Microsoft Azure is Microsoft’s public cloud computing platform, which provides on-demand compute, storage, networking, databases, analytics, identity and AI services from data centers in regions around the world, billed by actual consumption. In cloud fundamentals, Azure is one of the three largest hyperscale providers, alongside Amazon Web Services and Google Cloud, and offers the full range of IaaS, PaaS and SaaS-adjacent services.
Azure launched in 2010 as Windows Azure and was renamed in 2014, reflecting support for Linux and open-source workloads, which now make up a large share of its virtual machines. Its distinctive strengths are deep integration with Microsoft software, including Windows Server, SQL Server, Active Directory and Microsoft 365, and a strong hybrid cloud story that extends Azure management to on-premises and other clouds.
How Azure is organized
Every Azure deployment sits in a four-level management hierarchy. The hierarchy controls where policies, permissions and budgets apply, and it is the first design decision of any Azure environment.
| Level | Purpose | Typical use |
|---|---|---|
| Management group | Groups subscriptions for governance | Apply policy to all production subscriptions |
| Subscription | Billing and scale boundary | One per environment, team or workload |
| Resource group | Logical container with a shared lifecycle | All resources of one application tier |
| Resource | A single service instance | A virtual machine, storage account or database |
All operations pass through Azure Resource Manager (ARM), the control plane that authenticates requests, applies role-based access control and Azure Policy, and deploys resources. Infrastructure is described declaratively in ARM templates or Bicep, or with third-party tools such as Terraform. Identities are managed by Microsoft Entra ID, formerly Azure Active Directory, which also signs users in to Microsoft 365.
Regions, availability zones and resilience
Azure runs in more than 60 regions, each a set of data centers in a defined geography. Many regions contain three or more availability zones, physically separate locations with independent power, cooling and networking, connected by low-latency links. Most regions are also paired with another region in the same geography for platform updates and disaster recovery.
Choosing the right redundancy level is a trade-off between cost, latency and the failures to survive. The table shows the main options.
| Option | Protects against | Typical use |
|---|---|---|
| Locally redundant (LRS) | Disk or rack failure in one data center | Development, easily rebuilt data |
| Zone-redundant (ZRS) | Loss of one availability zone | Production workloads in one region |
| Geo-redundant (GRS) | Loss of an entire region | Backups, compliance copies |
| Geo-zone-redundant (GZRS) | Zone and region failures | Critical data needing both |
| Availability sets | Host and rack maintenance or failure | Legacy VMs in regions without zones |
Core Azure services by service model
Azure offers several hundred services, but most architectures are built from a small core. They map onto the classic service models, with the provider taking over more responsibility as one moves from IaaS toward fully managed services.
| Category | Main services | Service model |
|---|---|---|
| Compute | Virtual Machines, Virtual Machine Scale Sets | IaaS |
| Application hosting | App Service, Azure Functions, Container Apps | PaaS and serverless |
| Containers | Azure Kubernetes Service (AKS), Container Registry | Managed control plane |
| Storage | Blob Storage, Azure Files, Managed Disks | IaaS and PaaS |
| Databases | Azure SQL Database, Cosmos DB, PostgreSQL and MySQL flexible servers | PaaS |
| Networking | Virtual Network, Load Balancer, Application Gateway, Front Door | IaaS and edge |
Under the shared responsibility model, Microsoft secures the physical infrastructure and the platform, while the customer remains responsible for identities, data, network rules and configuration. With IaaS the customer also patches the guest operating system; with PaaS that duty moves to Microsoft.
How Azure compares with AWS and Google Cloud
The three hyperscalers offer broadly equivalent building blocks, so the choice usually depends on existing skills, licensing and specific services rather than raw capability.
| Aspect | Microsoft Azure | Amazon Web Services | Google Cloud |
|---|---|---|---|
| Identity foundation | Entra ID, shared with Microsoft 365 | IAM and IAM Identity Center | Cloud Identity and IAM |
| Hybrid approach | Azure Arc, Azure Local | Outposts, EKS Anywhere | Anthos, Distributed Cloud |
| Licensing advantage | Azure Hybrid Benefit for Windows and SQL Server | License-included instances | Sustained use discounts |
| Typical strength | Enterprise and Microsoft-centric workloads | Breadth and maturity of services | Data analytics and Kubernetes |
Many enterprises run a multi-cloud estate, using Azure for Microsoft-centric systems and identity while running other workloads elsewhere. Azure Arc supports this by projecting servers and Kubernetes clusters from any location into Azure Resource Manager for central policy and monitoring.
Cost models and common pitfalls
Azure is billed per second, per hour or per operation depending on the service. Pay-as-you-go suits variable workloads; reservations of one or three years and savings plans give substantial discounts for steady usage; Spot virtual machines use spare capacity at low prices but can be evicted at short notice. The Azure Hybrid Benefit lets organizations reuse existing Windows Server and SQL Server licenses.
- Ignoring egress charges: data leaving a region or Azure itself is billed and can dominate costs for data-heavy services.
- Flat subscription design: placing everything in one subscription makes access control and cost allocation difficult later.
- Overly broad permissions: Owner rights granted widely undermine role-based access control.
- Assuming a single VM is highly available: resilience requires zones, scale sets or availability sets.
- Idle resources: stopped but not deallocated VMs, orphaned disks and public IPs keep generating charges.
Azure Cost Management with budgets, tagging and regular reviews of reservations keeps spending aligned with real usage.
The idea in brief
Instead of buying and running its own servers, a company can rent computing power from Microsoft and pay only for what it uses, much like paying for electricity instead of owning a power plant. Microsoft Azure is that rental service.
A company can start a virtual server in minutes, store files, run databases or host a website, all in Microsoft data centers spread across the world. It is especially convenient for organizations that already use Microsoft products, because sign-in, licenses and tools fit together.
A practical scenario
A logistics company with 800 employees runs its ERP system on aging on-premises Windows servers and already uses Microsoft 365. It migrates the application servers to Azure virtual machines spread across three availability zones in a nearby region, moves the database to Azure SQL Managed Instance, and connects the head office through a site-to-site VPN.
Using the Azure Hybrid Benefit for existing Windows Server and SQL Server licenses and three-year reservations for the steady workload, the company keeps running costs close to its previous hardware budget. Staff sign in with the same Entra ID accounts they use for email, and a nightly backup to geo-redundant storage provides recovery in a second region. The old server room is decommissioned within a year.