Deploy and manage VMs inside an Azure virtual network with full OS-level control.
Choose Virtual Machines when you need full control over the operating system, runtime, and installed software — lift-and-shift migrations of existing servers, applications with custom OS-level dependencies, or workloads that can't be containerized. VMs are also the fallback when a workload doesn't fit the constraints of any managed service.
Avoid VMs when a managed platform can run the workload with less operational burden. If you only need to host a web app, run event-driven code, or operate containers, then App Service, Functions, or Container Apps remove the OS patching, scaling, and security maintenance that VMs require.
| Hosting model | IaaS |
|---|---|
| Container support | Container-compatible |
| Minimum nodes | 1 (2+ for a higher SLA) |
| State management | Stateless or stateful |
| Web hosting | Agnostic |
| Autoscaling | Virtual machine scale sets |
| Load balancer | Azure Load Balancer |
| Scale limit | 1,000 nodes per scale set (platform image); 600 (custom image) |
| Multiregion | Single region only — external router plus multiple VM deployments |
| Virtual network integration | Supported |
| Hybrid connectivity | Supported |
| GPU support | Supported |
| TLS | Set up in the VM |
| Architecture styles | N-tier, big compute (HPC) |
| Required skills | OS administration, networking, security patching |
| Operational overhead | High — full infrastructure management |
| Best for teams | Teams with traditional IT operations experience |
Not individually — a single VM doesn't autoscale. Use Virtual Machine Scale Sets to add or remove identical instances automatically based on metrics or a schedule.
A single VM with premium or ultra disks qualifies for an SLA, but two or more VMs across an availability set or availability zones are required for the higher uptime SLA.
Yes — a VM can run a container runtime such as Docker, but it provides no orchestration. For managed container hosting, AKS, Container Apps, or Container Instances are purpose-built.
This is one candidate from the Azure compute decision guide. The right choice depends on your full requirements — evaluate scaling, cost, and operational fit before committing.
Browse all tools →