Azure Cloud Billing Wiki

Azure Pricing Models

Understanding Azure's Pricing Options

Azure offers multiple pricing models to accommodate different usage patterns and commitment levels. For MSPs, understanding these models is critical to maximizing client value and partner margins.

Pay-As-You-Go (PAYG)

The default Azure pricing model where you pay only for what you consume.

Characteristics

No Upfront Commitment

  • Start and stop resources at any time
  • Pay per hour (VMs) or per transaction (storage, functions)
  • Ideal for variable or unpredictable workloads

Maximum Flexibility

  • Scale resources up or down instantly
  • No penalty for stopping or deleting resources
  • Perfect for dev/test environments

Highest Unit Cost

  • Most expensive per-hour rate for VMs and other services
  • Premium for flexibility
  • Can become costly for always-on production workloads

When to Use PAYG

  • Development and testing environments
  • Short-term projects or proof-of-concepts
  • Workloads with unpredictable demand
  • Resources that run <40% of the time
  • When cash flow doesn't permit upfront commitments

MSP Pricing Strategy

  • Mark up PAYG costs by 15-30%
  • Educate clients on optimization opportunities
  • Transition stable workloads to Reserved Instances over time

Reserved Instances (RIs)

Reserved Instances provide significant discounts in exchange for 1-year or 3-year commitments.

Discount Levels

1-Year Commitment

  • 20-40% savings vs. PAYG
  • Lower upfront capital requirement
  • Good for established workloads with uncertain long-term needs

3-Year Commitment

  • 40-72% savings vs. PAYG
  • Maximum cost efficiency
  • Best for stable, long-term production workloads

What Can Be Reserved

Virtual Machines

  • Most common RI purchase
  • Commit to specific VM series, region, and size
  • Apply to running VMs automatically

SQL Database

  • Save on vCore-based databases
  • Commit to compute capacity (vCores)
  • Storage charged separately at PAYG rates

Cosmos DB

  • Reserve throughput capacity (RU/s)
  • Significant savings for high-throughput databases

Storage (Reserved Capacity)

  • Available for Blob storage and Data Lake Storage
  • Commit to capacity in TBs
  • 1-year or 3-year terms

Azure Synapse Analytics, App Service, and More

  • Expanding list of services supporting reservations

Payment Options

All Upfront

  • Entire commitment paid at purchase
  • Typically highest discount
  • Largest immediate cash outlay

Monthly Payments

  • Spread payments over 1 or 3 years
  • No interest charges
  • Smaller initial capital requirement
  • Slightly lower discount than all-upfront

RI Flexibility Features

Instance Size Flexibility

  • Single RI can cover multiple VM sizes within same series
  • Example: 1x D16 RI can cover 16x D1 VMs or 2x D8 VMs
  • Automatic application requires same series and region

Scope Options

  • Single Subscription: RI applies only to one subscription
  • Shared: RI applies across all subscriptions in a billing account
  • Resource Group: RI applies to specific resource group (management group scope)

Reservation Exchanges

  • Exchange RIs for different VM series, regions, or terms
  • No penalty for exchanges
  • Useful when workload requirements change

Reservation Cancellation

  • Refund unused portion with 12% early termination fee
  • Annual refund limit of $50,000 per billing profile
  • Rare but provides safety net

MSP RI Strategy

Analyze Before Purchasing

  • Review 30-60 days of usage data
  • Identify steady-state workloads running 24/7
  • Use Azure Advisor RI recommendations as starting point

Start Conservatively

  • Purchase RIs to cover 60-80% of baseline usage
  • Leave room for PAYG growth and flexibility
  • Expand RI coverage as usage stabilizes

Client Communication

  • Explain savings potential clearly
  • Offer to manage RI purchases on client's behalf
  • Pass through savings or retain portion as management fee

Multi-Client Considerations

  • CSP partners can purchase shared RIs across customer tenants
  • Requires careful planning and cost allocation
  • Consider per-customer RIs for simpler accounting

Savings Plans

Azure Savings Plans offer flexibility similar to AWS Compute Savings Plans.

How Savings Plans Work

Hourly Commitment

  • Commit to spending $X/hour on compute for 1 or 3 years
  • Discount applies automatically to any compute usage up to commitment
  • Excess usage charged at PAYG rates

Flexibility Advantages Over RIs

  • Not locked to specific VM series or regions
  • Automatically applies to VMs, App Services, Azure Functions, and more
  • Change workload types without exchanging reservations

Discount Levels

Similar to RIs

  • 1-year: 15-30% savings
  • 3-year: 30-60% savings
  • Exact discounts vary by usage type

When to Use Savings Plans vs. RIs

Choose Savings Plans When:

  • Managing diverse workloads across regions
  • Expect to change VM series or regions
  • Want ultimate flexibility with discount commitment
  • Running mix of VMs, containers, and serverless

Choose Reserved Instances When:

  • Workload is stable and predictable
  • Using specific VM series in one or two regions
  • Want maximum discount percentage
  • Simpler cost allocation per client

MSP Consideration

Savings Plans require sophisticated tracking to allocate costs across clients accurately. RIs per subscription may be simpler for MSPs with many small clients.

Spot Virtual Machines

Azure Spot VMs provide access to unused Azure capacity at steep discounts.

Characteristics

Up to 90% Discount

  • Dramatic savings vs. PAYG
  • Price fluctuates based on available capacity
  • Can be evicted with 30-second notice when Azure needs capacity

Eviction Policies

  • Azure gives 30-second warning via metadata service
  • VM stops (deallocated) or deletes
  • Can set max price to limit exposure

Ideal Use Cases

Batch Processing Jobs

  • Data analysis, rendering, transcoding
  • Jobs that can pause and resume
  • No time-critical deadlines

Stateless Applications

  • Web apps with external state storage
  • Load-balanced applications that tolerate node failure

Dev/Test Environments

  • Non-critical development servers
  • Automated testing infrastructure

When NOT to Use Spot VMs

  • Production databases or state-sensitive apps
  • Applications requiring guaranteed uptime
  • Workloads without checkpoint/restart capability

MSP Caution

Clearly document Spot VM usage with clients. Evictions can surprise clients unfamiliar with the model. Use only for appropriate workloads.

Azure Hybrid Benefit

Not a pricing model per se, but a critical cost-saving mechanism for clients with existing licenses.

How It Works

Windows Server VMs

  • Use existing Windows Server licenses with Software Assurance
  • Reduces Windows licensing cost from Azure VM billing
  • Saves ~40% on Windows VM costs

SQL Server VMs

  • Use existing SQL Server licenses with Software Assurance
  • Significantly reduces database VM costs
  • Applies to SQL Server 2019 and earlier

SQL Database (PaaS)

  • Apply SQL Server licenses to Azure SQL Database (vCore model)
  • Save up to 55% with combined AHB + Reserved Instances

Requirements

  • Current Software Assurance or subscription licenses
  • Eligible Windows Server Datacenter or SQL Server Enterprise/Standard licenses
  • Compliance with Microsoft licensing terms

MSP Opportunity

License Assessment Services

  • Help clients inventory existing licenses
  • Identify Azure Hybrid Benefit opportunities
  • Ensure license compliance

Ongoing Management

  • Enable AHB in Azure portal for eligible VMs
  • Track license utilization
  • Report savings achieved

Many clients don't realize they can leverage existing licenses in Azure. Position this as a value-add service.

Pricing Model Decision Matrix

Workload Type Best Pricing Model Rationale
Production DB (24/7) RI (3-year) Maximum savings, stable usage
Web app tier (24/7) RI or Savings Plan Consistent usage pattern
Dev/Test (8hr/day) PAYG with auto-shutdown Low utilization %
Batch processing Spot VMs Eviction-tolerant, major savings
Variable traffic app PAYG + RI baseline RI for baseline, PAYG for bursts
Mixed compute needs Savings Plan Flexibility across services

Combining Pricing Models

Smart MSPs layer multiple models:

Example Architecture

  • 10 VMs: 8 covered by RIs (baseline), 2 PAYG (burst capacity)
  • Storage: Reserved Capacity for primary data, PAYG for temp storage
  • Batch jobs: All Spot VMs
  • Estimated savings: 50-60% vs. all-PAYG

Monitoring and Optimization

Regular Reviews

  • Quarterly: Assess RI utilization and coverage
  • Monthly: Review Spot VM eviction rates
  • Continuously: Monitor Azure Advisor for RI recommendations

Key Metrics

  • RI utilization percentage (target: >85%)
  • Savings achieved vs. all-PAYG scenario
  • PAYG spending as percentage of total compute

Next Steps

  • Review Azure Cost Management Tools to monitor RI utilization
  • Implement Azure Cost Optimization for MSPs strategies
  • Learn CSP Program Overview for partner-specific considerations

Mastering Azure's pricing models is foundational to MSP profitability. The right mix can double your effective margins.