Serverless containers built on Kubernetes, without exposing the Kubernetes API.
Choose Container Apps to run containers — microservices, background processors, or event-driven jobs — with autoscaling and scale-to-zero, but without operating Kubernetes. It's built on Kubernetes and KEDA yet hides the cluster, making it the middle ground between App Service and AKS.
Avoid Container Apps when you need direct access to the Kubernetes API, custom controllers, node-level configuration, or a specific CNI plugin — those require AKS. For a single container with no scaling, Container Instances is simpler; for non-containerized web apps, App Service is more direct.
| Hosting model | PaaS (serverless containers) |
|---|---|
| Container support | Container-only |
| Minimum nodes | Serverless |
| State management | Stateless or stateful |
| Web hosting | Agnostic |
| Autoscaling | Scaling rules |
| Load balancer | Integrated |
| Scale limit | 1,000 replicas per revision; 15 environments per region |
| Multiregion | Single region only — external router plus multiple environments |
| Virtual network integration | Supported |
| Hybrid connectivity | Supported |
| GPU support | Supported |
| TLS | Ingress controller |
| Architecture styles | Microservices, event-driven architecture |
| Required skills | Container basics, cloud-native patterns |
| Operational overhead | Low — abstracted Kubernetes management |
| Best for teams | Teams that want modern patterns without Kubernetes complexity |
Yes — it runs on a managed Kubernetes foundation with KEDA-based scaling, but the Kubernetes API and cluster are abstracted away, so you don't manage nodes or upgrades.
Yes. With the right scaling rules an app can scale to zero replicas when idle, so you pay nothing for inactive workloads — useful for event-driven and bursty services.
Choose AKS when you need the full Kubernetes API, operators, service-mesh control, or specific networking. Container Apps trades that control for far lower operational overhead.
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 →