Depending on the architecture, a second Azure region can improve resilience while adding cloud expenses that a failover budget may not account for. For a SaaS platform, where bytes travel matters as much as how many bytes leave. Azure egress pricing depends on the source region, destination, network route, service, and monthly volume.
The useful starting point is to separate customer traffic from replication and other internal transfers. Each follows a different pricing path.
Azure egress pricing starts with the traffic path
Internet egress has a monthly free allowance
Azure generally charges for data sent from Azure to the public internet. The first 100 GB of internet-bound traffic per month is covered by the eligible free tier for customers across Azure regions. Beyond that, Microsoft applies volume tiers based on the source geography and routing option. Data transfer into Azure is generally free.
The allowance is for internet-bound traffic, not every form of outbound transfer. Don’t subtract it from cross-region replication, VNet peering, or a service-specific network charge without checking that service’s billing rules.
Internal transfers belong in separate cost lines
A customer downloading an export creates outbound data transfer. An application replicating a database to another region and a pod calling a service across a peered virtual network create different charges. Even if all three move one gigabyte, they shouldn’t share one spreadsheet rate.
Start with a traffic inventory: record the source and destination, direction, monthly gigabytes, Azure service, and network path. Include private endpoints, and distinguish their service charges from bandwidth charges. Separate routine traffic from failover and migration bursts; this keeps estimates of cloud expenses auditable and highlights large monthly transfers.
2026 internet egress rates by source and route
Microsoft publishes progressive per-GB tiers for internet-bound traffic after the free allowance. The reference bandwidth pricing below is in US dollars, but it isn’t universal across Azure services or contracts.
| Route and source geography | Next 10 TB | Next 40 TB | Next 100 TB | Next 350 TB |
|---|---|---|---|---|
| Premium: North America, Europe | $0.087 | $0.083 | $0.070 | $0.050 |
| Premium: Asia, Australia, Middle East, Africa | $0.120 | $0.085 | $0.082 | $0.080 |
| Premium: South America | $0.181 | $0.175 | $0.170 | $0.160 |
| Transit ISP: North America, Europe | $0.080 | $0.065 | $0.060 | $0.040 |
| Transit ISP: Asia, Australia, Middle East, Africa | $0.110 | $0.075 | $0.070 | $0.060 |
| Transit ISP: South America | $0.120 | $0.085 | $0.080 | $0.075 |
Each tier applies only to the data volume within it, not retrospectively to every gigabyte billed that month. Check the Azure Pricing Calculator for rates beyond these tiers and for your actual deployment.
Premium routing keeps traffic on Microsoft’s network longer
With Microsoft Premium Global Network routing, outbound traffic stays on Microsoft’s network before reaching the public internet. North American and European sources start at $0.087 per GB in the first paid tier; South American sources start at $0.181.
These are source-geography billing groups, not Azure Availability Zones. A customer’s location alone doesn’t determine the rate, so check the deployment’s configuration before estimating cloud expenses.
Transit ISP routing changes the network path
The Azure Routing Preference documentation explains how the transit ISP network differs from Microsoft network routing. Routing Preference eligibility, public IP support, and service configuration matter, so teams should test performance before choosing a route.
For a North American source in the first paid tier, the reference rate is $0.087 per GB with Premium routing and $0.080 with transit ISP routing. This comparison applies only when the service and route are eligible; the route won’t automatically change for every storage endpoint, gateway, or application response.
Cross-region, cross-zone, and peering traffic need separate checks
Region pairs determine replication costs
Azure prices cross-region data transfers by source region and destination pair. Published examples list $0.02 per GB between regions within North America or Europe. A transfer from either area to another continent is listed at $0.05 per GB. These illustrate intra-continental data transfer and inter-continental data transfer pricing; actual rates depend on the region pair.
Other paths cost more. Transfers within Asia, Oceania, or the Middle East and Africa can be $0.08 per GB; within South America, $0.16. Rates between continents also vary by source, so model each replication direction separately. A primary-to-secondary database feed and a reverse-direction recovery sync may have different rates.
Availability Zones aren’t billing zones
Traffic within the same Availability Zone is generally free of bandwidth transfer charges, but don’t assume every cross-zone path is free or billed the same way. Microsoft’s published descriptions of cross-zone treatment differ, and the services involved matter. Confirm the exact deployment in the calculator and against observed charges. A private endpoint may involve separate service or network charges, so don’t automatically count it as an egress fee.
VNet peering creates another distinction: a peering connection itself may have no setup charge, while traffic across it is billed. A Microsoft answer on billable VM traffic also distinguishes inter-zone considerations from VNet peering charges. For an AKS application, map paths between clusters, databases, and shared services before assuming internal calls are free, since misclassification can skew cloud expenses.
A SaaS example shows why one rate falls short
Consider a North American application that sends 10,000 GB of replication data to Europe in a month. At the listed $0.05 per GB intercontinental rate, if it applies to this route under the 2026 pricing assumptions, the transfer costs $500.
If the application also sends 1,000 paid GB to internet users through Premium routing, its first-tier internet bandwidth estimate is $87. Together, these illustrative line items total $587 in modeled cloud expenses, not an all-in forecast. The first 100 GB of eligible internet egress is excluded from the paid 1,000 GB; it doesn’t reduce the replication charge.
The actual bill could also include VNet peering, load balancers, Front Door, storage requests, and data retrieval. Any applicable private endpoint or service charges are separate from these bandwidth calculations, and the endpoint itself doesn’t automatically incur egress. Cached responses may reduce traffic leaving the application origin, so measure these paths rather than treating $587 as an all-in application forecast.
Build a defensible estimate in the Azure pricing calculator
Enter traffic by service and direction
In the Azure pricing calculator, choose the actual deployment regions and add bandwidth alongside compute, storage, networking products, and any private endpoint in the design. Enter internet egress and cross-region transfer as separate assumptions. Confirm the selected route where applicable.
Then model at least two months: expected operation and a stressful month. The second model should include regional failover, a full dataset rebuild, a customer export surge, or another event your architecture permits. Validate 2026 rates for the actual source and destination regions, Azure services, traffic direction, route, and usage tier. Confirm the currency and agreement before approving the budget.
Storage needs its own entries. Microsoft’s Azure Storage network routing guidance describes routing choices for storage accounts, but a blob download can also involve storage transactions or retrieval fees. A bandwidth estimate alone won’t price that request.
Compare estimates with measured usage
Once traffic is live, use Azure Cost Management’s Cost analysis to inspect bandwidth and relevant service charges. Compare estimates with measured cloud expenses. Split views by subscription, resource group, service, and region where available. Set cloud budgets and alerts against expected spend, then investigate deviations in both gigabytes and dollars.
For shared infrastructure, retain the original billing records while allocating network costs to products. Azure cost exports for cross-cloud FinOps can support regular reconciliation when the same SaaS product also runs on another provider. A useful operating metric is cost per customer-facing gigabyte, alongside bytes transferred per API request or tenant.
Blob storage tiers can add charges before data leaves Azure
Hot, cool, cold, and archive tiers have different storage and access economics. Reading data from cooler tiers can add per-GB retrieval charges and transactions. Archive content must be rehydrated before normal access, so a recovery plan also needs time and rehydration costs.
Those fees are separate from network egress. Moving an archived backup to another cloud could involve retrieval, rehydration, cross-region movement, and internet transfer, depending on its path. Azure Blob retrieval and rehydration costs deserve their own estimate beside bandwidth. A cheap storage tier can be an expensive choice for data that customers export frequently.
Reduce outbound traffic without weakening the architecture
Keep routine reads near the users they serve
For cost optimization that preserves resilience, place read-heavy services and suitable data near their primary users. For static files, test whether a content delivery network lowers origin transfers enough to offset its delivery charges and reduce cloud expenses. Measure cache misses, since they still pull data from Azure.
For APIs, inspect response size before changing network routes. Pagination, compression where appropriate, and shorter repeated payloads can reduce billed bytes. Allocating egress costs to API usage helps connect those savings to product-level unit costs instead of a shared network bill.
Make replication intentional
Replicate data the second region needs, at the frequency the recovery target requires. Sending entire objects after small updates, duplicating event streams, or repeatedly rebuilding a standby can make cross-region transfer a standing expense.
Don’t remove replication merely to save bandwidth. Compare the saving with recovery time and data-loss requirements first. If a migration to another provider is planned, check Microsoft’s Azure data transfer fee policies for applicable conditions; don’t assume a general waiver covers routine multi-cloud operation.
Check the bill before a regional cutover
A cutover changes traffic direction. Customer responses may move to a different source continent, while synchronization back to the former primary creates a new cross-region flow. Update the estimate before moving production traffic.
During the first billing cycle, compare forecast and actual gigabytes for each path to reconcile cloud expenses. Investigate unexpected peering, storage, routing, and private endpoint service charges separately, where applicable. That gives the team a usable baseline for the next failover test rather than a single unexplained bandwidth total.
Key takeaways
- Price customer-download traffic by source, route, and volume tier, after the applicable 100 GB monthly allowance.
- Model cross-region replication, peering, and storage retrieval separately from customer downloads.
- Validate the calculator estimate with Azure Cost Management after deployment, especially following failover or migration.
Frequently asked questions
Is Azure data transfer between Availability Zones free?
Don’t apply one answer to every service or network path. Same-zone transfer is generally free, but cross-zone guidance and service-specific billing need verification. Peered-VNet traffic can carry its own charge.
Does the 100 GB allowance cover replication to another region?
No. The allowance applies to internet egress, not cross-region replication. Use the applicable cross-region pricing and check for additional service charges.
Can Routing Preference lower every SaaS egress bill?
No. Transit ISP rates can be lower for eligible internet traffic, but the route must suit the service and performance requirements. Estimate replication and other transfer categories separately.
Conclusion
A second region changes the bill according to the paths your data takes. Keep customer downloads, replication, peering, and storage access in separate cost lines, then price each against its actual region and service.
That makes the expense visible before failover turns it into a surprise.