Azure Platform Engineering: The Complete Guide to Building an Enterprise Landing Zone

By | Apr 1, 2026 min read

Your Azure bill arrives and one line item is $47,000. Nobody on your team knows which subscription it came from. You trace it back to a developer who spun up a GPU cluster three weeks ago “just to test something.” There was no policy to stop them. There was no budget alert configured. There was no shared networking — they punched a hole through the firewall themselves. That subscription is a silo. You have fourteen more just like it.

The bill is the part finance sees. The part that keeps you up is the question internal audit asks next: which controls applied to that subscription? The honest answer is none, and you can’t prove the other fourteen are any different.

This is what happens when your Azure estate grows without a foundation. Subscriptions accumulate as independent experiments. Security policies are applied by hand, subscription by subscription, whenever someone remembers. Networking becomes overlapping IP ranges and ad-hoc peerings that nobody fully understands. You aren’t running a cloud platform — you’re running a collection of accidents.

An Azure Landing Zone (ALZ) is the architectural foundation that prevents this. It provides the shared services — networking, identity, governance, and logging — that every application needs but no single team should manage alone. In 2026, the standard for building these foundations has shifted from monolithic modules to the Azure Verified Modules (AVM) initiative, emphasizing “Metadata-First” architectures and secret-less CI/CD.

Each layer below gets its own article in this series, with Terraform and Pulumi side by side. The full reading order is at the end.


1. The Anatomy of an Enterprise Landing Zone

A landing zone is not a single resource; it is a multi-layered ecosystem designed according to the Cloud Adoption Framework (CAF). Microsoft divides this ecosystem into eight distinct design areas, ranging from billing and identity to networking and operations.

Platform vs. Application Landing Zones

The most critical architectural distinction is between the Platform and the Application.

  • Platform Landing Zones: These are the shared services subscriptions (Connectivity, Identity, Management). They provide the hub VNet, the Entra ID tenant, and the central Log Analytics Workspace.
  • Application Landing Zones: These are the workload subscriptions where your apps actually run. They are “consumers” of the platform, inheriting its security policies and peering back to the hub.

ALZ Management Group Hierarchy

The hierarchy is the “Gavel” of your landing zone. It determines how policies and permissions cascade down to resources.

flowchart TD subgraph Root["Tenant Root Management Group"] subgraph Platform["[ Platform ]"] P_Sub["- Connectivity
- Identity
- Management"] end subgraph LZ["[ Landing Zones ]"] LZ_Sub["- Corp (Internal)
- Online (External)"] end subgraph Sandbox["[ Sandbox ]"] S_Sub["- (Experimentation)"] end subgraph Decomm["[ Decommissioned ]"] D_Sub["- (Resources to Purge)"] end end

2. Design Area 1: Network Isolation (Hub-and-Spoke)

Networking is the foundation of isolation. In a landing zone, we use a Hub-and-Spoke topology to centralize shared network services and enforce a single egress point to the internet.

Hub-and-Spoke Core Architecture

flowchart TD subgraph Hub["[ HUB VNET ] (Connectivity Sub)"] FW["[ Azure Firewall ]
(Central Egress)"] VPN["[ VPN / ER ]
(On-Premises)"] Bastion["[ Bastion ]"] DNS["[ DNS Private Resolver ]"] VPN --> FW end subgraph Spoke["[ SPOKE VNET A ] (Corp Sub)"] Apps["- App VMs / Private Endpoints
- (UDR to Firewall)"] end Hub <-->|VNet Peering - Transit| Spoke

In 2026, the Azure DNS Private Resolver is the standard for hybrid name resolution, replacing the need for complex, manual VM-based DNS forwarders. When combined with Azure Firewall Premium, you gain deep packet inspection and TLS termination at the network boundary.

3. Design Area 2: Identity & Governance (Zero Trust)

API keys and standing “Owner” permissions are the primary risks in the modern cloud. Our blueprint moves to a “Keyless and Standing-Access-Free” model.

Managed Identity and OIDC

For CI/CD, we use Workload Identity Federation (OIDC). GitHub Actions and Azure DevOps runners authenticate to Azure using short-lived tokens, eliminating the need to store secrets in your repository settings.

Privileged Identity Management (PIM)

For human administrators, we enforce PIM-only access. No user holds a permanent “Owner” or “Contributor” role. Instead, they are eligible for the role and must activate it for a 4-8 hour window with a business justification and MFA.

# Example: Grant the pipeline's federated identity Management Group Contributor
# on your intermediate management group, not the tenant root.
az role assignment create \
  --role "Management Group Contributor" \
  --scope "/providers/Microsoft.Management/managementGroups/<intermediate-mg-id>" \
  --assignee "<pipeline-identity-object-id>"

Scope the pipeline to the intermediate management group, not the tenant root. Root-scope assignments need a Global Administrator to elevate access, and anything granted there reaches every subscription in the tenant, including ones your landing zone doesn’t govern yet.

4. The First Call: Terraform or Pulumi?

Every article in this series shows both, because you’ll inherit whichever one your organization already runs. If you get to choose, choose once. Running both for the platform layer doubles your review burden and splits your team’s expertise.

Both keep state, so state doesn’t decide it. Either way you need a hardened backend with private access, versioning and RBAC-only keys.

Pick Terraform when you already run it for other clouds or on-premises, or you want Azure Verified Modules as your starting point. A second toolchain for Azure alone is a cost you’ll pay forever.

Pick Pulumi when your platform engineers already write TypeScript and you want type-checking, unit tests and real loops in your infrastructure code. @pulumi/azure-native is generated from Azure’s REST API specifications, so new resource types and API versions show up without anyone hand-writing them.

What I’d rule out: picking Pulumi and then using all of TypeScript. Class hierarchies and clever abstractions make a preview harder to trace back to the code, and your reviewers pay for that on every PR. Keep platform programs as flat as a Terraform configuration.

5. Building with Azure Verified Modules (AVM)

In 2026, Microsoft retired the monolithic terraform-azurerm-caf-enterprise-scale module in favor of Azure Verified Modules (AVM).

AVM modules are atomic, high-quality, and follow a strict “Contract” for parameters and outputs, so modules from different authors compose without surprises. On Pulumi, you use azure-native resources directly and carry the same defaults (diagnostic settings, locks, private endpoints) in your own component resources.

Terraform AVM Example

module "spoke_vnet" {
  source  = "Azure/avm-res-network-virtualnetwork/azurerm"
  version = "~> 0.7"

  name                = "vnet-prod-app-001"
  resource_group_name = "rg-prod-app-001"
  location            = "eastus"
  address_space       = ["10.1.0.0/16"]

  # AVM includes standard telemetry and diagnostic support
  diagnostic_settings = {
    to_central_law = {
      workspace_resource_id = var.log_analytics_workspace_id # Full Resource ID
    }
  }
}

6. The Deployment Lifecycle (CI/CD)

Deploying a landing zone from a laptop is a single point of failure. A production-grade foundation requires a GitOps workflow.

The Deployment Lifecycle (AVM + CI/CD)

sequenceDiagram participant PR as PR CREATED participant GH as GITHUB ACTIONS participant Azure as AZURE PLATFORM PR->>GH: 1. Lint & Scan GH->>Azure: 2. Plan / Preview GH-->>PR: 3. Post Comment note over PR: 4. HUMAN REVIEW
5. MERGE TO MAIN PR->>GH: Merge note over GH: 6. ENVIRONMENT GATE GH->>Azure: 7. APPLY / PULUMI UP

7. Hands-On: Scaffolding Your Landing Zone

Start with the ALZ Accelerator. This tool generates the repository structure, OIDC identities, and initial GitHub Actions workflows for your specific environment. The defaults are designed for a demo. You’re not running a demo — run through the prompts and override what you need.

# Install the ALZ Accelerator PowerShell module
Install-Module -Name ALZ -Force -Scope CurrentUser

# Scaffolding a new environment for GitHub Actions
New-ALZEnvironment -Path "C:\Source\MyALZ" -DeploymentStrategy "GitHubActions"

Key Takeaways

  1. Defaults are Risks: Azure’s default public endpoints and standing permissions must be explicitly hardened for enterprise use.
  2. Hierarchy is the Foundation: A well-designed Management Group tree is the only way to scale policy and cost controls.
  3. Standardize on AVM: Stop building custom modules for standard Azure resources. Use Microsoft-verified patterns to reduce your maintenance debt.
  4. Shift-Left with PRs: Every change to the platform must be previewed via terraform plan or pulumi preview before a human signs off.

The series, in build order:

  1. Design Your Azure Management Group and Subscription Hierarchy
  2. Hub-and-Spoke Networking for Azure Landing Zones
  3. Identity and Access Architecture for Azure Landing Zones
  4. Governance at Scale: Writing and Deploying Azure Policies
  5. Subscription Vending: Automating New Workload Onboarding
  6. Centralized Monitoring: Log Analytics, Diagnostic Settings, and Workbooks
  7. Security Baseline: Defender for Cloud and Microsoft Sentinel
  8. CI/CD Pipeline for Your Landing Zone
  9. Cost Governance in the Landing Zone
  10. Day-2 Operations: Maintaining and Evolving Your Landing Zone

Start with the management group hierarchy. It’s the hardest thing to change later.

Sources