This chapter covers the consumption-based pricing model, a core concept in cloud computing that distinguishes Azure from traditional on-premises IT. Understanding this model is critical for the AZ-900 exam, as it underpins how you estimate and manage costs in Azure. The 'Describe Cloud Concepts' domain, which includes this objective, carries approximately 25-30% of the exam weight. By the end of this chapter, you will grasp why consumption-based pricing is a key advantage of the cloud and how it influences architectural decisions.
Jump to a section
Imagine you need to get around a city for a month. Option A: buy a car. You pay a large upfront cost, plus insurance, maintenance, and fuel regardless of how much you drive. Even if you leave it parked for a week, you still pay. Option B: use taxis. You pay only when you take a ride, and the fare depends on distance and time. No upfront cost, no ongoing fixed payments. With taxis, you can also scale – use a small car for short trips or a van for moving furniture, paying only for that trip. The consumption-based pricing model in Azure works exactly like taxi rides: you pay only for the compute, storage, or network resources you actually use, measured in units like hours, gigabytes, or operations. There is no upfront commitment, and you can stop paying at any time. Just as a taxi company maintains the fleet and handles fuel, Azure manages the underlying hardware and software. You, the customer, are free from capital expenditure and capacity planning. This model shifts IT spending from a fixed, predictable cost (like a car loan) to a variable, usage-driven cost (like taxi fares). It enables businesses to align costs directly with revenue – if your app gets more users, you use more resources and pay more, but you also earn more. If demand drops, costs drop too. This is the fundamental business shift that Azure's consumption model enables.
What is Consumption-Based Pricing and the Business Problem It Solves
In traditional on-premises IT, organizations purchase hardware (servers, storage, networking) as capital expenses (CapEx). They must forecast their needs years in advance, often over-provisioning to handle peak demand. This leads to wasted capacity and high fixed costs. Cloud computing, and specifically Azure's consumption-based pricing, solves this by allowing you to pay only for what you use, turning IT spending into an operational expense (OpEx). This model is also called 'pay-as-you-go'.
How It Works – The Mechanism Step by Step
Azure meters every resource you provision. For example:
Virtual Machines (VMs) are billed per second (or per hour for some older plans) based on their size and operating system.
Storage is billed per gigabyte stored per month, plus transaction costs.
Outbound data transfer is billed per gigabyte.
At the end of each billing cycle (monthly), Azure aggregates all usage metrics from your subscriptions and generates an invoice. The total is the sum of each resource's usage multiplied by its unit price. You can see real-time cost estimates in the Azure Portal under Cost Management + Billing.
Key Components: Pricing Tiers and Models
Azure offers several pricing tiers within the consumption model: - Pay-as-you-go: No upfront commitment, you pay standard rates. - Reserved Instances: Pre-pay for 1 or 3 years of VM usage and get up to 72% discount compared to pay-as-you-go. This still fits the consumption model because you are committing to a certain amount of usage, but you pay upfront. - Spot VMs: Use unused Azure capacity at deep discounts (up to 90%) but can be evicted at any time. This is pure consumption for non-critical workloads. - Azure Hybrid Benefit: Use your existing on-premises Windows Server or SQL Server licenses with Software Assurance to reduce VM costs.
Comparison to On-Premises Equivalent
On-premises, you buy a server for $10,000 and it has a fixed capacity. Whether you use 10% or 100%, you paid the same. In Azure, you could run a VM that costs $0.10 per hour. If you run it for 100 hours, you pay $10; if you run it for 1,000 hours, you pay $100. The cost scales linearly with usage. Additionally, you can scale up or down instantly. For example, if your e-commerce site sees 10x traffic during holiday season, you can increase VM size or add more VMs, then reduce them after the spike – paying only for the extra capacity when needed.
Azure Portal and CLI Touchpoints
To see consumption-based pricing in action:
- Azure Portal: Go to 'Subscriptions' > select a subscription > 'Cost analysis' to view usage and charges.
- Azure Pricing Calculator: Estimate costs by selecting services and configuring specifications.
- TCO Calculator: Compare on-premises costs with Azure costs.
- Azure CLI: Use az billing commands to retrieve invoices and usage details.
- PowerShell: Get-AzConsumptionUsageDetail to get usage data.
Example CLI command to list usage for a subscription:
az consumption usage list --subscription <subscription-id> --billing-period-name <period>This command returns JSON with quantities and costs. The exam does not test specific CLI commands, but understanding that these tools exist is important.
Concrete Business Scenarios
Scenario 1: Startup with unpredictable traffic. A new social media app starts with 100 users. They deploy a small VM and a database. As the user base grows to 10,000, they scale up. With consumption-based pricing, their costs grow proportionally to revenue. If they had bought servers upfront, they would have wasted money on idle capacity.
Scenario 2: Batch processing. A data analytics company runs big data jobs every night for 2 hours. They use Spot VMs to reduce costs. They pay only for those 2 hours per day. On-premises, they would need a dedicated cluster.
Scenario 3: Dev/Test environments. A development team needs VMs for testing. They shut them down on weekends and evenings. With consumption-based pricing, they save 60% compared to running 24/7. In on-premises, the servers consume electricity and cooling even when idle.
How Azure Meters Usage
Each resource has a meter. For example: - Virtual Machines: Metered per minute of runtime (after the first minute). A stopped VM does not incur compute charges, but attached storage still incurs costs. - Azure Blob Storage: Metered by average daily storage in GB, plus write/read operations per 10,000. - Azure Functions: Metered by number of executions and execution time (in GB-seconds). The first 1 million executions per month are free.
Reservations and Savings Plans
While pay-as-you-go is pure consumption, Azure also offers reservations (1 or 3 years) and savings plans (1 or 3 years) that reduce costs for committed usage. These are still consumption-based because you pay for a certain amount of compute capacity (e.g., $100/month for a VM) regardless of actual usage, but you get a discount. The exam considers these as part of the consumption model because they are optional and you pay only for what you commit to.
Cost Management Tools
Azure provides several tools to monitor and control costs: - Azure Cost Management + Billing: Dashboards, budgets, alerts. - Azure Advisor: Recommends right-sizing VMs or using reserved instances. - Azure Policy: Enforces tagging and limits. - Budgets: Set spending limits and receive alerts when thresholds are exceeded.
Summary of Benefits
No upfront costs: Aligns with operational budgeting.
Elasticity: Scale resources up or down automatically.
Cost variability: Costs match business activity.
No over-provisioning: Only pay for what you use.
Global pricing: Same model across regions, though rates vary.
Exam Tip
On the AZ-900 exam, you will be asked to distinguish between CapEx and OpEx, and to identify which model corresponds to consumption-based pricing. Remember: consumption-based pricing = OpEx. Also, know that reserved instances are a form of consumption-based pricing with a commitment discount.
Choose a resource to provision
In the Azure Portal, you select a service like Virtual Machines, Azure SQL Database, or Blob Storage. Each service has a pricing tier that defines the unit of consumption (e.g., VM size, storage tier). At this stage, you do not pay anything yet – you are just configuring what you want to use.
Deploy and start using the resource
Once deployed, the resource begins consuming Azure capacity. For a VM, this means it starts running and using CPU, memory, and network. Azure immediately begins metering usage. For example, a Standard_D2s_v3 VM running in East US incurs charges per second. Behind the scenes, Azure records every minute of runtime in a usage log.
Azure meters usage continuously
Azure uses meters to track consumption. Each resource has one or more meters. For a VM, there is a compute meter (per second), a storage meter (for managed disks), and a network meter (outbound data). The meters aggregate usage into hourly or daily buckets. You can view real-time meter data in the Azure Portal under 'Metrics'.
View cost estimates and set budgets
Using Cost Management + Billing, you can see your estimated costs before the billing period ends. You can set budgets (e.g., $500 per month) and configure alerts to notify you when spending reaches 80% or 100% of the budget. This is proactive cost control – a key feature of consumption-based pricing.
Receive monthly invoice and analyze costs
At the end of the billing cycle, Azure generates an invoice that lists all usage and charges. You can download a CSV file with detailed line items. Use Cost Analysis to break down costs by resource, tag, or region. This helps identify which services are driving costs and optimize usage.
Scenario 1: E-Commerce Startup with Seasonal Peaks
A startup launches an online store. They expect low traffic for most of the year but massive spikes during Black Friday. On-premises, they would need to buy servers for the peak, leaving them idle 11 months. In Azure, they deploy VMs behind a load balancer. During normal times, they use 2 small VMs. For Black Friday, they enable auto-scaling to add 20 VMs. After the sale, they scale back down. With consumption-based pricing, they pay only for the extra VMs during the few days of high traffic. They also use Azure SQL Database with serverless compute, which pauses when not in use, further reducing costs. The key lesson: consumption-based pricing allows them to match costs exactly to revenue, avoiding wasted capital.
Scenario 2: Media Company Processing Video Files
A media company receives raw video files from clients, transcodes them into multiple formats, and delivers them. Transcoding is CPU-intensive but only runs when files arrive. On-premises, they would need a render farm. In Azure, they use Azure Batch or Azure Functions. Each job triggers a function that runs for a few minutes. With consumption-based pricing, they pay only for the compute time used during transcoding. They also store the original and transcoded files in Blob Storage, paying per GB per month. If a client uploads a 10 GB video, they pay for storage and compute for that job. If no clients upload anything for a week, they pay nothing for compute. This is ideal for variable workloads.
Common Mistakes and Pitfalls
Forgetting to stop resources: A developer leaves a test VM running over the weekend, incurring charges for 48 hours of idle compute. Always shut down VMs when not in use.
Ignoring data egress costs: Moving large amounts of data out of Azure can be expensive. For example, downloading 1 TB of data costs around $87 (depending on region). Many new users underestimate these costs.
Not using reserved instances for steady workloads: A company runs a production database 24/7 for 3 years. Pay-as-you-go costs $10,000/year. A 3-year reserved instance costs $6,000/year (40% discount). By not reserving, they waste $12,000 over 3 years.
AZ-900 Objective: 1.4 – Describe the consumption-based pricing model
This objective falls under 'Cloud Concepts' and is tested in roughly 3-5 questions. The exam expects you to:
Define consumption-based pricing (pay-as-you-go).
Contrast it with CapEx (on-premises) and OpEx (cloud).
Identify scenarios where consumption-based pricing is beneficial.
Understand that reserved instances and spot VMs are variations.
Common Wrong Answers and Why Candidates Choose Them
'Consumption-based pricing means you pay a fixed monthly fee.' – This is wrong because fixed fees are subscription-based, not consumption-based. Candidates confuse SaaS subscription (like Office 365) with IaaS pay-as-you-go.
'You pay for allocated capacity, not actual usage.' – This describes reserved instances (committed capacity) but not pure consumption. For pay-as-you-go, you pay for actual usage.
'Consumption-based pricing is only for compute resources.' – It applies to all services: storage, networking, databases, etc.
'You can never predict costs with consumption-based pricing.' – Azure provides cost management tools and budgets to forecast and control costs.
Specific Terms and Values That Appear on the Exam
Pay-as-you-go: The default consumption model.
Reserved Instances: Pre-pay for 1 or 3 years.
Spot VMs: Discounted, but can be evicted.
Azure Hybrid Benefit: Use on-prem licenses to reduce costs.
OpEx: Operational expenditure – ongoing costs.
CapEx: Capital expenditure – upfront hardware costs.
Edge Cases and Tricky Distinctions
Reserved Instances vs. Pay-as-you-go: Both are consumption-based, but RIs require a commitment. The exam might ask: 'Which model requires no upfront commitment?' Answer: Pay-as-you-go.
Serverless services: Azure Functions and Logic Apps are pure consumption – you pay per execution. The exam may test that serverless is a subset of consumption-based pricing.
Free tier: Some services offer free tiers up to a certain limit (e.g., 1 million function executions). Beyond that, consumption pricing applies. The exam may ask about the free tier as an example of consumption.
Memory Trick: 'Pay as you go, grow as you go'
When you see 'consumption-based', think 'no upfront, pay for what you use'. For CapEx, think 'buy it all upfront'. Use this decision tree:
Does the customer pay only for actual usage? → Consumption-based.
Does the customer pay a fixed fee regardless of usage? → Not consumption (e.g., reserved instance with full commitment, or SaaS subscription).
Consumption-based pricing means you pay only for the resources you use, with no upfront costs.
It is also called pay-as-you-go and is an operational expense (OpEx).
All Azure services (compute, storage, networking, databases) support consumption-based pricing.
Reserved Instances and Spot VMs are variations that offer discounts for commitment or risk.
Azure provides Cost Management + Billing to monitor and control consumption costs.
Data egress (outbound data transfer) is a significant cost that must be considered.
Stopping a VM deallocates it and stops compute charges, but storage charges continue until the disk is deleted.
These come up on the exam all the time. Here's how to tell them apart.
Pay-as-you-go
No upfront commitment
Highest per-unit cost
Billed per second/minute
Ideal for short-term or variable workloads
Can stop anytime without penalty
Reserved Instances
Requires 1 or 3 year commitment
Up to 72% discount
Upfront or monthly payment
Best for steady-state workloads
Early termination incurs fees
Mistake
Consumption-based pricing means you never have to pay upfront for anything.
Correct
While pay-as-you-go has no upfront cost, reserved instances require a one-time or monthly upfront payment for the commitment period. However, the ongoing cost is still based on consumption of the reserved capacity.
Mistake
You can only use consumption-based pricing for virtual machines.
Correct
All Azure services support consumption-based pricing. For example, Azure SQL Database charges per DTU or vCore per hour, and Blob Storage charges per GB stored per month.
Mistake
Consumption-based pricing is always cheaper than on-premises.
Correct
For steady, predictable workloads, reserved instances or on-premises can be cheaper. Consumption-based pricing is cost-effective for variable or unpredictable workloads.
Mistake
Once you stop using a resource, you stop paying immediately.
Correct
For VMs, you stop paying compute charges when deallocated, but attached storage (disks) still incurs costs. You must delete the resource entirely to stop all charges.
Mistake
Consumption-based pricing means you cannot budget or predict costs.
Correct
Azure provides Cost Management, budgets, and alerts to forecast and control costs. You can also use pricing calculators to estimate costs before deployment.
Consumption-based pricing is a model where you pay only for the Azure resources you actually use, measured in units like hours, gigabytes, or operations. It requires no upfront commitment, making it an operational expense (OpEx). For example, a virtual machine running for 100 hours costs exactly 100 times the per-hour rate. This model is ideal for variable workloads because costs scale with usage.
Pay-as-you-go is pure consumption pricing with no commitment – you pay standard rates per unit of usage. Reserved instances require you to commit to using a specific VM size in a region for 1 or 3 years, in exchange for a significant discount (up to 72%). Both are consumption-based, but reserved instances involve a commitment. Use pay-as-you-go for short-term or unpredictable workloads; use reserved instances for steady-state workloads.
When you stop (deallocate) a VM, you stop incurring compute charges, but you still pay for the attached managed disks and any static public IP addresses. To stop all charges, you must delete the VM and its associated resources. Always check that you have deleted or stopped resources you no longer need.
The Azure free tier offers limited quantities of certain services for 12 months (e.g., 750 hours of B1s VM, 5 GB of Blob Storage). Beyond these limits, consumption-based pricing applies. The free tier is a promotional offering, but it uses the same consumption meters – you just get a credit applied. It's a great way to learn without cost, but remember that exceeding free limits incurs charges.
Consumption-based pricing makes budgeting more challenging because costs vary with usage. However, Azure provides tools like Cost Management, budgets, and alerts to help you forecast and control spending. You can set a budget of $500 per month and receive an email when you hit 80% or 100%. For predictable workloads, reserved instances can stabilize costs.
Serverless computing (e.g., Azure Functions) is a type of consumption-based pricing where you pay per execution and per execution time, with no provisioned capacity. All serverless services use consumption pricing, but not all consumption-priced services are serverless (e.g., VMs are not serverless). Serverless is a subset.
Spot VMs use unused Azure capacity at a deep discount (up to 90%). They are pure consumption – you pay only for the time the VM runs. However, Azure can evict them at any time with 30 seconds notice. They are ideal for batch processing, dev/test, or stateless workloads that can handle interruptions. They are a form of consumption-based pricing with lower cost but higher risk.
You've just covered Consumption-Based Pricing Model — now see how well it sticks with free AZ-900 practice questions. Full explanations included, no account needed.
Done with this chapter?