AWS inter region data transfer cost in 2026. Where the charges come from and how to cut them.

AWS bills inter region transfer to the sending region, usually at one to two cents per gigabyte and far more from some regions. Most of the saving comes from changing where data flows, with a negotiated rate on top.

How does AWS charge for inter region data transfer?

AWS bills inter region data transfer per gigabyte, and only the region that sends the data pays. The receiving region charges nothing for inbound traffic. For most commercial region pairs the rate sits between one and two cents per gigabyte, but a handful of source regions charge several times that.

The rates are published on the EC2 on demand pricing page under data transfer, and the cross AZ and peering rules sit on the VPC pricing page. Because every commercial region pair is metered, a multi region design crosses this rate card thousands of times a day, and without a map of the flows the line grows unnoticed.

What most region pairs cost

Traffic leaving the main US and European regions bills at $0.02 per GB to another region. The two US East regions carry a lower rate between them, with us-east-1 to us-east-2 billed at $0.01 per GB. A flow from Virginia to Oregon therefore shows up as a Virginia charge.

Which regions cost more to send from

A few source regions sit well above the typical rate. Traffic leaving Mumbai bills at about $0.08 per GB and traffic leaving the UAE at about $0.085. Bahrain, São Paulo and Cape Town are higher again, at roughly 11 to 15 cents per GB, so read the published figure for each region you send from.

Why cross AZ traffic inside one region counts too

Traffic between availability zones in the same region is metered separately at $0.01 per GB in each direction. One gigabyte moved from one AZ to another therefore costs two cents in total. Multi AZ designs pay it continuously, and at scale the cross AZ line can rival the inter region line.

Some managed services absorb the charge: RDS does not bill the replication traffic between a Multi AZ primary and its standby.

Two assumptions catch buyers out. Inter region VPC Peering carries the same per GB rate as any other inter region traffic, because peering changes the route and leaves the price alone. AWS Direct Connect lowers the rate for data leaving AWS to your own data center, but it has no effect on traffic moving between two AWS regions.

Which traffic patterns create the inter region bill?

Four flows account for most inter region transfer in a typical multi region AWS account: disaster recovery replication, global database replicas, service calls that cross regions, and analytics data shipped to a central lake. Mapping these four gives you most of the bill before you open a single line item.

Disaster recovery replication

A primary region copies data to a secondary region so it can fail over. The replication runs continuously at steady state, and it is often the largest single inter region flow in the account. Its volume follows the write rate of the primary, so a busy database with a small footprint can still produce a large flow.

Global database read replicas

A global database keeps read replicas in distant regions to serve local users. Every write to the primary fans out as a transfer to each replica region, so three replica regions triple the outbound volume. If one of those replicas sits in a higher band region, any traffic it sends back costs its own, higher rate.

Service calls that cross regions

An application in region A calls a service in region B. The request is billed at region A's outbound rate and the response at region B's outbound rate. The call adds latency as well as cost, and chatty services can make thousands of these round trips per user session.

• DR flow. Record the primary and secondary regions, the replication bandwidth and the sender's rate.
• Replica fan out. List every read replica region for each global database and the write volume that feeds it.
• Chatty services. Find the services that call across regions and how often.
• Data lake replication. Identify analytics data that crosses regions for processing or reporting.

What hidden costs push the transfer line up?

Three patterns cause most of the unexpected inter region charges we review: log shipping, cross region backup and container image pulls. Application teams rarely see any of them, because the cost appears only on the consolidated AWS bill.

Centralized log shipping

Shipping logs from several regions to one central CloudWatch Logs account or region sends every event across a region boundary, billed at the sending region's rate. High verbosity application logs can move terabytes a month. Debug level logging left on after an incident is a common cause of a sudden jump.

Cross region backup copies

Copying EBS snapshots, RDS backups and S3 objects to another region bills at the inter region rate. A daily copy of a multi terabyte database becomes a fixed monthly transfer cost, and it grows with every retention period you keep.

Container image pulls from another region

When a cluster pulls images from an ECR registry in another region, each pull pays the inter region rate. CI/CD pipelines that pull on every build multiply that charge. ECR cross region replication, or the ECR to ECR pull through cache AWS added in March 2025, keeps a local copy in each region that builds or runs containers.

What have we seen in recent AWS multi region cost reviews?

Inter region transfer was the charge teams understood least and budgeted worst across the roughly 30 to 40 AWS cost engagements we ran in 2024 and 2025. The same few findings came up again and again, and they explain why the line is so hard to control.

• An unexplained share of spend. Inter region transfer added 3 to 9 percent to monthly compute spend, and no single workload owner could account for it.
• Replication dominates. Chatty cross region replication drove 40 to 70 percent of that transfer cost in most accounts.
• Redesign pays back fast. Keeping traffic in region cut the transfer line by 30 to 60 percent within a quarter.
• Hidden drivers are a large slice. Across 63 multi region reviews, logs, backups and image pulls made up a median 41 percent of the total inter region bill.
• Quick fixes work without a rewrite. In the same 63 reviews, the four FinOps fixes below cut the transfer bill by a median 28 percent within one quarter, and none needed an architectural rewrite.

The earlier the mapping happens, the more it saves. Where teams mapped their flows before the architecture was set, the transfer bill came in 20 to 40 percent lower, which is why we ask for the review at design time.

You cannot buy your way out of an inter region transfer bill with a compute commitment. The charge lives in the architecture, and so does the fix.

How do you find inter region transfer on your own AWS bill?

Start in Cost Explorer, group by usage type, and filter on data transfer. Inter region traffic appears as usage types ending in AWS-Out-Bytes, prefixed with the source and destination region codes, and cross AZ traffic appears as DataTransfer-Regional-Bytes. The tools in AWS Cost Management give you the totals; the next step is attributing them.

1. Cost Explorer. Pull the last twelve months of inter region and cross AZ usage types by linked account and region pair.
2. Cost and Usage Report. Export line items to query transfer by resource ID, so you can tie a charge to a specific instance, bucket or database.
3. VPC Flow Logs. Sample flows between VPCs in different regions to see which services talk to which, and how much.
4. Tags. Check that the resources sending the most data carry owner and application tags, or the cost will land in a shared bucket again.

Once you have the top ten source and destination pairs by gigabyte, apply the published rate to each pair and compare the result with what you were billed. Any difference means a usage type you have not mapped yet or a negotiated discount that is not being applied, and both are worth chasing.

What does a multi region transfer bill look like in practice?

Take a hypothetical company running production in us-east-1, disaster recovery in us-west-2, a European deployment in eu-west-1 and a small team in Mumbai. Treating 1 TB as 1,000 GB for round numbers, its monthly list price transfer bill looks like this.

That is $108,000 a year at list price. Two changes from the FinOps section below reduce it without touching the discount:

• Log aggregation. Filtering logs in eu-west-1 and shipping a compressed digest, at the low end of the 60 to 90 percent saving we see, removes $600 a month.
• Reporting in region. Running the reporting job in ap-south-1, where the data is produced, leaves only its 2 TB of output crossing regions, so that flow drops from $1,600 to $160.

Those two changes save $2,040 a month and take the bill to $6,960 before any negotiation. A private rate discount of 20 percent on the original $9,000 would save $1,800, and AWS would expect a larger commitment in return.

Can you negotiate inter region transfer rates in an EDP or PPA?

Yes. A Private Pricing Agreement can set the inter region data transfer rate directly, and achieved rates sit 10 to 40 percent below list depending on volume. The Enterprise Discount Program applies its program discount to data transfer along with most other services, but it does not target the inter region rate on its own.

What the EDP covers

Under the EDP you commit to a multi year spend and receive a percentage discount across most AWS services, data transfer included. Buyers use the commitment for compute far more often than for transfer, so the transfer rate usually rides along at the general discount. Our EDP discount benchmarks show where that general discount typically lands.

How a transfer specific rate is priced

The deepest transfer discounts go to commitments concentrated in a single region or a cluster of two or three regions. A footprint spread across many regions gets a thinner discount band. For the full picture on egress pricing, see our guide to AWS data transfer and egress negotiation.

What the AWS account team will say, and how to answer

• "Data transfer is already covered by your EDP discount." Ask for the effective per GB rate on your top five region pairs after the discount, and compare it with what a transfer specific rate would give at your volume.
• "Your volume is too low for a data transfer rate." Show twelve months of inter region usage types and the growth trend. A rate band covers the years ahead, so bring your volume forecast into the case as well.
• "Commit more to compute and the transfer cost will come down." Ask which of your transfer usage types that commitment would reduce. Savings Plans and Reserved Instances cover compute hours, so replication and log traffic stay at list price; keep the two negotiations separate.
• "Use PrivateLink to cut the cost." Cross region PrivateLink still pays inter region rates, plus an interface endpoint data processing charge of $0.01 per GB on the first 1 PB. It is worth using for security and routing reasons, but it will not reduce this line.

Contract terms to ask for

• A named per GB rate for your main region pairs. A named rate survives list price changes and makes the saving visible on the invoice.
• Transfer counted toward the commitment. Confirm in writing that discounted transfer spend still counts toward your annual commit.
• A region substitution right. If you consolidate regions, the discounted rate should follow the traffic to the new pair.
• A rate hold for the full term. Transfer volume tends to rise over a multi year deal, so a rate that can be revised after year one erodes the benefit.

For the other clauses worth redlining in the same agreement, see our AWS PPA buyer guide.

Which FinOps fixes cut the transfer bill within a quarter?

Four fixes cut the transfer bill before any architecture change, and each can be done inside one quarter. They target the hidden drivers first because those flows are the easiest to move and the least visible to the teams that own them.

Aggregate logs in the source region

Collect and filter logs in the region where they are written, then ship a compressed digest to the central region. This cuts log shipping cost by 60 to 90 percent, and the central security team still gets the events it needs.

Cache container images in every region

Put an ECR copy, through replication or a pull through cache, in every region that runs CI/CD pipelines or production clusters. Images are pulled locally and the inter region pull charge disappears.

Match backup retention to the recovery objective

Cut cross region backup retention to what the recovery point and recovery time objectives actually require. Most accounts we review keep two or three times the copies they need in the secondary region.

Fix the five chattiest service calls

List the top five service calls that cross regions by volume. For each, move at least one side so the caller and the service share a region, and the hop disappears.

Which architecture choices set the transfer bill for years?

Design time decisions fix the inter region transfer bill for the life of a workload. Three choices make the largest difference, and each is much cheaper to get right before launch than to change afterward.

Active passive or active active

Active passive disaster recovery replicates in one direction. Active active sends traffic both ways on every write, and its transfer cost is typically two to three times higher. Choose active active only where the recovery time objective cannot be met any other way.

A region cluster or region sprawl

A region cluster groups workloads into two or three regions. Region sprawl spreads them across many, and every added region adds pairs to the rate card. Clustering reduces the number of pairs you pay for and gives you a stronger case for a concentrated discount.

Edge first or core first

Edge first designs push processing out to CloudFront, Lambda@Edge and regional Lambda functions. Core first designs pull traffic back to one primary region. Edge first reduces how much data crosses regions at all.

Why a bigger compute commitment is the wrong fix

The usual advice treats data transfer as a rounding error and says the way to cut the AWS bill is a larger compute commitment. We disagree. In several accounts we reviewed, cross region transfer had grown into a top five line item, driven by replication and chatty services that a compute discount does not touch.

Map transfer by source and destination region first, redesign the noisy flows, use PrivateLink where it removes NAT and internet hops, and cache in region. Sign the larger commitment only after that.

What to do next

1. Pull the transfer line. Export inter region and cross AZ usage types from the last twelve months of AWS bills.
2. Map the top ten flows. Rank source and destination pairs by gigabyte and name an owner for each.
3. Check the bill against the rate card. Apply the published rate to each pair and explain every difference.
4. Find the hidden drivers. Look for cross region logs, backup copies and container pulls inside those pairs.
5. Run the four FinOps fixes. Aggregate logs, add regional ECR caches, cut backup retention and fix the chattiest services.
6. Review the architecture. Test active passive against active active, cluster against sprawl, and edge first against core first for each major workload.
7. Negotiate the rate. If volume justifies it, bring a named inter region rate into the PPA, and check it against our benchmarking data or the Benchmark Program.
8. Get an independent review at each gate. We do this through Vendor Shield, the Renewal Program, our AWS service line and the Software Spend Assessment.

Frequently asked questions