AZ-900 Study Guide 2026
This guide covers what you need to know to pass AZ-900 in 2026 — the domain breakdown, what is actually tested in each area, how to study efficiently, and what changed in the January 2026 update.
Exam overview
- Questions: 40–60 (exact number varies per test)
- Time: 45–60 minutes
- Passing score: 700 / 1000 (scaled scoring — aim for 80%+ on practice tests)
- Format: Multiple choice, drag-and-drop, scenario-based, case studies
- Retake policy: Wait 24 hours after first failure; restrictions apply after multiple fails
- Expiry: AZ-900 is a Fundamentals certification and does not expire
- Delivery: Online proctored or at a Pearson VUE test centre
Domain weights — the most important thing to understand
The three exam domains are not equally weighted. Most people who fail do so because they studied the wrong proportions.
| Domain | Weight | Study priority |
|---|---|---|
| Domain 1: Cloud Concepts | 25–30% | Solid foundation, but don't over-invest here |
| Domain 2: Azure Architecture & Services | 35–40% | Heaviest domain — go broad, know each service's use case |
| Domain 3: Management & Governance | 30–35% | Most under-studied — give it equal time to Domain 2 |
Domain 1: Cloud Concepts (25–30%)
What the exam tests
- Cloud definition: The NIST 5 characteristics — on-demand self-service, broad network access, resource pooling, rapid elasticity, measured service
- Shared Responsibility Model: Who manages what in IaaS, PaaS, and SaaS. Customer always owns data and user access. Microsoft always owns physical security.
- Service models: IaaS (you manage OS and above), PaaS (you manage app and data only), SaaS (you manage data and access only)
- Deployment models: Public (shared, OpEx), Private (dedicated, CapEx), Hybrid (connected), Multi-cloud (2+ providers)
- 8 cloud benefits: High Availability, Scalability, Elasticity, Reliability, Predictability, Security, Governance, Manageability
- CapEx vs OpEx: On-premises = CapEx. Cloud = OpEx. Know scenario-based questions.
- NEWServerless computing: Event-driven, scales to zero, pay per execution, no server management
- NEWHA vs DR / RTO vs RPO: High Availability minimises downtime; Disaster Recovery recovers from catastrophic failure. RTO = restore time. RPO = data loss window.
- NEW6 Rs of migration: Rehost, Refactor, Rearchitect, Rebuild, Replace (SaaS), Retire
Domain 2: Azure Architecture & Services (35–40%)
Core infrastructure
- Regions, Availability Zones, Region Pairs — AZs protect DC failure; Region Pairs protect regional failure
- Management Groups → Subscriptions → Resource Groups → Resources. Rules flow down, billing flows up.
- NEWARM Templates (JSON) and Bicep: declarative IaC, idempotent, version-controlled. Bicep compiles to ARM JSON.
Compute
- Virtual Machines: deallocated ≠ stopped (stopped from OS still bills). Scale Sets for auto-scale.
- App Service: PaaS web hosting, no OS management
- Azure Functions: serverless, event-driven, pay per execution
- ACI: fastest way to run a container, no cluster needed. AKS: managed Kubernetes at scale.
Networking
- VNets, Subnets, NSGs (L4, lower number = higher priority)
- VPN Gateway: encrypted over internet. ExpressRoute: private line, NOT over internet.
- DDoS Basic (free). Standard (paid, per-VNet, ML-tuned).
- NEWAzure DNS: public zones (internet-facing) vs private zones (internal VNet resolution)
Storage
- Redundancy: LRS → ZRS → GRS → GZRS (most resilient)
- Blob tiers: Hot → Cool (30d min) → Cold (90d min) → Archive (180d min, hours to rehydrate)
Identity & Security
- Entra ID: cloud identity, SSO, MFA, Conditional Access
- NEWPasswordless authentication: Windows Hello, FIDO2, Microsoft Authenticator — co-equal with MFA and SSO
- NEWManaged Identity: System-Assigned (1 resource, deleted with it) vs User-Assigned (independent, shareable). No credentials in code.
- NEWEntra ID roles (directory) vs Azure RBAC roles (resources) — completely separate systems. Global Admin ≠ Owner.
Other services tested
- NEWMessaging: Service Bus (reliable delivery) vs Event Hubs (high-throughput streaming) vs Event Grid (event routing)
- NEWAzure AI Foundry (renamed from Azure AI Studio): build and deploy AI models including GPT-4
- Azure SQL DB (PaaS, new apps) vs SQL Managed Instance (near-100% compat, lift-and-shift) vs Cosmos DB (globally distributed NoSQL)
Domain 3: Management & Governance (30–35%)
Cost tools — the most-tested three-way distinction
| Tool | When to use it |
|---|---|
| Pricing Calculator | BEFORE deploying — estimate future costs |
| TCO Calculator | BEFORE migrating — compare on-prem vs Azure |
| Cost Management + Billing | AFTER deploying — track actual spend |
| Azure Advisor | Anytime — free recommendations across 5 pillars |
Cost reduction
- NEWSavings Plans: flexible (any compute, commit $/hr). Up to 65% savings.
- Reservations: specific VM SKU + region, 1 or 3 years. Up to 72% savings.
- Hybrid Benefit: bring existing Windows/SQL licences. Up to 85% savings.
- Spot VMs: spare capacity, 90% discount — can be evicted with 30s notice.
Governance
- Azure Policy effects: Deny, Audit, Append, Modify, DeployIfNotExists, AuditIfNotExists, Disabled
- Resource Locks: CanNotDelete and ReadOnly. Override RBAC — even Owner cannot delete a locked resource.
- Tags: max 50 per resource, not inherited by children, enforce via Policy
- Blueprints: versioned, bundled RG + ARM + RBAC + Policy. Tracked assignments.
RBAC
- Roles: Owner > Contributor > Reader > User Access Administrator
- NEWEntra ID roles manage the directory. Azure RBAC roles manage resources. Separate systems, no automatic overlap.
Monitoring
- Azure Monitor: Metrics (93 days) + Logs (2 years) + Alerts + Action Groups
- Azure Service Health: personalised alerts for your services and regions
- Azure Status (status.azure.com): global public page, not personalised
- Resource Health: health of your specific resource instance
SLAs — know these numbers
| SLA | Max monthly downtime | Example |
|---|---|---|
| 99% | 7.3 hours | Some free services |
| 99.9% | 43 minutes | Single VM |
| 99.95% | 22 minutes | Availability Set |
| 99.99% | 4 minutes | Availability Zones |
Composite SLA (services in series) = multiply individual SLAs. 99.9% × 99.9% = 99.8%. Free and Preview services have no SLA.
Compliance
- Microsoft Trust Center: public overview of Microsoft's compliance and security practices
- Service Trust Portal: download actual third-party audit reports (login required)
- Microsoft Purview: data governance across all cloud environments
How to study efficiently
Recommended study schedule (4 weeks)
- Week 1: Domain 1 — cloud concepts, service models, shared responsibility, CapEx/OpEx, cloud benefits, serverless, HA/DR
- Weeks 2–3: Domain 2 — go through each service area (compute, networking, storage, identity). Domain 2 is the widest; don't rush it.
- Week 3–4: Domain 3 — cost tools, policy, RBAC, monitoring, SLAs. This is the most commonly under-studied domain.
- Final 2–3 days: Full timed practice tests. Aim for 80%+ before booking. Focus Wrong Only and Weak Spots modes.
What the exam actually tests
AZ-900 questions are scenario-based. You will not be asked "define IaaS." You will be asked: "A company needs to run a web application without managing the OS or runtime. Which Azure service fits?" The answer is App Service (PaaS). Knowing the definition of IaaS is not enough — you need to know which Azure service maps to which scenario.
Practice this actively: for every service you study, ask yourself — what problem does it solve? When would I use this instead of a similar service?
Frequently asked questions
Ready to practice?
595 flashcards and 540 practice questions — all weighted to these domain proportions, all updated for January 2026.
See the full bundle at az900prep.com →