/azure-pricing

Cost Estimation Formulas

Reference formulas for converting Azure unit prices into monthly and annual estimates when using the azure-pricing skill.

Standard Time-Based Multipliers

Azure billing uses 730 hours/month as the standard period (365 days × 24 hours ÷ 12 months).

Reservation prices are lump-sum totals, not hourly rates. Despite unitOfMeasure: "1 Hour" in the API response, Reservation retailPrice values represent the total commitment cost for the full term. Divide by 8,760 (1-year) or 26,280 (3-year) to get a comparable hourly figure.

Service-Specific Formulas

Virtual Machines

Monthly (always-on)      = hourly_price × 730
Monthly (business hours) = hourly_price × 176   # 8h/day, 22 working days

Linux and Windows are separate SKUs with different prices.
Always query both if OS hasn't been confirmed.

⚠️ Azure Hybrid Benefit: Retail prices reflect full PAYG rates. AHB can reduce Windows VM and SQL Server costs by 40%+. Always flag this for Windows or SQL workloads.

Spot pricing (interrupt-tolerant workloads only):

Query via filter: "contains(meterName, 'Spot') and armSkuName eq '<sku>' and armRegionName eq '<region>'" with price-type: Consumption.

Live reference (uksouth, Standard_D4s_v5, tested March 2026):

Spot is not suitable for databases, persistent workloads, or anything that can't tolerate interruption with ~30 seconds notice.

Azure App Service

Monthly = plan_price × 730   # all plans priced hourly

All App Service plans (Basic, Standard, Premium v3, Isolated) return unitOfMeasure: "1 Hour" in the API. Multiply the hourly retailPrice by 730 to get monthly cost for an always-on plan.

SKU name spacing: P2v3 returns no results — the API stores it as P2 v3. If sku returns empty, use filter: "skuName eq 'P2 v3' and serviceName eq 'Azure App Service'".

Azure SQL Database

Two separate queries required — compute and storage are distinct meters.

Monthly compute = hourly_compute_price × 730
Monthly storage = price_per_GB_month × storage_GB
Monthly total   = Monthly compute + Monthly storage

SKU format mismatch: The ARM SKU GP_Gen5_4 maps to skuName: "4 vCore" in the API. Filter by both skuName and productName (containing General Purpose - Compute Gen5) to avoid mixing General Purpose, Business Critical, and DC-Series rows.

Azure Functions

Three-component cost — always ask about invocation volume before estimating.

Execution cost = price_per_execution × invocations_per_month
Compute cost   = price_per_GBs × (memory_GB × duration_seconds × invocations)
Total monthly  = Execution cost + Compute cost

Free grant (Consumption plan): 1 million executions and 400,000 GB-s per month are free. For low-volume functions, the actual cost may be £0.

Azure Blob Storage

Three separate meters — a single query won't produce a single combined figure.

Storage cost     = price_per_GB × stored_GB
Transaction cost = price_per_10k_ops × (operations ÷ 10,000)
Egress cost      = price_per_GB × egress_GB
Monthly total    = Storage + Transactions + Egress

Ask about access pattern (Hot/Cool/Cold/Archive tier) — prices differ significantly between tiers.

Azure Cosmos DB

Provisioned throughput:

Monthly = (RU_per_second ÷ 100) × price_per_100_RUs × 730

Serverless:

Monthly = (total_RUs_consumed ÷ 1,000,000) × price_per_million_RUs

Ask the user which model they're using and, for serverless, their expected RU consumption per month before estimating.

Azure Container Apps

Cannot query via service parameter — use OData filter: serviceName eq 'Azure Container Apps' and armRegionName eq '<region>'. Pricing has three separate billing dimensions depending on plan type:

Consumption plan (scale-to-zero):

vCPU cost    = vCPU_price_per_second × vCPU_count × active_seconds_per_month
Memory cost  = memory_price_per_GiB_second × memory_GiB × active_seconds_per_month
Request cost = request_price_per_million × (requests_per_month ÷ 1,000,000)
Monthly total = vCPU + Memory + Request costs

Free grant on Consumption plan: 180,000 vCPU-seconds and 360,000 GiB-seconds per month. Low-traffic apps may have near-zero cost.

Dedicated plan (always-on workload profiles):

Monthly = (dedicated_vCPU_price × vCPU_count × 730)
        + (dedicated_memory_price × memory_GiB × 730)
        + dedicated_plan_management_price × 730

Ask about: minimum/maximum replicas, vCPU + memory per replica, and request volume before estimating Consumption plan costs.

Azure Kubernetes Service (AKS)

Monthly = node_VM_price × 730 × node_count

The AKS control plane is free on the Standard tier. Factor in node pool VMs only. If autoscaling is in use, estimate against average node count rather than maximum.

Reservation Savings Formula

Savings % = ((PAYG_price - Reserved_price) ÷ PAYG_price) × 100

Always present all pricing models side by side when Reservation or Savings Plan pricing is requested:

Pre-Estimation Clarifying Questions

Before querying for consumption-based services, ask:

For always-on, fixed-tier services (VMs, App Service Premium, SQL Database provisioned), no clarifying questions are needed — proceed with the pricing query.

Known MCP Tool Limitation: AI Services

The Azure MCP pricing tool returns a 500 error for any service in the AI + Machine Learning service family. This includes:

• Foundry Models (Azure OpenAI)
• Azure OpenAI (legacy service name)
• service-family: "AI + Machine Learning"
• Any OData filter that returns results from that family

Workaround: Direct users to:

• Azure OpenAI Pricing
• Azure Pricing Calculator

Token-based pricing formula (for reference once rates are obtained manually):

Monthly cost = (input_tokens ÷ 1,000) × input_price_per_1k
             + (output_tokens ÷ 1,000) × output_price_per_1k

Ask for expected prompt and completion token volumes per month before estimating.