How to Estimate Azure Costs

Why Cost Estimation Matters

Cloud billing is consumption-based, which means costs appear after resources have already run. Without an upfront estimate, teams often discover that a workload costs three times what they expected — after the bill arrives. Good estimation habits prevent this by modelling costs before deployment and creating a baseline to compare actual spend against.

Estimation also shapes architecture decisions. Knowing that Azure SQL Database costs $0.724/hour on a General Purpose 4 vCore instance, while Azure SQL Serverless can scale down to $0.145/hour minimum during idle periods, may change your database tier choice entirely.

The Azure Pricing Calculator

The Azure Pricing Calculator at azure.microsoft.com/en-us/pricing/calculator/ is a browser-based tool that lets you build up a cost estimate by adding services and configuring their parameters.

Building an estimate step by step

1. Navigate to the Pricing Calculator and click + Add to select a service category.
2. Search for the service you want (e.g., “Virtual Machines”) and click Add to estimate.
3. Configure the parameters for that service: region, tier, instance size, operating system, hours per month, and any optional features like managed disks or reserved pricing.
4. Repeat for every service in your architecture: databases, storage accounts, networking, monitoring, and so on.
5. Review the monthly and annual totals at the bottom of the page.
6. Click Save to store the estimate in your account, or Export to download it as a CSV or PDF.

Example: Three-tier web application estimate

Consider a standard three-tier web application with two web servers, one application server, and one database. A realistic estimate for a production-grade East US deployment might look like this:

CLI-Based Cost Estimation

The Azure CLI provides commands for querying retail pricing data, which lets you build cost estimates programmatically or check current prices without opening a browser.

Query the Azure Retail Prices API

Azure exposes a public REST API for pricing data that does not require authentication. The Azure CLI can query it directly:

# Get current PAYG price for Standard_D4s_v5 in East US (Linux)
az rest \
  --method GET \
  --url "https://prices.azure.com/api/retail/prices?\$filter=armRegionName eq 'eastus' and skuName eq 'D4s v5' and serviceName eq 'Virtual Machines' and priceType eq 'Consumption' and contains(productName, 'Linux')" \
  --query "Items[].{sku:skuName, price:retailPrice, unit:unitOfMeasure}" \
  --output table

# Get prices for Azure SQL Database General Purpose
az rest \
  --method GET \
  --url "https://prices.azure.com/api/retail/prices?\$filter=armRegionName eq 'eastus' and serviceName eq 'SQL Database' and skuName eq 'General Purpose 4 vCores'" \
  --query "Items[].{sku:skuName, price:retailPrice, unit:unitOfMeasure}" \
  --output table

Estimate costs for existing resources

Once resources are deployed, you can query actual usage and project monthly costs:

# Get usage for the current billing period
az consumption usage list \
  --billing-period-name $(az billing period list --query "[0].name" -o tsv) \
  --query "[].{Name:instanceName, Cost:pretaxCost, Currency:currency}" \
  --output table

# List all VMs with their sizes to estimate RI savings
az vm list \
  --query "[].{Name:name, Size:hardwareProfile.vmSize, Location:location, RG:resourceGroup}" \
  --output table

Estimating Costs from Bicep Templates

The az deployment what-if command shows what resources a Bicep template will create, which you can cross-reference with pricing data. While it does not calculate costs directly, it helps you build a manifest of resources to estimate.

// Example: document expected costs with Bicep metadata comments
// D4s_v5 Linux VM in East US
// PAYG: ~$140/month | 1-yr RI: ~$85/month | 3-yr RI: ~$55/month
resource appVm 'Microsoft.Compute/virtualMachines@2024-03-01' = {
  name: 'app-vm-prod'
  location: 'eastus'
  properties: {
    hardwareProfile: {
      vmSize: 'Standard_D4s_v5'
    }
    // ... other properties
  }
}

# Preview what a template will deploy before running it
az deployment group what-if \
  --resource-group my-rg \
  --template-file main.bicep \
  --parameters environment=prod

# After deployment, check actual costs in Cost Management
az costmanagement query \
  --type Usage \
  --timeframe MonthToDate \
  --dataset-granularity Daily \
  --dataset-aggregation "{\"totalCost\":{\"name\":\"Cost\",\"function\":\"Sum\"}}" \
  --scope "/subscriptions/$(az account show --query id -o tsv)"

TCO Calculator for Migration Projects

The Total Cost of Ownership Calculator is specifically designed for teams migrating from on-premises to Azure. It accounts for server hardware depreciation, data centre costs, networking hardware, power and cooling, IT staffing, and software licences — then compares the full cost against equivalent Azure services.

To use the TCO Calculator:

1. Navigate to azure.microsoft.com/en-us/pricing/tco/calculator/
2. Define your current on-premises servers: server count, CPU cores, RAM, storage, and whether they run Windows or Linux
3. Define your databases: engine type (SQL Server, Oracle, MySQL), version, and data volume
4. Set your local assumptions: electricity costs per kWh, IT labour rate, and hardware amortisation period
5. Review the 5-year comparison report

Most mid-sized organisations find that Azure is 20–40% cheaper over 5 years when accounting for capital expenditure, staffing, and maintenance costs — even before accounting for cloud-native features like auto-scaling and managed services.

Tips for More Accurate Estimates

• Add a 15–20% buffer: Initial estimates typically miss minor services like Azure Monitor, Key Vault, and DNS zones. These individually cost very little but add up across a large environment.
• Model your actual usage hours: A development VM that runs 8 hours/day, 5 days/week uses only 160 hours per month — not the 730-hour full-month figure. Adjust accordingly.
• Include data egress in every estimate: Egress costs are invisible until the bill arrives. Estimate your outbound data transfer and add it explicitly.
• Check regional price differences: Prices vary by region. East US is typically cheaper than West Europe or Australia East. If your workload has no regional data residency requirements, compare prices across regions.
• Revisit estimates after 30 days: Compare your first real bill against your estimate and identify which line items were significantly under or over-estimated. Refine your estimation model for future projects.