AZ-104 vs AZ-103 in 2026: What Changed for Azure Administrators

AZ-103 belongs to an earlier version of Microsoft’s Azure Administrator path. The current exam is Microsoft AZ-104, and Microsoft’s April 17, 2026 skills outline defines the administrator role around identity and governance, storage, compute, virtual networking, and monitoring.

That makes a direct “AZ-103 versus Microsoft AZ-104” comparison useful only as historical context. Candidates should not try to prepare for both. The better question is which older administrator skills still transfer and what the current Azure Administrator Associate role expects beyond them.

The answer is not that Azure administration became completely different. Core infrastructure responsibilities remain recognizable. What changed is the platform around them: Microsoft Entra ID, governance, managed services, containers, modern monitoring, infrastructure as code, and cloud operating practices are now more integrated into the job.

AZ-103 is history; AZ-104 is the current operational target

AZ-103 was part of the transition from Microsoft’s older certification model into role-based Azure credentials. It tested administration skills appropriate to the Azure platform of its time. Anyone who earned it built knowledge that can still be useful, but the exam itself should not shape a new study plan in 2026.

AZ-104 is maintained as the current Azure Administrator exam. That distinction matters because cloud platforms change continuously. A five-year-old course may explain virtual machines or storage accurately at a conceptual level while missing current identity, governance, networking, container, security, and monitoring expectations.

Use legacy material only when the concept is durable. Verify procedures against current Azure behavior before treating them as preparation.

Identity and governance are now inseparable from administration

Azure administration is not only about creating resources. Administrators manage who can act, where they can act, and which configurations are allowed.

That means Microsoft Entra ID, Azure role-based access control, subscriptions, management groups, resource groups, Azure Policy, resource locks, and cost governance belong in the same operating model. A resource can be technically healthy and still be poorly administered if permissions are excessive or governance is inconsistent.

Cloud identity and least privilege are operational concepts, not just exam terminology. Ask which identity is acting, which role grants the permission, what scope applies, and how the action will be audited.

Storage administration is about access, durability, and lifecycle

Current Azure administrators need more than the ability to create a storage account. They should understand Blob Storage and Azure Files, access controls, redundancy choices, lifecycle management, shared access, encryption, and operational behavior.

Storage scenarios become easier when you start with the workload. Does the application need object storage, a shared file system, or a disk attached to compute? How durable must the data be? Who needs access? What happens if credentials are exposed? How will old data be tiered or removed?

These questions are more reusable than memorizing portal sequences. They also help distinguish administration from architecture: the administrator has to implement and operate the storage design correctly, even when somebody else chose the overall pattern.

Compute now spans virtual machines, scale, and containers

Virtual machines remain important, but Azure administration is not a VM-only role. Administrators need to deploy and manage compute resources, understand availability options, extensions, images, disks, networking, and operational controls.

Containerized workloads also appear in the current role. The administrator does not need to become a full application developer, but should understand how Azure container services are provisioned, connected, monitored, and secured.

Practice lifecycle rather than creation. Deploy a workload, update it, resize it, attach storage, restrict access, monitor it, troubleshoot it, and remove it cleanly. That sequence exposes far more administrator knowledge than a one-time deployment exercise.

Virtual networking remains one of the strongest transferable skills

Networking knowledge from older Azure administration remains valuable because addressing, routing, name resolution, segmentation, and connectivity still determine whether resources can communicate.

The current environment brings more managed networking patterns, private access, hybrid connectivity, DNS options, and security controls into routine administrator decisions. Understanding virtual networks, subnets, routes, and peering as a connected traffic path is more useful than memorizing the portal location of individual settings.

Build a small network with multiple subnets and deliberate restrictions. Break a route, a DNS setting, or a security rule and diagnose the symptom from both ends. Administrators become effective when they can explain the path a packet should take and identify where reality differs.

Monitoring is part of running Azure, not an exam afterthought

A production administrator needs to know when something is unhealthy before a user opens a ticket. Azure Monitor, logs, metrics, alerts, activity records, resource health, and diagnostic settings provide different pieces of that visibility.

Monitoring should answer operational questions. Is CPU pressure causing the slowdown? Did a configuration change occur immediately before the incident? Is a dependency failing? Has storage latency increased? Are requests being denied by policy or networking?

Practice creating an alert that has a meaningful response. Then create the failure condition and confirm that the signal appears where you expect. Observability becomes useful when it is tied to a troubleshooting workflow.

Automation is now normal administrator work

Microsoft expects Azure administrators to be comfortable with the portal, PowerShell, Azure CLI, and Azure Resource Manager or Bicep-based deployment. That does not mean every administrator must become a software engineer.

It means repetitive configuration should not depend entirely on manual clicks. Scripts and declarative templates make changes reproducible, reviewable, and easier to apply consistently across environments.

Start small. Query resource state with CLI or PowerShell. Deploy a repeatable resource group. Parameterize a Bicep template. Compare desired state with what actually exists. The value is not typing syntax from memory; it is understanding what state you are creating and how to recreate it safely.

Administrative boundaries are especially important in teams. An Azure administrator may be responsible for implementing network or identity settings designed by specialists, but still needs enough context to recognize when a requested change creates an operational conflict. That is why broad platform understanding matters even when ownership is shared.

Cost should also be visible during daily operations. Orphaned disks, oversized virtual machines, forgotten public IPs, unnecessary snapshots, and idle resources are administration problems as much as finance problems. Build cost review into the lab so that cleanup and right-sizing become part of the operating routine.

Know the boundary of the administrator role

The Azure Administrator role has broad platform coverage, but it does not make someone an architect, network specialist, security architect, developer, or DevOps engineer automatically. AZ-104 and AZ-305 separate administration from architecture responsibility, while AZ-104 and AZ-700 separate broad platform administration from deeper network-engineering ownership.

The useful progression is not “collect the next exam.” It is to notice which part of administration you are increasingly being asked to design, troubleshoot, or own.

Document the environment before you change it. Record the intended access model, network layout, resource ownership, monitoring signals, and deployment method. Then compare that intended state with what exists after several rounds of changes. This is how configuration drift becomes visible.

For troubleshooting practice, begin with symptoms rather than services. “The application cannot reach storage” should force you to check identity, network path, DNS, permissions, firewall rules, and service health in a sensible order. Cross-domain diagnosis is much closer to administrator work than isolated feature questions.

Create a small Azure environment that forces the domains to interact. Add identities and role assignments. Apply policy. Deploy storage and compute. Create virtual networks. Configure monitoring. Automate at least part of the deployment.

Then operate it for several days. Change permissions. Rotate access. Introduce a network problem. Fill a storage quota. Trigger an alert. Rebuild a resource from code. This makes the role feel like administration instead of a list of services.

Azure Administrator skills are easiest to diagnose against a hands-on environment: if you cannot implement, observe, troubleshoot, or restore a capability there, it is probably still a real gap rather than a missing fact from a checklist.

Backup and recovery belong in normal administration

Protection settings are only useful when the administrator understands what can actually be restored. Different workloads may need different recovery points, retention periods, and restore methods, and those choices should reflect the business requirement rather than a default policy.

Backup and recovery should be included in that operating environment as well. Configure protection for a workload, verify the recovery points, and perform a restore into a safe target. Administration is not complete when a backup policy shows green; the important evidence is that the resource and its data can be recovered in the way the organization expects.

Use the activity log and resource history during troubleshooting. Reconstructing who changed what and when is often faster than inspecting every setting from scratch. This also reinforces why governance, identity, and observability belong together in the administrator role.

Review the current Microsoft skills outline shortly before the exam because Azure services and objective wording continue to change. Use the blueprint as a boundary for final revision, but keep the lab centered on operational behavior. That combination prevents both outdated study and shallow feature memorization.

Use current documentation when a lab behaves differently from older training. The discrepancy is often the lesson: service defaults, supported options, and management experiences evolve, while the underlying administrator responsibility remains to understand the resulting state.

AZ-103 matters because it is part of the history of Microsoft’s Azure Administrator certification, not because candidates still need to study for it. In 2026, AZ-104 is the current target. Keep transferable infrastructure fundamentals, discard obsolete exam mechanics, and build current hands-on skill across identity, governance, storage, compute, networking, monitoring, and automation.