📌 Current exam facts (2026): 40–60 questions · 45 minutes · passing score 700/1000 · 3 domains (Cloud Concepts 25–30%, Architecture & Services 35–40%, Management & Governance 30–35%) · no case studies or labs.
📌 Suggested study plan: 1) Read this guide (1–2 days) → 2) Do the free Microsoft Learn AZ-900 learning path → 3) Take our
practice tests until you score 85%+ → 4) Book the exam. Total: 1–2 weeks.
1. Cloud Concepts (25–30% of the exam)
Deployment models
| Model | Description |
| Public cloud | Services over the internet, shared infrastructure, pay-as-you-go |
| Private cloud | Used by a single organization — more control, higher cost |
| Hybrid cloud | Public + private together (e.g., via VPN/ExpressRoute) |
| Multi-cloud | Using more than one public cloud provider |
Benefits of the cloud
- High availability — minimal downtime through redundancy
- Scalability — vertical (bigger VM) and horizontal (more VMs)
- Elasticity — automatic scaling to match demand
- Fault tolerance — keeps running despite component failures
- Disaster recovery — data and apps survive regional outages
- Agility — deploy resources in minutes
- Economies of scale — lower costs from shared infrastructure
CapEx vs OpEx
CapEx = upfront spending on physical infrastructure. OpEx = pay-as-you-go operational spending. The cloud shifts you from CapEx to OpEx via the consumption-based model — you pay only for what you use.
Service models & shared responsibility
| Model | Example | You manage |
| IaaS | Azure VMs | OS, apps, data (most control) |
| PaaS | App Service, SQL Database | Apps, data only |
| SaaS | Microsoft 365 | Data & access only |
The provider is always responsible for physical security and the underlying infrastructure. You are always responsible for your data and identities.
2. Core Azure Services (35–40% of the exam)
Compute
| Service | What it is |
| Virtual Machines | IaaS — full control, you patch the OS |
| App Service | PaaS for web apps/APIs — managed, auto-scaling |
| Functions | Serverless — event-driven code, pay per execution |
| AKS / ACI | Kubernetes / serverless containers |
| Virtual Desktop | Managed Windows desktops & apps |
Storage
| Service | What it is |
| Blob Storage | Unstructured data (images, videos, backups) — hot/cool/archive tiers |
| Files | Managed file shares over SMB/NFS |
| Disk Storage | VHDs attached to VMs |
| Queue Storage | Message queues between app components |
| Table Storage | Key-value NoSQL storage |
Databases
| Service | What it is |
| Azure SQL Database | PaaS relational SQL database |
| Cosmos DB | Globally distributed multi-model NoSQL |
| SQL on a VM / Managed Instance | IaaS / more compatible SQL options |
Networking
| Service | What it is |
| Virtual Network (VNet) | Isolated network with subnets & peering |
| Load Balancer | Layer 4 (TCP/UDP), regional traffic |
| Application Gateway | Layer 7 (HTTP) + WAF |
| VPN Gateway | Encrypted tunnel over the internet |
| ExpressRoute | Private dedicated connection (not over the internet) |
| Traffic Manager | DNS-based global routing & failover |
| DNS | Domain hosting & resolution |
Identity & management
- Microsoft Entra ID (formerly Azure AD) — identity, SSO, MFA, Conditional Access
- Resource groups — logical containers; a resource belongs to exactly one
- Management hierarchy: Management groups → Subscriptions → Resource groups → Resources
Monitoring & other services
- Azure Monitor — metrics, logs, alerts
- Service Health — status of services affecting you
- Advisor — recommendations (cost, security, reliability, performance, operations)
- Azure AI Services (pre-built AI APIs — formerly Cognitive Services), Azure OpenAI (GPT models), AI Foundry (build & deploy AI), Machine Learning, IoT Hub, Marketplace
3. Security, Privacy, Compliance & Trust (10–15% of the exam)
Defense in depth
Layers from outside in: Physical → Identity → Perimeter → Network → Compute → Application → Data. Data is the innermost, most protected layer.
Zero Trust (3 principles)
- Verify explicitly — always authenticate and authorize
- Use least privilege — give only the access needed
- Assume breach — segment access and monitor continuously
Key security services
| Service | Purpose |
| NSG | Rules that allow/deny network traffic |
| Azure Firewall | Managed, centralized firewall |
| DDoS Protection | Protects against DDoS attacks |
| Key Vault | Stores secrets, keys, certificates |
| RBAC | Roles: Owner, Contributor, Reader |
| Azure Policy | Enforces standards (deny/audit/append) |
| Blueprints | Packages policy + RBAC + templates |
| Defender for Cloud | Security posture, secure score, threat protection |
| Sentinel | SIEM + SOAR — detect & respond to threats |
| Purview | Data governance, compliance manager, information protection |
| PIM | Just-in-time privileged access |
Compliance
- Service Trust Portal — audit reports & compliance documents (ISO 27001, SOC 2, GDPR…)
- Compliance Manager — compliance score and improvement actions
- Encryption: at rest (default for Azure services) and in transit (TLS/HTTPS)
4. Pricing, SLA & Lifecycle (15–20% of the exam)
Cost factors
- Resource type and size
- Location (region)
- Egress (outbound data transfers are billed; ingress is free)
- Reserved capacity — 1–3 year commitments save up to ~72%
- Azure Hybrid Benefit — reuse existing Windows/SQL licenses
Tools
| Tool | Purpose |
| Pricing Calculator | Estimate cost of a specific Azure solution |
| TCO Calculator | Compare on-premises vs Azure total cost |
| Cost Management + Billing | Budgets, alerts, cost analysis |
| Advisor (cost) | Right-sizing & reserved-instance recommendations |
| Tags | Organize resources & allocate costs |
SLAs
- SLA = Microsoft's uptime commitment; breach → service credits
- Single VM: 99.9% · 2+ VMs in availability set: 99.95% · 2+ VMs across zones: 99.99%
- Composite SLA = multiply the SLAs of dependent services (0.999 × 0.999 ≈ 99.8%)
- Preview features have no SLA — don't use them in production
Support plans
| Plan | Best for |
| Basic (free) | Billing & subscription support only |
| Developer | Non-production workloads |
| Standard | Production workloads |
| Professional Direct | Business-critical, proactive guidance |
Lifecycle
- Private preview → Public preview (testable, no SLA) → General Availability (GA) (full SLA & support)
- Microsoft publishes retirement dates for services/features — plan upgrades in advance
5. Azure Management & Governance (30–35% of the exam)
The AZ-900 exam objectives were updated — security, compliance, and pricing topics now live under Azure Management & Governance. Key items:
| Topic | What to know |
| Management groups | Contain subscriptions; policies & RBAC are inherited downward |
| Resource groups | Logical containers; a resource belongs to exactly one group |
| Resource locks | CanNotDelete (no deletion) or ReadOnly (no changes) |
| Tags | Key-value metadata for organization & cost allocation |
| Azure Policy | Enforces rules with Deny / Audit / Append effects |
| Blueprints | Packages policy + RBAC + templates for repeatable compliance |
| ARM templates | Declarative JSON infrastructure-as-code |
| Azure Arc | Manage on-premises & multi-cloud resources with Azure tools |
| Azure Automation | Runbooks for scheduled tasks (e.g., stop VMs at night) |
| Application Insights | Web app performance & dependency monitoring |
| Log Analytics | KQL queries across all resource logs |
6. AI, Copilot & Responsible AI (NEW — January 2026 refresh)
The 2026 AZ-900 refresh added Azure AI Services, Microsoft Copilot, and Responsible AI across all domains. Key things to know:
| Service / Concept | What to know |
| Azure OpenAI Service | Access OpenAI models (GPT, embeddings) as a managed Azure API with enterprise security |
| Azure AI Foundry | Unified platform to build, evaluate, and deploy AI solutions & custom copilots |
| Azure AI Services | Pre-built APIs: Vision (image/OCR), Speech (STT/TTS), Language (sentiment/entities), Document Intelligence, Content Safety, AI Search (formerly Cognitive Search) |
| Microsoft Copilot in Azure | AI assistant that helps manage Azure resources via natural language — assists admins, doesn't replace them |
| Prompt engineering | Crafting effective prompts to get the desired output from generative AI models |
| RAG (Retrieval-Augmented Generation) | Grounding a model's answers in your own data by retrieving relevant documents first |
| Responsible AI (6 principles) | Fairness · Reliability & Safety · Privacy & Security · Inclusiveness · Transparency · Accountability |
🤖
Exam tip: Expect 2–4 questions on AI topics. Know the six Responsible AI principles cold, and be able to match each Azure AI service to what it does. Practice with our
AI, Copilot & Responsible AI test.
⚠️ Important: These notes summarize the official skills. Always combine them with the free official learning path on
Microsoft Learn (search "AZ-900" on learn.microsoft.com) and our
practice tests.