Stop Overpaying for Azure VMs: RIs vs. Savings Plans (With AVD-Specific Recommendations)

Introduction

Controlling cloud costs is a constant challenge for organisations running Virtual Machines in Azure. With multiple pricing models and commitment options, it is easy to make the wrong choice and end up paying more than expected and exceed budget.

Moving ahead than standard pay-as-you-go (PAYG) pricing, Reserved Instances and Savings Plans are two of the most powerful tools that Microsoft offers to help reduce VM spend. However, they differ in how they work, quite significantly.

Choosing the right one depends on your workload stability, resource planning and how well your architecture supports their scope and assignment.

In this post, I’ll break down what Reserved Instances and Savings Plans actually are, how they differ, when to use each, and why scope and assignment are critical to getting the discounts you expect. I’ll also share real-world guidance based on pricing calculator estimates, AVD architecture, and common pitfalls that can drive unexpected costs.

Let’s get into it. 💸

What Are Reserved Instances

Reserved Instances (RIs) are a commitment-based pricing model in Azure that allow organisations to pre-pay for virtual machine usage over a one-year or three-year term. In return, Microsoft offer significant discounts compared to pay-as-you-go pricing. Looking at the UK South region, organisations can see a discount of approx. 60%!

Here's what you need to know about how Reserved Instances work, and why they are best suited to predictable workloads.

Key Characteristics of Reserved Instances

• Fixed Commitment: Choosing either a one-year or three-year term. The longer commitment generally offers the better discounts. For example, in the UK South region, a one-year term for a B2ms VM will provide approx. 41% discount, whereas a three-year term will provide approx. 62% discount.
• Scope Assignment: Reservations can be applied at subscription, resource group, or a management group level. The scope determines which VMs the reservation applies to, which can have cost implications if misaligned.
• Applies to Matching Resources: The RI discount typically only applies to VMs that match the size, series, region, OS type and scope selected at the time of purchase.
• VM Size Flexibility: For certain VM families, such as D-Series v3, Azure allows the discount to apply across different sizes within the same family, offering limited flexibility, provided the reservation is configured correctly.
• Billing: RIs do not reserve capacity or lock in a specific VM. RIs provide discounted usage against the billing commitment.
• Payment Options: Organisations can pay upfront or choose monthly payments. The level of discount remains the same either way.

What Are Savings Plans?

Savings Plans are a flexible, commitment-based pricing model in Azure that provides discounted compute costs in exchange for committing to a fixed-hourly spend over a one-year or three-year term. Still looking at the UK South region, organisations can see a discount of approx. 45%!

Unlike Reserved Instances, which are tied to specific VM characteristics, Savings Plans apply more broadly to eligible compute usage, regardless of region, VM series, or operating system.

This flexibility makes them ideal for dynamic workloads where VM sizes, types or deployment locations could change over time.

Key Characteristics of Savings Plans

• Spending Commitment: Organisations commit to a consistent hourly spend (e.g. $1.50 / per hour) over a one-year or three-year term. Any usage within this committed spend is charged at the discounted Savings Plan rate.
• Broader Coverage: Savings Plans apply across VM series, regions, OS types and deployment types. This makes them well-suited to changing environments.
• Scope Assignment: Savings Plans can be scoped to a shared billing account, a specific subscription or an individual resource group. This allows organisations to target cost savings to nearly the exact workload required. For example, assigning a Savings Plan to the resource group hosting your AVD environment ensures the discount is applied correctly, rather than being consumed by other compute resources in the same subscription.
• VM Flexibility: If you change the size, region, or series of your VM, the discount still applies. Providing compute usage stays within the committed hourly spend.
• Billing and Usage: Organisations are charged the committed hourly rate regardless of whether the full amount is used. If customers exceed the hourly commitment, the overage is billed at PAYG rates.
• Payment Options: Similarly to RIs, Microsoft offer the option to pay upfront or on a monthly basis. The discount remains the same regardless of the payment method.

Reserved Instances vs. Savings Plans: What's The Real Difference?

While both RIs and Savings Plans offer significant discounts for Azure compute workloads, their structure, flexibility and application differ in a number of ways. Understanding these differences is essential for choosing the right model based on your workload, environment and commercial strategy.

Here's a handy table to compare RIs and Savings Plans.

When To Use RIs & When To Choose Savings Plans

Choosing between Reserved Instances and Savings Plans depends on the nature of your workloads, your cost control objectives, and how predictable your compute requirements are.

When Are Reserved Instances A Good Fit?

Reserved Instances work best for stable, long-term workloads where consistency and predictability are key.

• Organisations running consistent workloads such as domain controllers or fixed application servers.
• Business' VM sizing, region and platform requirements are stable and unlikely to change during the commitment period.
• You want predictable cost control and can commit to long-term infrastructure planning.
• Customers are comfortable with lower flexibility in exchange for a higher discount.
• A well-defined Azure architecture that will not change significantly during the commitment term.

When Are Savings Plans A Better Choice?

Savings Plans are more suitable for dynamic environments where flexibility and simplicity are more important than a maximum discount.

• Customers running dynamic workloads i.e. AVD session hosts. This includes autoscaling VMs and services that scale up and down based on demand.
• You anticipate changes in VM families, regions, or types, and need flexibility over exchanging reserved instances.
• An environment which is still evolving, such as during migrations, proof-of-concepts or scaling up new AVD services.

Configuring a Reserved Instance

Because of customer privacy, and my Azure subscription is credit-based, I can't add a video demonstration of configuring a reserved instance. Here are the steps to do so:

1. Sign into the Azure Portal.
2. Navigate to the Reservations blade.
3. Click Add or Purchase.
4. Choose the Product Type e.g. Virtual Machine.
5. Configure the Reservation
   1. Select the Azure region where your machines are deployed.
   2. Choose the VM Series - Azure is clever enough to show you RIs that will compatible with your VMs at this point.
   3. Term: Select the 1-year or 3-year term.
   4. Billing Frequency: Choose Monthly or Upfront. The discount remains the same so monthly is usually the preferred option.
   5. Quantity: Define how many instances you want to reserve. Remember, this is one RI per VM.
6. Select the Scope and assign the RI to the required scope.
7. Review and Purchase.

Configuring a Savings Plan

1. Sign into the Azure Portal.
2. Go to Savings Plans.
3. Click Add or Purchase.
4. Select the Plan Type as Compute Savings Plan.
   1. Hourly Commitment: Define how much you're willing to spend per hour. Based on the scope you select, this should automatically add the required compute hours for the maximum discount from Azure Advisor.
   2. Term: Select the 1-year or 3-year term.
   3. Billing Frequency: Choose Monthly or Upfront. The discount remains the same so monthly is usually the preferred option.
   4. Select the Scope and assign the Savings Plan to the required scope. For AVD workloads, assigning at resource group level gives you fine-grained control over where the discount applies.
   5. Review and Purchase.

Licensing & Commitment Terms

Reserved Instances and Savings Plans are financial commitments. You are agreeing to either a specific resource setup, or a set hourly spend, for one or three years.

Three-year terms unlock the highest discounts, but require long-term confidence in the environment and workload. One-year terms offer more flexibility, with a smaller discount.

However, you are not buying a VM, you are only committing to usage. These commitments sit at the billing account level and apply for the full term, regardless of how usage changes.

For CSP customers, any changes or cancellations often need to be handled by your provider e.g. Pax8. Direct customers manage everything in the Azure portal.

The Key To Accurate Azure Costs

Arguably the most important section, particularly for finance directors.

One of the most common mistakes with Reserved Instances and Savings Plans is misalignment of scope. If your commitment is scoped too broadly, the benefit may be consumed by the wrong workload. Too narrow, and you might not get full utilisation.

A clean way to leverage Savings Plans for AVD, is to place session hosts into a dedicated resource group. This makes it easier to assign a Savings Plan directly to that resource group, so the discount is focused on the compute you intended.

For RIs, similar logic applies, however, typically RIs will be utilised based on the billing subscription. Therefore, if you have a VM SKU, within a subscription, then Azure is clever enough to discount the cost of that SKU, in that subscription.

Timing Matters

Timing matters as well as structure.

Let your AVD environment run for 30 days on PAYG pricing - and thank me later. (Ensure that the customer, or your organisation is aware that the first 30-days will be PAYG).

This will give you a chance to validate performance, tune session host sizing and confirm auto-scaling is working properly. After 30 days, Azure Advisor will surface the required hourly compute discount to achieve the maximum discount. The 30-day window ensures your commitment is based on real usage, not guesswork.

Getting this wrong often means underutilised discounts, duplicated costs, or scope overlap that benefits other workloads instead. In short, structure and patience both play a part in accurate forecasting and cost optimisation.

Final Thoughts

Reserved Instances and Savings Plans are not just discount mechanisms, they're strategic tools for controlling and optimising cloud spend. But the benefit is only realised when they're applied deliberately, scoped correctly, and aligned with the real-world behaviour of your workloads.

For stable, predictable environments, Reserved Instances offer powerful long-term savings. For dynamic, fast-changing workloads like AVD, Savings Plans provide the flexibility needed without giving up too much on cost.

If there’s one takeaway, it’s this: get your architecture in order, let usage patterns settle, and commit based on data, not assumptions. The more mature your environment, the more confident your cost commitments will be.

Spend wisely. Architect with purpose.