Blog Azure Cloud Costs

Azure VM Pricing: What They Don't Tell You (But You Need To Know)

Azure virtual machines (VMs) are computing services that enable users to host their Windows and Linux-based systems or applications in the cloud.

Virtual machines (VMs) can be costly in Microsoft Azure, and Azure VM pricing is complex: How much do they cost, and can you really get a VM for free? And how do I reduce VM costs in Azure? 

This article outlines all the ins and outs of Azure VM prices and shows you some examples and tips to reduce unnecessary spending in Azure.

Niels Kroeze

Author

Niels Kroeze IT Business Copywriter

Reading time 18 minutes Published: 17 April 2025

In this guide, we’ll break down the following: 

 

What is Azure VM Pricing? 

Azure Virtual Machine pricing refers to the cost of running VMs on Microsoft Azure. You choose from different VM types, series, sizes, and configurations based on your compute needs. 

Azure VM pricing is calculated per second of usage and based on several factors including the VM instance type (SKU) and the different pricing models (pay-as-you-go, reserved instance, spot VM, etc.). 

 

Azure VM Types: Which are the cheapest and most expensive? 

There’s plenty of choice when running your apps and workloads in Microsoft Azure. Microsoft has a vast offering of hundreds of VMs, providing you with scalability and performance across either your Linux or Windows Server-based apps. 

Their VMs are grouped into so-called “families” and “series” based on different use cases. The A-series are the cheapest option, and the HB family is typically the most expensive VM. Azure has distinct VM families optimised for every kind of workload, where each family has its speciality. This includes compute, memory, and storage-intensive workloads in addition to AI, machine learning, and mission-critical scenarios. 

In total, Azure VM series consists of 7 main categories (or call them families), shown in the table below: 

VM family Description Family Series
General Purpose General Purpose VMs offer a balanced CPU-to-memory ratio. They are best suited for development, testing, small to medium databases, and low to moderate web traffic. A-family Av2-series, Previous-gen
B-family Bsv2, Basv2, Bpsv2, Previous-gen
D-family Dpsv6, Dplsv6, Dpdsv6, Dpldsv6, Dasv6, Dadsv6, Dalsv6, Daldsv6, Dpsv5, Dpdsv5, Dpldsv5, Dlsv5, Dldsv5, Dv5, Dsv5, Ddv5, Ddsv5, Dasv5, Dadsv5, Previous-gen
DC-family DCasv5, DCadsv5, DCas_cc_v5, DCads_cc_v5, DCesv5, DCedsv5, DCsv3, DCdsv3, Previous-gen
Compute Optimised Compute-optimised VMs offer a high CPU-to-memory ratio. They’re a good fit for medium-traffic web servers, batch processing, application servers, and network appliances. F-family Fasv6, Falsv6, Famsv6, Fsv2, Previous-gen
FX-family FX
Memory Optimised Offer a high memory-to-CPU ratio. They’re ideal for relational databases, data analytics and large in-memory workloads. E-family Epsv6, Epdsv6, Easv6, Eadsv6, Ev5, Esv5, Edv5, Edsv5, Easv5, Eadsv5, Epsv5, Epdsv5, Previous-gen
Eb-family Ebdsv5, Ebsv5
EC-family ECasv5, ECadsv5, ECas_cc_v5, ECads_cc_v5, ECesv5, ECedsv5
M-family Mbsv3, Mbdsv3, Msv3, Mdsv3, Mv2, Msv2, Mdsv2
Storage Optimised Offer high throughput and low latency. They are great for running big data, SQL and NoSQL databases, or large transactional databases and data warehouses. L-family Lsv3, Lasv3, Previous-gen
GPU GPU-optimised VMs are designed for compute-heavy and graphics-intensive tasks. They come with single, multiple, or fractional GPUs for workloads like rendering and machine learning. NC-family NC, NCads_H100_v5, NCCads_H100_v5, NCv2, NCv3, NCasT4_v3, NC_A100_v4
ND-family ND_MI300X_v5, ND-H100-v5, NDm_A100_v4, ND_A100_v4
NG-family NGads V620
NV-family NV, NVv3, NVv4, NVadsA10_v5, Previous-gen
FPGA accelerated Specialised VMs for compute-intensive workloads, such as machine learning inference, video transcoding, database search and analytics. NP-family NP
High-performance compute Built for demanding HPC workloads like fluid dynamics, finite element analysis, EDA, rendering, molecular modelling, geoscience simulations, weather modelling, and financial risk calculations. HB-family B, HBv2, HBv3, HBv4
HC-family HC
HX-family HX

 

You can learn more about Azure virtual machines here: Microsoft Azure VMs overview.

 

How does Azure VM pricing work? 

Since you’ve got here, you’ve probably also seen Microsoft's price overview page. While Microsoft provides complete lists of their VM prices, it’s often still unclear where the costs come from. Let us help you out. 

Looking at the pricing of a general-purpose VM, in this case, the D2 v3, we can conclude that the price build-up is consistent: 

Azure Virtual machine instance pricing table showing vCPU, RAM, and storage options, including pay-as-you-go, one-year, and three-year plans, and spot pricing.

The number after the "D", for example, 2 for D2 v3, indicates the number of cores. D2 has two cores, D4 has four cores, and so forth. In other words, the build-up is linear. 

As you can see, the Gibibytes of RAM to core ratio in this D series is consistently four, meaning the Gibibytes of RAM follow the number of cores multiplied by a factor of four. 

The monthly price (without any operating system) grows linearly with the increase in the number of cores and Gibibytes of RAM. The price also doubles when doubling the number of cores and Gibibytes of RAM. This consistent pricing method is similar across all newer Azure VM series.  

Below, you can see VMs of the DV3 series specified: 

 

Bar graph showing the price per GBRAM for different Azure Dv3 series models (D2 v3, D4 v3, etc.) in EU West, categorized by cores, Linux per GBRAM, and Windows per GBRAM.
Prices of the Dv3 series virtual machines in Microsoft Azure

 

The D8 V3 is priced around $350 a month, pay-as-you-go and running a free Linux OS.

Table showing prices of the Azure D8 v3 VM

Let us now calculate the price per Gibibyte (GiB) of RAM per month for this VM:
$350.40/32GiB= $10.95, let’s round it off at 11 dollars a month per GiB of RAM.

The same is true for all DV3 series VMs. For example, the D4 v3:
$175.2000/16= $10.95. 

As you can see, the prices are constant at $11 per Gibibyte of RAM (GBRAM).

However, when you don’t use Azure Hybrid Benefit (also known as AHUB or AHB), the price goes up as you then need to pay for the OS as well.

That means, the same D8 V3 pay-as-you-go but now including the Windows OS, results in a price of $19.4 per GBRAM. And this pattern stays constant per DV3 series and for other VM series.

Here's an overview of the aggregated pricing of different Azure VM Series in terms of GBRAM:

Bar graph showing the price per gigabyte of RAM (GBRAM) for various DV3 series components, comparing Azure Windows and Linux costs in the EU West region.

As you can see, there are significant price differences between VM series, and therefore, a competitive price offering starts by selecting the VM series with the best price-to-performance ratio.

 

Azure Virtual Machine (VM) Pricing Factors

Azure VM prices aren’t fixed – instead, they are based on several factors. Pricing is influenced by the VM size, the operating system (Linux or Windows), the region where the VM is deployed, and the duration the VM is running and many more factors. Before deploying a VM in Azure, you should know what drives their costs.

 

1. Operating System (OS) licence: Windows vs Linux 

The operating system, known as OS, is often one of the first things that impacts the cost of running a virtual machine in Azure. Azure VMs support both Windows and Linux. Pricing differs based on OS licensing and support fees.

Linux vs Windows Server

Azure Windows VM pricing

Windows VMs in Azure require a Windows Server license, which is baked into the hourly VM price. Hence, it’s often more expensive.

Azure Linux VM pricing

Most Linux virtual machines (VMs) in Azure don't incur additional licensing fees (like Ubuntu, Debian, and CentOS). Therefore, they are often a more cost-effective option than Windows VMs.

Popular Linux distributions available include Ubuntu, CentOS, Debian, Red Hat Enterprise Linux, and SUSE Linux Enterprise Server and more, as you can see in the image below:

Linux distributions for Azure virtual machines

For example: A D2s v3 series running Ubuntu is less expensive than running Windows as OS.

Price table of the D2s v3 Azure virtual machine

The D2s v3 series would cost $154 per month running Windows as OS. Whereas with Ubuntu as OS, you pay only around $87 a month (PAYG).

the D2s v3 virtual machine price table with Azure Hybrid Benefit

In the end, choosing between one or the other depends on your use case. Linux may be cheaper, but if you rely on Microsoft tools, the extra cost of a Windows VM might be worth it. 

Perhaps you’ve already got the required licences to bring to Azure (BYOL) instead of paying twice for the OS.

 

2. Virtual machine size

The size of the virtual machines directly impacts the Azure VM pricing. Each VM size comes with a specific number of virtual CPUs (vCPUs) and memory (RAM). Azure charges more for virtual machines with higher vCPU counts and larger memory allocations.

Larger VMs with more CPU and memory (like D or E-series) are therefore pricier than smaller ones (like B-series).

For example: a B1s VM with 1vCPU and 1 GiB RAM costs far less than a D16s v5 VM with 16 vCPUs and 64 GiB RAM.

B1s virtual machine in azure pricing table
B1s: $8.76 per month
Price table of the D16s v6 virtual machine
D16ss: $705.18 per month

High-performance sizes (like F or H series) cost more because they offer faster CPUs or high-throughput capabilities.

Overall, the more powerful and capable the VM, the higher the hourly cost. Always choose the right SKUs that match your workload needs, as overprovisioning leads to unnecessary waste. 

 

3. Geography

Microsoft’s worldwide data centre infrastructure consists of many regions, and the price depends on the selected region.

A map of Azure 100+ data centres across 36 regions worldwide

The price of Azure VMs depends on the region due to factors like operational costs and demand. Meaning, the same VM can cost more or less depending on where it's deployed. 

For example: a virtual machine (VM) in the US East region and the same VM (DSv3) in West Europe will have different costs. Besides, the price differences are on single items, and the total price difference depends very much on the mix of components you build the pricing on.

Currently, the cheapest region to deploy your VM in is Central India, with an average price of $1.06 per hour. On the other hand, Brazil South East is the most expensive region, with VMs coming at an average price tag of $2.64 per hour.

Tip

When choosing an Azure region, make sure to compare prices at different regions and if possible, choose a less expensive region if compliance isn’t an issue. 

View all average VM prices for Azure Regions.

 

4. Storage types

Storage significantly affects Azure VM costs, often in Azure ways that aren’t immediately visible. Azure offers several storage tiers:

Disk type Description Price starting from
Standard HDD Cheapest option, lower performance. $1.54 per month for 32 GiB
Standard SSD Better performance than HDD at a moderate cost. $2.40 per month for 32 GiB
Premium SSD High IOPS and low latency. More expensive, and often used for production workloads. $5.81 per month for 32GiB
Ultra Disk Highest performance tier. Expensive, for very demanding apps. A 512 GiB disk costs $118.08/month**

*Selected region: West Europe
**Including IOPS and throughput fees

Using Premium storage for workloads that don’t need it drives up your VM costs, which is why you should always try to match your storage against actual needs.

 

5. Network interface

Finally, the network interface affects costs indirectly. Each VM has at least one NIC (network interface card), but Azure charges for outbound data transfers, not the NIC itself. Sending data outside the Azure region, or across peered regions, adds cost. Inbound traffic is generally free, but poorly architected network setups can lead to high egress charges.

 

Azure VM Pricing Models

When you visit Azure’s VM pricing page, you’ll notice Microsoft offers several pricing models for virtual machines. 

  • Pay-As-You-Go
  • Azure Savings plan for compute
  • Reserved Instances
  • Spot

Azure cloud computing pricing options, including Pay as you go, Azure savings plan for compute, Reserved Instances, and Spot options, with descriptions of each pricing model.Understanding how each works helps you avoid overspending and pick the right one for your use case. Let’s break them down:

 

1. Pay-As-You-Go

While prices are always indicated per month – or per hour, Microsoft actually charges their VMs per second. Meaning that, you pay only for the time when the virtual machine is running (pay for what you use) – which makes it cost-effective for more temporary tasks.

You can quit anytime and you don’t need to pay upfront or commit which is why it's the most flexible choice for provisioning VMs in Azure. 

That said, there’s a catch: with PAYG, you pay the list price, which is typically the highest price compared to other pricing models (which we will discuss next).

The D2s v3 instance costs $0.2120/hour with PAYG.

 

2. Reserved Instances

Azure Reservations (reserved instances) are ideal for companies with steady and long-term workloads that want to commit to a fixed number of resources for extended periods.

A VM becomes a reserved instance if the customer is committing to use a volume of specific VM series for 1 or 3 years. If you have a predictable and steady-state workload, and know you are going to need VMs for 1 or 3 years, this model can save a lot of costs. 

Cost comparison of reserved and on-demand D3v2 Linux VMs in US West 2, showing 1-year and 3-year reserved VM instance cost savings.

The longer you commit, the more you can potentially save. Discounts go up from 40% to 70% (even more if you can combine that with AHB).

Think about long-term workloads such as web hosting or a database running (a self-hosted database running onto it like this). At Intercept, we recommend it for stable, always-on applications

Example: Applications such as web servers that need to be running 24/7.

Looking at the D2s v3 instance, you’ll pay:

  • 1 year plan: $0.1685/hour, saving 20% 
  • 3 year plan: Pay 31% less: $0.1446/hour

Price comparison of a virtual machine comparing pay-as-you-go with 1 and 3 year reserved instances

 

3. Azure Spot VMs

Spot VMs are designed for flexible and interruptible workloads. These are perfect for tasks that don’t need uptime and can handle interruption. Azure Spot VMs offers discounts up to 90%. They are useful when running batch jobs or doing a test environment.

Azure Spot VM architecture showing unused and used virtual machine capacity, with a graph illustrating the relationship between discount and unused capacity.

Important: Spot VMs don’t have an SLA. They are ephemeral. In other words, Microsoft can decide any moment to evict these VMs. So, plan accordingly and always have a backup plan for running such instances because these instances are only for non-critical tasks.

The price of the D2s v3 series is $0.0276/hour, which saves you 86% compared to PAYG pricing. Let that sink in…

Table showing Azure Reservation pricing for 1 year, 3 year, and spot options, including hourly rates and percentage savings.

 

4. Azure Free Tier

Yes, you can provision virtual machines in Azure for free. With Azure free tier, you get a free credit 30-day trial worth $200 in Azure credits.

You get 750 hours per month of free burstable VMs: B1s, B2pts v2 (Arm), and B2ats v2 (AMD). These are both for Linux and Windows, valid for 12 months.

Azure VMs for free, offering 12 months of 750 hours burstable virtual machines for Linux and Windows.

 

5. Azure Hybrid Benefit

Azure Hybrid Benefit (AHB) is a pricing offer that helps to maximise the value of existing on-prem Windows Server and SQL Server licence investment while migrating to Azure. Azure Windows VM cost structure with Azure Hybrid Benefit

It allows you to bring your own licence (BYOL) to Microsoft Azure. This way, you don’t need to pay again for the OS licence normally baked in for Windows VMs in Azure.

Without using AHUB, the D8s v3 would cost $619.0400 per month (PAYG). 

Table displaying D8sv3 azure VM with 8, 32, and 64 GB and corresponding monthly costs.

However, sourcing the OS licence under AHB means paying only $350. 

Table showing D8s v3 instance along with the monthly costs, ranging from D2s v3 to D8s v3, with varying vCPUs, RAM, temporary storage, and pay-as-you-go pricing.

Instead of paying the $269 for the Windows Server licence for an entire month running the D8 v3, you can take advantage of AHB and save a lot of money.

Azure Cost Management Whitepaper

Want to save on your monthly Azure cost?

Get our Azure Cost Management Whitepaper! With the best tips, tricks, and background knowledge to optimise your cloud costs.

Download for free!

Azure Virtual Machine Pricing Models compared

As you’ve probably noticed, it can get complex with all the different pricing models. That’s why we’ve outlined an example of the differences between the prices for the D2s v3 Azure VM in the region West Europe:

Instance D2s v3 D4s v3 D8s v3
Pay as you go $154.7600/month $309.5200/month $619.0400/month
Pay as you go with AHB $87.6000/month $175.2000/month $350.4000/month
1 year savings plan

133.7360/month
13% savings

$267.4720/month
13% savings
$534.9440/month
13% savings
1 year savings plan with AHB

$66.5760/month
23% savings

$133.1520/month
23% savings
$266.3040/month
23% savings
3 year savings plan

$115.3400/month
25% savings

$230.6800/month
25% savings
$461.3600/month
25% savings
3 year savings plan with AHB

$48.1800/month
44% savings

$96.3600/month
44% savings
$192.7200/month
44% savings
Spot

$20.1495/month
86% savings

$40.2997/month
86% savings
$80.5993/month
86% savings
Spot with AHB

$11.4055/month
86% savings

$22.8110/month
86% savings
$45.6221/month
86% savings

As you can see, leveraging Azure Hybrid Benefit (AHB) with other pricing models such as reserved instances (either 1-year or 3-year) and Spot results in great cost savings.

Rekenmachine

Azure Cost Scan

We'd like to help you make sense of your costs. Our in-house experts provide a professional savings recommendation, based on your current Azure cloud usage. 

Yes, I want it!

Quick saving tips for Azure VMs

Pick the right VM size

Some VMs offer more memory, some more CPU, others come with GPU. Picking the right size or type of VM for your workload is key. Getting it right the first time might be challenging. At Intercept, our experts can help you by advising you on which VM size to choose so you don’t have to do guesswork or lose unnecessary money in the cloud.

 

Combine Azure Hybrid Benefit with Azure Reservations

A graph showing savings up to 80% with Reserved Instances and Azure Hybrid Benefit

Reserved instances can work if you use the same virtual machine size consistently in your production environment, and the sizing won’t change. But that discount only applies to the compute cost – not the Windows Server licence, which can still make up a large part of the monthly bill.

When you combine them with Azure Hybrid Benefit, you can save up to 80%.

 

Leverage VM Spot pricing with Azure Hybrid Benefit

Even though Spot VMs include a Windows Server licence, you still pay around 60% of the licence cost. That’s where AHB again gives you an edge. Use AHB with Spot VMs to avoid the licence cost from already-discounted pricing and get savings on non-critical workloads. 

 

Resize your virtual machines

Always ask yourself: Do I really need that large VM instance? Overprovisioning is a common happening. Pick the right size for your VM and resize your virtual machines based on actual usage. There are many sizes and types to choose from, each one with a different price. 

Assess your VM usage regularly, use Azure Monitor to track VM performance and identify underutilised resources. Scale the size of your VM to what you need. You might not need a large, expensive VM. Instead, consider switching to a smaller size or a more cost-effective VM family. This avoids paying for unused capacity and helps cut costs.

 

Enable VM autoscaling

Azure VM Scale Sets lets you automatically scale the number of VMs based on real-time demand. This helps lower costs and reduces manual effort by adding or removing instances as needed. You can set rules to scale based on:

  • CPU thresholds
  • Time-of-day schedules
  • Custom metrics

For instance, you might scale out when CPU usage stays above 60% for 15 minutes, and scale in during off-peak hours. This setup gives you flexibility, high availability, and better cost control.

 

Use Spot VMs with VM Scale Sets (VMSS)

Use Spot VMs with VM Scale Sets or auto-scaling to handle variable workloads more efficiently. It’s a cost-effective way to manage demand spikes and save significantly.

 

Virtual Machine Scheduling

Do all your VMs need to run all the time, the whole week? Review each VM in your environment. You might be able to reduce costs by limiting runtime. Set your VM to shut down automatically at a specific time. That way, it's not running 24/7, and you avoid paying for unused hours.

 

Leverage Azure Free Tier

Azure Free Tier lets you explore certain Azure services at no cost, making it ideal for testing and small-scale projects. It includes limited free usage of key services, allowing you to experiment without upfront investment.

Closing thoughts

In this article, we’ve shown you how the pricing of VMs in Azure work, along with some practical examples and tips to save money with your VMs in Azure.

Read for more Azure Cost Optimisation tips to reduce your cloud bill. 

If you are considering migrating to Azure and want to explore the costs of Azure services, the Azure pricing calculator will help you calculate the costs. The tool includes all Microsoft Azure services. You select the services, and the calculator calculates the total price.

As a Microsoft partner, we at Intercept can help you with precise and detailed cost calculations.  

Working Jack

Get in Touch!

Are you looking for ways to save on your Azure costs? Contact us and we will happily help you out.