🗾 Google Cloud Regions and Zones: Zonal, Regional and Global Things

What a region and a zone are on Google Cloud, how zone names work, which resources are zonal, regional, multi-regional or global, how to choose, and how it compares with Azure and AWS

💡 How this page was written. From Google Cloud's documentation ("Geography and regions", "Compute Engine regions and zones" and the locations page), read on 6 October 2026. Nothing here was run on an account. Google adds regions often, so the counts below are a snapshot. The same topic on the other clouds: Azure and AWS.

Contents
   The short version
   Choosing a region
      See the list yourself
   What is zonal, regional or global
   A tiny example of the zonal idea
   Side by side
   Common mistakes
   The cheat sheet

The short version

A region is a geographic area, such as us-central1 (Iowa), us-east1 (South Carolina), us-west1 (Oregon), europe-west1 or asia-east1. Google's locations page counts 43 regions, 130 zones and over 200 network edge locations (read on 6 October 2026).
A zone is a deployment area inside a region. Google's definition: a region has three or more zones housed in three or more physical data centers. A few regions (Stockholm, Mexico, Osaka and Montreal) have three zones in one or two data centers, and are being expanded.
A zone's name is its region plus a letter: us-central1-a. The newer AI-focused zones use a longer form such as us-west4-ai2b.
Resources are zonal, regional, multi-regional or global, and you have to know which, because it decides what survives a failure.

Choosing a region

Google's documentation names the things to weigh, and the Region Picker tool (cloud.withgoogle.com/region-picker) lets you filter by them:

Latency. Close to users, or to you.
Failure tolerance. Spread across zones, and for important things, across regions.
Availability of what you need. GPUs and TPUs exist only in some zones, and some machine types too. Check with gcloud compute machine-types list.
Price and carbon. Prices differ by region. Google also publishes how low-carbon each region's power is, and the Region Picker shows it.

Two practical things for a beginner:

The free tier VM is limited to three regions: us-west1, us-central1 and us-east1 (see the account page).
Quotas apply per project and per region. Check the quotas of a region before you plan something big, and remember that requesting more quota is not allowed during the free trial.

See the list yourself

gcloud compute regions list
gcloud compute zones list
gcloud compute zones list --filter="region:us-central1"

(From the documentation, not run here: they need a project with the Compute Engine API enabled.) They print the names and the status of each region or zone (and, for regions, quota usage). Without an account, the locations page shows the same data.

What is zonal, regional or global

This is the heart of Google Cloud's model, and the documentation puts it clearly:

Zonal resources live in one zone. A Compute Engine VM and a persistent disk are zonal, and a disk and the VM it is attached to must be in the same zone. If the zone has an outage, they are affected.
Regional resources are redundantly deployed across several zones of a region, so they survive the loss of one zone. Examples in the documentation: regional managed instance groups and App Engine applications. A static external IP address is a regional resource too.
Multi-regional resources are managed by Google so as to be redundant within and across regions: Cloud Storage (when you choose a multi-region), BigQuery, Firestore and Spanner. They survive the loss of a region, with trade-offs in latency and consistency.
Global resources are not tied to a place: VPC networks and images are global. This is an unusual and nice feature of Google Cloud: one VPC network spans the world, and its subnets are regional.

Google's own guidance on failure:

A zone failure affects the resources in that zone only.
To be safe, deploy redundantly across zones, and replicate data to another region, or snapshot it to a multi-regional storage resource, for disaster recovery.
"Communication within regions will always be cheaper and faster than communication across different regions." Check the network pricing page for what traffic between zones and regions costs: do not assume it is free.

A tiny example of the zonal idea

You create a VM in us-central1-a. Its boot disk is created in us-central1-a too. You cannot attach that disk to a VM in us-central1-b: to move it you make a snapshot and a new disk from it in the other zone. In the console, you can pick the zone "Any" for some resources to let Google choose by capacity, and from the CLI you give --zone us-central1-a, or set a default:

gcloud config set compute/region us-central1
gcloud config set compute/zone us-central1-a
gcloud compute instances create demo-vm --zone us-central1-a --machine-type e2-micro

(From the documentation, not run here. The compute/zone default means you do not have to type the zone every time, which is also how you end up creating things in a zone you forgot you had set: check gcloud config list.)

Side by side

Region = region on all three clouds. Google: 43. AWS: 34 (standard). Azure: over 70 (Microsoft's figure).
Zone = zone on Google, Availability Zone on AWS and Azure. Naming: Google us-central1-a, AWS us-east-1a (plus the zone ID), Azure numbers 1, 2 and 3.
Global network: a Google VPC network is global. On AWS a VPC and on Azure a virtual network belong to one region.
No region opt-in was described in the pages used here. (AWS has opt-in regions, and Azure has restricted sovereign ones.)
Region pairs exist on Azure only. On Google and AWS you choose your own second region.
The "resource group" idea on Azure is a project on Google, with the extra role of being the unit of billing and API switching.

Common mistakes

Putting the disk and the VM in different zones. They must match.
One VM in one zone, and calling it highly available.
Using a region far from the users, or not in the free tier's three regions, and being surprised by the bill.
Mixing up the project default zone and the zone you meant.
Asking for a GPU in a zone that does not have them.

The cheat sheet

Region (us-central1) has 3 or more zones; zone is region plus a letter (us-central1-a). Counts: 43 regions, 130 zones (October 2026).
Choosing: latency, failure tolerance, availability of what you need, price and carbon (Region Picker). Learning: one region, and for the free tier, us-west1, us-central1 or us-east1.
Zonal: VMs and persistent disks (the same zone). Regional: regional instance groups, static IPs, subnets. Multi-regional: Cloud Storage multi-region, BigQuery, Firestore, Spanner. Global: VPC networks, images.
Safe: several zones, plus data replicated to another region.
CLI: gcloud compute regions list, gcloud compute zones list, gcloud config set compute/zone ZONE, --zone on commands.

← Creating a Google Cloud Account: Billing Accounts, Projects and the Free Trial · ↑ All tutorials · Google Cloud Services: A Map for People Who Know Linux →