and process in another move large volumes on a schedule. The cost scales with the data, not the query.
  • Disaster recovery: Standby environments in a second cloud need current data. Replication runs whether or not the failover is ever used.

Each of these is a deliberate architectural choice made for resilience or capability. The transfer cost is a consequence rather than a mistake. What changes the economics is the path the traffic takes.

How Tata Communications helps reduce egress costs

Egress economics change when traffic stops crossing the public internet. IZO™+ Multi Cloud Connect (MCC) provides private, direct connections into AWS and other cloud service providers from more than 700 access locations and over 160 cloud on-ramps. Setup takes under 10 minutes.

One insurer consolidating a fragmented multi-cloud estate onto a common multi-cloud hub reduced internet egress cost by approximately 30 per cent, alongside 25 to 30 per cent lifecycle cost savings. For traffic moving between providers, the Tata Cloud Router, part of the MCC Flex variant, carries it cloud-to-cloud without hair-pinning through your own data centre.

Talk To Our Team Now.

See how IZO+ Multi Cloud Connect provides private, direct connectivity into AWS. View AWS Direct Connect

Discuss your transfer volumes, regions and architecture with our cloud connectivity specialists. Schedule A Conversation

Model what private connectivity would save against your current AWS egress spend. Calculate Your Savings

Frequently asked questions

What is AWS egress cost per GB?

Internet egress from AWS starts at US$0.09 per GB after a free allowance of 100 GB each month, tiering down to US$0.05 per GB above 150 TB. Full tier breakdown is in the pricing table above. Some regions cost more, with Asia Pacific (Singapore) starting at US$0.12 per GB.

What is the difference between AWS ingress and egress cost?

Ingress is data entering AWS and egress is data leaving it. AWS does not charge for ingress, so uploading, importing or migrating data into AWS costs nothing in transfer fees, because AWS meters outbound traffic only. Every major cloud provider prices this way. It is one reason moving a large dataset in costs almost nothing, while moving the same dataset out later can cost a great deal.

How can I avoid AWS egress fees entirely?

Avoiding them entirely is unrealistic for most production workloads, but several paths genuinely cost nothing. Data transfer into AWS is free. So is traffic between instances in the same Availability Zone over private addresses, S3 to EC2 within one region, and S3 to CloudFront. Designing so that high-volume paths fall into those categories removes most of the charge rather than trimming it.

Does the 100 GB free allowance apply per service or per account?

It applies per account, aggregated across all AWS services and all regions. Outbound traffic from S3, EC2, RDS and Lambda draws on the same 100 GB before billing begins. Splitting a workload across services or regions therefore gains you nothing. The allowance excludes the China and GovCloud regions, and it runs permanently rather than forming part of any introductory free tier.

Why does egress cost more than expected on multi-region setups?

Because several meters run at once. Cross-region replication carries a per-gigabyte charge in its own right. Cross-zone traffic inside each region adds more, and any traffic reaching end users then pays internet egress on top. A multi-region design can therefore pay three separate charges on data that started life as a single write, so replication topology deserves a cost review before it goes live.

Schedule a Conversation
Thank you for reaching out.

Our team will be in touch with you shortly.