Azure Global Backbone, Regions, Region Pairs & Availability Zones
AI-103: Developing AI Apps and Agents on Azure
When building AI applications and agents on Azure, it's easy to focus immediately on models, prompts, RAG, tools, and agents.
But underneath all of these technologies is Azure's global infrastructure.
Your AI application needs compute. Your models need infrastructure. Your agents may need access to data, search indexes, APIs, and other Azure services.
Understanding where these resources run - and how Azure provides resilience and connectivity - is therefore an important foundation for developing AI solutions on Azure.
1. What Is the Azure Global Backbone?
Cloud computing may appear to be entirely virtual, but Azure ultimately runs on physical infrastructure.
Microsoft connects this infrastructure through its global network backbone. This worldwide infrastructure allows Azure services to operate across a massive geographic footprint.
We can explore Microsoft's global infrastructure using the Microsoft Global Infrastructure Explorer
Azure's Global Infrastructure
Microsoft currently lists:
| Component | Current figure |
|---|---|
| Azure Regions | 70+ |
| Datacenters | 400+ |
| Lit fiber optic & undersea cable systems | 440K+ km |
| Points of presence | 190+ |
These are Microsoft's current global infrastructure figures. Microsoft continues to expand its infrastructure as demand for cloud and AI workloads grows.
For AI developers, this global infrastructure is important because AI applications can require:
- High-performance compute
- Low-latency access to users
- Large-scale data processing
- Connections between multiple Azure services
- Reliable access to models and AI tools
2. Azure Regions
An Azure Region is a geographically distinct location where Microsoft hosts Azure infrastructure and services.
When you deploy many Azure resources, you select the region in which the resource will operate.
For example:
or:
The region you choose can affect:
- Network latency
- Data residency
- Compliance
- Service availability
- Cost
- Resiliency options
For an AI application, these considerations can be particularly important.
3. Why Does Region Selection Matter for AI Applications?
Imagine you're building an AI application for users in India. Your application might contain:
If the application and its supporting resources are located close to the users, network latency can generally be reduced. But imagine that the application is hosted in one continent while its supporting services and data are hosted far away.
Additional network distance can increase latency.
For interactive AI applications and agents, latency can be particularly noticeable because an application may make multiple calls during a single user interaction.
4. Key Benefits of Azure Regions
Lower Latency
Choosing a region close to your users can help reduce network latency. This is especially relevant for interactive AI applications where users expect rapid responses.
Data Residency
Organizations may have requirements concerning where their data is stored and processed. Azure's global region footprint provides geographic choices that can help organizations address data residency requirements.
Compliance
Region and geography selection can be an important consideration when designing applications that must satisfy regulatory requirements. However, developers should always verify the specific requirements of the Azure service being used.
Service Availability
One of the most important things to remember is:
This is particularly important when working with newer AI services and models. Microsoft provides a regional product availability table so you can check which services are available in specific regions.
For AI developers, this can affect where you deploy:
- AI models
- Foundry resources
- AI services
- Search resources
- Storage
- Supporting application infrastructure
5. Azure Regions and Datacenters
An Azure region is not simply one physical building. A region contains one or more datacenters that host Azure infrastructure.
Conceptually:
For regions that support Availability Zones, the infrastructure can be further separated into physically independent zones. This provides additional options for building resilient applications.
6. Global vs Regional Services
Azure services can have different geographic scopes. Some services operate at a global or nonregional scope, while other services are deployed into a specific Azure region.
For example, when deploying a Virtual Machine, you select the region in which the VM will run.
Similarly, when designing an AI solution, you may need to consider the regions associated with its individual components.
These components may have different regional availability and configuration requirements.
Therefore, the region of one resource does not automatically tell you the region or availability of every other service your application depends on.
7. Azure Region Pairs
Azure also has the concept of Region Pairs. A region pair consists of two Azure regions that Microsoft associates with each other.
For example:
Region pairs can provide benefits for certain Azure services and disaster recovery scenarios.
However, there is an important point for modern Azure architecture:
Microsoft's current documentation notes that many newer regions are nonpaired regions. Therefore, don't assume that every new Azure region automatically has a paired region.
8. Why Are Region Pairs Useful?
For supported Azure services, region pairs can provide several benefits.
Disaster Recovery
Some Azure services can use paired regions for geo-replication and redundancy. For example, Azure Storage supports geo-redundant configurations that can replicate data to another region. This can help protect against large-scale regional failures.
Staged Updates
Microsoft can stagger certain platform updates across paired regions. The objective is to reduce the risk of a single maintenance event affecting both regions simultaneously.
Recovery Prioritization
Microsoft also uses region pairing as part of its broader approach to recovery prioritization during certain large-scale Azure incidents.
9. Important: Region Pairing Does NOT Automatically Give You High Availability
This is an important architecture concept.
Suppose your AI application is deployed in East US. Microsoft has West US as the associated region.
That does not automatically mean your application is running in both regions.
You still need to design and configure an appropriate disaster recovery or multi-region architecture.
Microsoft explicitly states that deploying resources into paired regions does not automatically provide:
- High availability
- Disaster recovery
- Automatic failover
For AI applications, this is especially important because your application may depend on multiple resources.
If your AI application uses:
- A model
- AI Search
- Storage
- Application hosting
- Databases
- APIs
you need to consider the availability and recovery strategy for the whole solution, not just one Azure resource.
10. Azure Availability Zones
Now let's move inside a single Azure region.
An Availability Zone is a physically separate location within an Azure region. Availability Zones have independent:
- Power
- Cooling
- Networking
They are designed to protect applications and data from datacenter-level failures.
The zones are connected through Azure's network infrastructure.
11. How Availability Zones Improve Resilience
Consider an application deployed into only one physical location:
If that datacenter experiences a failure, the application may become unavailable.
Now consider a zone-aware architecture:
If one zone fails, the application can continue operating from the other zones if the application has been designed and deployed appropriately.
This is the key idea behind Availability Zones:
12. Availability Zones for AI Applications
Availability Zones can be relevant when building resilient AI applications.
Imagine an AI application with several components:
If the underlying services support Availability Zones, you can potentially design the solution to reduce dependence on a single physical location.
However, Availability Zone support is service- and region-specific. Not every Azure service supports Availability Zones, and a service may support them in one region but not another.
13. Zone-Redundant Services
Some Azure services can automatically distribute or replicate components across Availability Zones. These are commonly referred to as zone-redundant services.
The exact behavior depends on the service. Therefore, always check the documentation for the specific Azure service you're using.
14. Region Pairs vs Availability Zones
This distinction is extremely important.
| Feature | Availability Zones | Region Pairs |
|---|---|---|
| Scope | Within an Azure region | Between Azure regions |
| Primary purpose | Protection from zone/datacenter failures | Regional disaster recovery capabilities |
| Physical separation | Separate locations within one region | Separate Azure regions |
| Automatic HA | No | No |
| Available everywhere? | No | No |
| Example | East US AZ1 / AZ2 / AZ3 | East US / West US |
Easy way to remember
Region Pairs = Microsoft-associated regions
And remember:
Your application architecture must be designed for the desired level of resilience.
15. Azure Infrastructure and AI-103
Why does this matter for AI-103?
AI-103 focuses on developing AI applications and agents using Microsoft Foundry and Azure services.
The current exam includes planning and managing Azure AI solutions (25-30%), including designing Azure infrastructure for AI applications and agent-based solutions, choosing deployment options, and managing quotas, scaling, cost, monitoring, and security.
This means infrastructure knowledge supports decisions such as:
- Where should an AI workload be deployed?
- Is the required model or service available in the selected region?
- How should the application handle regional failures?
- Should supporting resources use Availability Zones?
- How can latency be reduced?
- How should the solution meet data residency requirements?
- How should the AI workload scale?
- What infrastructure is required for an agent-based solution?
These are not just traditional cloud infrastructure questions. They become part of designing and operating modern AI applications.
16. The Big Picture
Let's put everything together.
The Azure Global Backbone connects Microsoft's global infrastructure.
Azure Regions provide geographically distributed locations where Azure services and resources can operate.
Datacenters provide the physical infrastructure behind those services.
Availability Zones provide physical separation within supported Azure regions.
Region Pairs provide Microsoft-defined associations between certain Azure regions and support capabilities for some Azure services.
17. Exploring Azure Regions, Datacenters and the Global Infrastructure Experience
Understanding Azure regions conceptually is important, but as an AI-103 developer, you should also know where to find this information when designing an actual AI solution.
Microsoft provides an up-to-date list of Azure regions that allows you to inspect the characteristics of individual regions.
18. Azure Regions List
Microsoft maintains an official Azure Regions List that provides detailed information about the regions currently available in Azure.
The list can be used to check:
- Azure region name
- Availability Zone support
- Paired region
- Physical location
- Geography
- Programmatic region name
- Region availability status
The list is organized geographically and can be filtered by areas such as:
- Americas
- Europe
- Middle East
- Africa
- Asia Pacific
Microsoft's current region list also identifies regions that are coming soon and regions where access is restricted to specific customer scenarios.
19. Understanding the Region Status Icons
When looking at the Azure region list, pay attention to the status indicators.
Coming Soon
A region marked as coming soon is not yet generally available.
Microsoft provides this information so customers can understand upcoming infrastructure expansion and availability-zone support.
Restricted Access
Some regions have restricted access. These regions are intended for specific customer scenarios, such as certain disaster-recovery requirements.
Access may require an Azure subscription to be approved for that region.
Standard Region
If there is no special status indicator, the region is generally available through the normal Azure deployment process.
20. What Information Can You Check for a Region?
One of the most useful features of the Azure region list is that you can inspect the characteristics of each individual region.
For example, consider East US. The current Microsoft region list shows:
| Property | East US |
|---|---|
| Availability Zones | Yes |
| Paired Region | West US |
| Physical Location | Virginia |
| Geography | United States |
| Programmatic Name | eastus |
This information can be extremely useful when designing an AI application.
For example, suppose you're deciding where to deploy an AI workload. You can check:
Does the region support Availability Zones?
Does it have a paired region?
What geography does the region belong to?
Is the Azure service I need available there?
21. Physical Location Does Not Mean Exact Datacenter Location
The region list provides a physical location such as Virginia for East US.
However, this doesn't mean Microsoft publishes the exact physical location of every individual Azure datacenter. This is intentional.
Microsoft does not generally expose precise datacenter locations as a public directory because of security considerations.
Therefore, when you see:
you should understand that this identifies the region's general physical location rather than giving you the street address of every datacenter.
22. Availability Zone Support Can Vary
The region list also shows whether a region supports Availability Zones. For example, the current Microsoft list shows:
However, there is an important distinction:
Microsoft explicitly warns that even when a region provides Availability Zones, some services may not support them there.
Therefore, when designing an AI solution, you need to check both:
- Does the region support Availability Zones?
- Does the specific Azure service support Availability Zones in that region?
23. Region Pairs Can Also Be Checked
The Azure region list also shows the paired region, when one exists.
East US | +-- West US
The current Microsoft documentation shows East US → West US, but it also shows many regions with N/A for their paired region.
This is an important update compared with older Azure training material.
Not Every Region Has a Pair
Many newer Azure regions are nonpaired regions.
Microsoft states that many newer regions instead use Availability Zones as their primary resiliency mechanism, and many Azure services can support geo-replication between regions without relying on traditional region pairs.
For example, the current region-pair documentation lists regions such as:
- Austria East → N/A
- Belgium Central → N/A
- Chile Central → N/A
- Indonesia Central → N/A
- Italy North → N/A
- Mexico Central → N/A
So don't memorize the old rule "Every Azure region has a paired region." That is no longer correct.
24. Example: Central India
For AI developers in India, the region list is particularly useful. The current Microsoft region list shows:
| Region | Availability Zones | Paired Region | Physical Location |
|---|---|---|---|
| Central India | Yes | South India | Pune |
| South India | No | Central India | Chennai |
| West India | No | South India | Mumbai |
| India South Central | Yes | Central India | Hyderabad |
This illustrates why you shouldn't simply choose a region based on geographic proximity. You should also consider:
- Availability Zones
- Paired region
- AI service availability
- Model availability
- Data residency
- Latency
- Cost
- Disaster recovery requirements
25. Azure Global Infrastructure Experience
Microsoft also provides an interactive Azure Global Infrastructure experience.
Instead of looking at the information only as a table, you can explore Azure's infrastructure visually using a globe, map, or table.
You can inspect:
- Regions
- Geographies
- Networking
- Points of presence
- WAN links
- Sustainability information
- Compliance information
- Availability Zones
- Disaster recovery characteristics
This makes it a useful resource when learning how Azure's infrastructure is distributed globally.
26. Exploring Compliance
The Azure Global Infrastructure experience can also be used to explore compliance-related information. For example, you can filter for standards and regulations relevant to your organization.
Regional Compliance
- GDPR
- Regional data requirements
Industry Compliance
- HIPAA
- GxP
- PCI DSS
Global Compliance
- SOC
- ISO standards
This can be useful when deciding where to deploy an AI application. However, remember that compliance is not simply a property of an Azure region.
You must also consider:
- The specific Azure service
- Your configuration
- Your data
- Your application architecture
- The applicable regulatory requirements
27. Exploring Disaster Recovery
The infrastructure experience can also help you investigate resiliency characteristics. For example, you can explore where:
- Region pairing exists
- Availability Zones are available
- Regions are available or coming soon
This can help when you're designing an AI application's disaster-recovery strategy.
However, the infrastructure map is only the starting point. Your actual disaster-recovery architecture must consider the individual services used by your application.
28. Exploring Sustainability
Microsoft also provides information about sustainability initiatives associated with its global infrastructure.
The infrastructure experience can be used to explore sustainability-related information, including projects associated with renewable energy sources such as:
- Wind
- Solar
- Mixed energy sources
This is increasingly relevant when organizations evaluate cloud infrastructure not only from a technical perspective, but also from environmental and sustainability requirements.
29. Availability Zone Filtering
The infrastructure experience can also help identify regions based on Availability Zone support. You can distinguish between regions where Availability Zones are:
- Available
- Coming soon
- Not available
This can be useful when comparing potential deployment locations. For an AI application where high availability is important, you might prefer a region where the required AI services support Availability Zones.
But remember:
Always verify the specific service.
30. Virtual Microsoft Datacenter Tour
Microsoft also provides an interactive virtual datacenter experience. This lets you explore a representation of a Microsoft datacenter and learn about the physical infrastructure behind Azure.
You can explore areas such as:
- Datacenter facilities
- Networking
- Power infrastructure
- Cooling
- Physical security
- Server infrastructure
This is optional for AI-103 preparation, but it can be useful if you want to understand what actually exists behind the cloud services you use.
31. Why This Matters for AI-103
As an AI-103 developer, you may eventually build an architecture such as:
Choosing the region is therefore not just a deployment checkbox. You may need to answer:
- Is the required AI service available in this region?
- Is the required model available?
- Does the region support Availability Zones?
- Does the specific service support Availability Zones?
- Is there a paired region?
- Does the deployment satisfy data-residency requirements?
- Is the latency acceptable for the application's users?
- What is the disaster-recovery strategy?
These are all practical considerations when developing production AI applications and agents on Azure.
32. Practical Workflow for Choosing an Azure Region
When designing an AI-103 solution, a useful workflow is:
Step 1 - Identify your users
Where are the users of the AI application located?
Step 2 - Check latency requirements
Choose regions that provide acceptable network latency.
Step 3 - Check AI service availability
Verify that the required Azure AI services are available.
Step 4 - Check model availability
For generative AI solutions, verify that the required model and deployment options are available in the intended region.
Step 5 - Check Availability Zones
Determine whether the region supports Availability Zones and whether the services you need support them.
Step 6 - Check region pairing
Determine whether the region has a paired region and whether your architecture benefits from paired-region capabilities.
Step 7 - Check compliance and data residency
Verify that the selected region and services satisfy your organization's requirements.
Step 8 - Evaluate cost
Compare the pricing of the required services across candidate regions.
33. AI-103 Exam Takeaways
Azure Regions
- Azure currently has 70+ regions.
- Regions are geographically distinct Azure locations.
- Region selection can affect latency, data residency, compliance, cost, and service availability.
- Not every Azure service is available in every region.
Azure Availability Zones
- Availability Zones are physically separate locations within an Azure region.
- They have independent power, cooling, and networking.
- They help protect against datacenter-level failures.
- Availability Zone support varies by region and service.
Azure Region Pairs
- Region pairs are Microsoft-associated Azure regions.
- Not every region has a pair.
- Some Azure services use region pairs for geo-replication and disaster recovery capabilities.
- Microsoft can stagger updates between paired regions.
- A paired region does not automatically provide high availability or disaster recovery for your application.
Finding the Current Facts
- Azure's region list is the place to verify current regional characteristics - Availability Zone support, paired region, physical location, geography, and region status.
- Not every Azure region has a paired region. Many newer regions are nonpaired.
- Region-level Availability Zone support doesn't guarantee that every service supports zones. Always check the service-specific documentation.
- A region's physical location is not an exact public datacenter address. The published location represents the Azure region's geographic location.
- Region selection is an architectural decision. For AI applications, consider service/model availability, latency, compliance, resilience, and cost - not just geographic proximity.
Quick Revision
| Concept | Remember This |
|---|---|
| Azure Global Backbone | Global physical and network infrastructure connecting Azure |
| Azure Region | Geographically distinct Azure location |
| Datacenter | Physical infrastructure hosting Azure services |
| Availability Zone | Physically separate location within a region |
| Region Pair | Microsoft-associated Azure regions |
| Latency | Deploy appropriate resources close to users |
| Data Residency | Region selection can help address geographic requirements |
| Service Availability | Services and models vary by region |
| High Availability | Requires appropriate application architecture |
Where to Check
| Question | Where to Check |
|---|---|
| Is the region available? | Azure Regions List |
| Is it coming soon? | Region status |
| Is access restricted? | Region status |
| Does it support Availability Zones? | Azure Regions List |
| Does it have a paired region? | Azure Regions List |
| Where is it physically located? | Azure Regions List |
| Which geography does it belong to? | Azure Regions List |
| Is my AI service available there? | Azure service documentation |
| Is my model available there? | Model/service availability documentation |
| What compliance options exist? | Azure compliance/infrastructure resources |
| What global infrastructure exists? | Azure Global Infrastructure |
Final Takeaway
When developing AI applications and agents on Azure, don't think about the AI model in isolation.
A production AI solution is usually a collection of interconnected components:
The Azure Global Backbone provides the underlying connectivity.
Regions determine where resources operate.
Availability Zones provide physical fault isolation within supported regions.
Region pairs provide Microsoft-defined relationships between certain regions for specific platform and service capabilities.
Don't treat Azure regions as simply a list of places where you can deploy resources - a region affects the architecture of your AI application. When evaluating a region, think:
And remember that Azure's infrastructure is constantly evolving. Use Microsoft's current Azure Regions List and Global Infrastructure pages rather than relying on static numbers or screenshots from older training material.
Understanding these concepts gives you an important foundation for designing, deploying, and managing the AI applications and agents covered throughout AI-103.
AI-103 - Table of Contents
Develop AI apps and agents on Azure with study guides covering every exam domain.