How to estimate Azure costs from Terraform

Azure cost estimation from Terraform follows the same workflow as AWS and GCP but has a few Azure-specific patterns worth knowing. Reserved Instances work differently than AWS RIs. Premium SSD disks use tier-based pricing instead of pure per-GB. App Service plans bill differently from VMs. This post walks through the workflow end-to-end with Azure-specific examples.

Step 1: Install and verify

brew install c3xdev/tap/c3x

c3x --version

Step 2: Estimate an Azure project

Point c3x at your Terraform directory:

c3x estimate --path /path/to/terraform

Sample output for a small Azure stack:

 Name                                                  Monthly Qty  Unit         Monthly Cost

 azurerm_linux_virtual_machine.api
 ├─ Compute (Standard_D2s_v5, Linux, pay-as-you-go)             730  hours          $70.08
 └─ OS disk (Premium SSD P10)                                                       $19.71

 azurerm_postgresql_flexible_server.main
 ├─ Compute (GP_Standard_D2s_v3, Zone-redundant HA)             730  hours         $394.20
 └─ Storage (general-purpose SSD)                               200  GB              $23.00

 azurerm_storage_account.assets
 ├─ Blob storage (Hot, LRS)                                     500  GB               $9.50
 └─ Read operations (estimated)                                                       $0.10

 OVERALL TOTAL                                                                       $516.59

Notice three Azure-specific things in the output: the disk shows as P10 (Premium SSD tier 10), not "128 GB at $0.10/GB"; the Postgres compute charge reflects the Zone-Redundant HA premium (roughly 2x base price); the storage account shows blob storage in the Hot tier with LRS replication.

For per-resource pricing details, see the catalog pages for azurerm_linux_virtual_machine, azurerm_postgresql_flexible_server, and azurerm_storage_account.

Step 3: Azure-specific patterns

Reserved Instances

Azure Reserved Instances work at the subscription level. You commit to a VM size (Standard_D2s_v5, for example) for 1 or 3 years and Azure applies the discount to any matching VM in the scope. Configure in c3x-usage.yml:

# c3x-usage.yml
resource_usage:
  azurerm_linux_virtual_machine.api:
    purchaseOption: "reserved"
    purchaseOptionTerm: "1yr"

c3x applies the Reserved Instance rate from the Retail Prices API. Typical discount is 30-50% for 1-year commitments, up to 65% for 3-year all-upfront.

Azure Hybrid Benefit

If you have eligible Windows Server or SQL Server licenses with Software Assurance, you can bring your own license to Azure and save on the per-hour licensing premium. Add to the resource:

resource "azurerm_windows_virtual_machine" "app" {
  name          = "app-vm"
  license_type  = "Windows_Server"  # AHB enabled
  size          = "Standard_D2s_v5"
  # ...
}

c3x detects license_type and applies the base Linux rate (minus the Windows premium). The savings are substantial: a Windows VM typically costs 40-50% more than the Linux equivalent. AHB closes most of that gap.

Spot VMs

For fault-tolerant workloads, Spot VMs cost 60-90% less:

resource "azurerm_linux_virtual_machine" "batch_worker" {
  name              = "batch-worker"
  size              = "Standard_D2s_v5"
  priority          = "Spot"
  eviction_policy   = "Deallocate"
  max_bid_price     = -1   # use current Spot price
  # ...
}

c3x reads priority and applies the Spot pricing tier. Same principles as AWS Spot — see our Spot instances post for the design patterns.

Premium SSD tier mapping

Azure Premium SSDs are billed per tier (P1 through P80), not per GB. Each tier has fixed IOPS, throughput, and a fixed monthly price. c3x reads the disk size and maps it to the smallest tier that fits.

Important pricing quirk: a 130 GB disk maps to P15 (256 GB tier) because it crosses the 128 GB threshold. You pay the P15 price even if you use less storage. Right-sizing means staying just under a tier boundary, not just provisioning what you need.

See the azurerm_managed_disk catalog page for tier-by-tier pricing.

Step 4: Diff branches

c3x diff --path . --compare-to main

Sample output for a PR that resizes a VM and adds a new managed disk:

 ~ azurerm_linux_virtual_machine.api
   └─ Compute (Standard_D2s_v5 -> Standard_D4s_v5)               +$70.08 ($70.08 -> $140.16)

 + azurerm_managed_disk.data
   └─ Premium SSD P20 (512 GB tier)                              +$73.73/month

 Monthly cost change: +$143.81 ($516.59 -> $660.40)

Step 5: CI integration

The setup-c3x GitHub Action works for Azure Terraform identically to AWS:

# .github/workflows/cost.yml
on: pull_request

jobs:
  cost:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0
      - uses: c3xdev/setup-c3x@v1
        with:
          path: .

For full CI integration including Azure DevOps Pipelines, see the CI/CD documentation. For budget gating, see budget guardrails in CI.

Azure-specific cost dimensions to watch

Application Insights ingestion

Bills via Log Analytics at $2.30/GiB. For high-traffic apps with 100% telemetry capture, this can be massive. Set sampling_percentage on the resource (10-20% is usually sufficient). See the azurerm_application_insights catalog page.

Log Analytics for AKS

Container Insights on AKS sends logs to Log Analytics. A standard AKS cluster generates 5-20 GiB/day. At $2.30/GiB ingest, that's $345-1,380/month per cluster. See the azurerm_log_analytics_workspace catalog page for mitigations.

Bandwidth and ExpressRoute

Azure outbound bandwidth costs $0.087/GB after the first 100 GB free per month. ExpressRoute Direct ports add fixed per-hour fees. Cross-region traffic within Azure also has separate rates.

Premium tier Cosmos DB and Service Bus

Cosmos DB Premium and Service Bus Premium have significantly higher per-RU and per-hour rates than Standard. Make sure the features (e.g., VNet integration, longer message retention) actually justify the cost.

A real Azure cost comparison

Typical small production stack:

• 2× azurerm_linux_virtual_machine (Standard_D2s_v5): $140/month
• 1× azurerm_postgresql_flexible_server (D2s_v3 HA): $394/month
• 1× azurerm_storage_account (500 GB Hot LRS): $10/month
• 1× azurerm_application_insights (sampling at 10%): $50/month
• 1× azurerm_redis_cache (C1 Standard): $43/month
• Total: ~$637/month

With 1-year Reserved Instances on compute and 1-year Reserved Capacity on Cosmos: ~$420/month (34% savings).

FAQ

Where to go from here

• Full resource catalog with Azure-specific pricing dimensions for every supported resource
• For the AWS equivalent workflow: how to estimate AWS costs from Terraform
• For GCP: how to estimate GCP costs from Terraform
• For PR-level cost gating: budget guardrails in CI