GCP network egress and inter-region pricing trips up more teams than any other line on the bill, because data transfer is invisible until the invoice arrives. The principle is consistent across clouds: data coming in is generally free, data going out costs money, and the price rises with distance. On Google Cloud that means internet egress is charged per gigabyte by destination, inter-region traffic is charged when data moves between regions, and even inter-zone traffic within a region carries a small per-gigabyte fee. Understanding which boundary your traffic crosses is the whole game.
This article is part of our Google Cloud cluster. For where networking sits among compute, storage and commitment savings, start with our complete guide to Google Cloud cost optimization, the pillar this piece links up to.
The four kinds of data transfer you pay for
Google Cloud meters traffic by the boundary it crosses. Knowing the tiers tells you where the cost is hiding.
| Traffic type | What it is | Typical cost |
|---|---|---|
| Ingress | Data into Google Cloud | Generally free |
| Internet egress | Data out to the public internet | Per GB, varies by destination continent |
| Inter-region | Between two Google Cloud regions | Per GB, varies by region pair |
| Inter-zone | Between zones in one region | Small per-GB fee each direction |
Internet egress is usually tiered by volume and priced by where the data lands, with traffic to nearer continents costing less than far ones. Inter-region pricing depends on the source and destination region pair. Confirm the current per-gigabyte rates for your specific regions and destinations in the Compute Engine and Cloud Networking pricing documentation before modeling, because they vary widely and change.
Where egress hides on a real estate
The biggest surprises come from architecture, not raw traffic. Cross-region replication of databases and storage buckets generates continuous inter-region egress. Chatty microservices split across regions pay egress on every internal call. Pulling large datasets out of BigQuery or Cloud Storage to an external tool or another cloud is internet egress. Logging and monitoring pipelines that ship data to a central region cross region boundaries. And serving user traffic from a single region to a global audience racks up internet egress to distant continents. Each of these is fixable, but only once you can see it, which is why detailed billing export matters; see Google Cloud Storage classes and lifecycle management for the storage side of the same data picture.
How to cut egress: the architecture moves
Co-locate services that talk to each other. The single biggest win is keeping chatty components in the same region, and ideally the same zone, so internal calls never cross a billed boundary. Cache aggressively at the edge: a CDN in front of static and semi-static content serves repeat requests from cache, so the origin pays egress once instead of every request. Compress and batch: smaller payloads and fewer round trips mean fewer gigabytes. Reconsider cross-region replication: replicate only what disaster recovery genuinely requires, not everything by default. And keep analytics close to the data, running queries in the region where the data lives rather than exporting it.
Want your egress mapped and cut?
Our Google Cloud cost audit traces every gigabyte of egress to its source, ranks the cost by service and region pair, and ships the architecture changes that bring it down. On the performance model you pay only from realized savings. No savings, no fee.
Book a GCP cost audit →CDN and network tiers
Two Google Cloud features directly affect egress cost. Cloud CDN caches content at Google edge locations, so cacheable traffic is served from cache at a lower cost than repeated origin egress, and it improves latency at the same time. Network Service Tiers let you choose between Premium tier, which carries traffic on Google's backbone for most of the journey, and Standard tier, which hands traffic to the public internet sooner at a lower price. Standard tier can cut egress cost for workloads that do not need the performance of the premium backbone, though it changes the latency and routing profile, so test before switching production traffic.
The review we run on engagements
We start from the billing export, breaking egress down by SKU, source service, and region pair so the largest contributors are obvious. Then we map each one to a cause: cross-region replication, chatty inter-region service calls, uncached serving traffic, or large data exports. We co-locate what should be together, put a CDN in front of cacheable serving traffic, trim replication to what disaster recovery requires, and evaluate Standard network tier where latency allows. Finally we set budget alerts on the network cost so a new egress pattern is caught the month it appears; see how to set budgets and alerts in Google Cloud.
Traffic tiers, network service tiers and CDN behavior above reflect Google Cloud as of May 2026. Verify current per-gigabyte pricing for your regions and destinations in Google Cloud documentation before acting, as the platform changes.
The Google Cloud Cost Optimization Field Guide includes our egress mapping queries and the architecture checklist we use to drive data transfer down. It is the downloadable companion to this article.
The short version
Ingress is free, egress is not, and the price rises with distance: internet egress by destination, inter-region by region pair, inter-zone within a region. The cost hides in cross-region replication, chatty multi-region services, uncached serving traffic, and large data exports. Co-locate, cache with a CDN, trim replication, query data in place, and consider Standard network tier where latency allows. When you want egress mapped and cut across the whole estate, that is exactly what our Google Cloud cost optimization service delivers.