In short: An Azure Reserved Instance is a billing commitment. You agree to use a specific resource for one or three years, and Microsoft discounts it by up to 72 percent versus pay-as-you-go. Nothing about how the workload runs changes. The discipline that makes it pay is simple: right-size first, then reserve the true baseline, so you never lock in three years of the wrong size. For an MSP, that whole sequence, spotting the reservable spend, validating it against real usage, and packaging it for the client, is billable advisory work. Spotto surfaces the reservable baselines, the resources worth right-sizing, and the coverage gaps across a client's subscriptions, with the numbers, the term options and an effort estimate, so your team can scope and price the work and take it to the client.
How does an Azure reservation work?
You commit to a one-year or three-year term for a given resource type, region and scope. Microsoft then applies the reservation discount automatically to any usage that matches those attributes, so nothing has to move or be reconfigured. You can pay up front or monthly for the same total cost. (Source: Microsoft, What are Azure Reservations.)
Because a reservation is a pure billing construct, it does not affect the runtime state of your resources. The VM runs exactly as before, just billed at the reserved rate.
How much can you save with Azure Reserved Instances?
- Up to 72 percent versus pay-as-you-go on committed compute. (Source: Microsoft.)
- Up to around 80 percent for Windows Server, and up to 85 percent for SQL Server, when a reservation is combined with Azure Hybrid Benefit on eligible workloads. The reservation discounts the compute while Hybrid Benefit removes the licence charge on top. (Source: Microsoft, Azure Hybrid Benefit.)
These are Microsoft's headline maximums. What a client actually realises depends on the estate mix, how much of it is steady, reservable compute versus bursty or already-modernised workloads. Estates with a heavy, always-on VM footprint, common in the client books MSPs manage, have the most reservable baseline to work with, so the blended saving across a real estate is often substantial.
One-year or three-year: which term should you choose?
A three-year term carries the deeper discount but the longer commitment. Match the term to how confident you are the workload will persist. Core, unchanging production workloads suit a three-year term. Workloads you expect to re-architect within a year suit a one-year term, or a savings plan.
What can you reserve?
Reservations are broader than a compute lever. They span three categories:
- Compute: VMs, Dedicated Host, App Service.
- Storage: Blob and Files storage capacity.
- Databases and analytics: Azure SQL Database and Managed Instance, Cosmos DB, Azure Database for MySQL and PostgreSQL, Databricks, Data Explorer, Redis, and more.
So a reservation strategy reaches past "reserve your VMs." Steady spend in storage and analytics is reservable too, and often overlooked. (Source: Microsoft.)
How do Reserved Instances, Savings Plans and Azure Hybrid Benefit fit together?
There are three commitment-and-discount levers, and the biggest saving usually comes from using them together rather than choosing one.
- Reserved Instances commit to a specific resource (SKU and region). They carry the deepest discount, up to 72 percent, with the least flexibility. Best for a stable, predictable baseline. (Source: Microsoft.)
- Azure Savings Plans commit to an hourly dollar amount across compute. They are more flexible, up to 65 percent, and the discount follows usage as the mix shifts, so they suit the flexible remainder. They cannot be cancelled or refunded, so commit only what you are confident you will use. (Source: Microsoft, Azure savings plan for compute.)
- Azure Hybrid Benefit is a different lever again. It lets a client reuse existing on-premises Windows Server or SQL Server licences (with active Software Assurance or a qualifying subscription) to drop the licence charge on Azure, so it attacks the software cost that a reservation and a savings plan both leave untouched. It stacks on top of a reservation rather than competing with it. Microsoft puts the combined saving at up to 80 percent for Windows Server and up to 85 percent for SQL Server versus pay-as-you-go. (Source: Microsoft, Azure Hybrid Benefit.)
The mature pattern reserves the stable baseline with Reserved Instances for the deepest discount, covers the flexible remainder with a savings plan, and stacks Hybrid Benefit on every eligible Windows or SQL workload. Three levers, each attacking a different part of the bill, working together.
Why right-size before you reserve?
A reservation locks in a specific size for up to three years. Buy one on an oversized or idle VM and you have committed to three years of the wrong thing. The order that avoids it is right-size first, using real usage telemetry, establish the true baseline, then reserve that baseline. Microsoft's own guidance is to analyse your usage data, or use reservation recommendations, before you buy. (Source: Microsoft.)
Right-size, then reserve. That two-step is the difference between a real optimisation and a locked-in mistake.
How do you do this across a book of clients?
On your own, the raw signals exist but they sit in different places. Azure Advisor offers reservation recommendations for VMs, and right-sizing lives in a separate view again. Piecing the two together, tenant by tenant, across a book of clients, is the manual work that often does not get done.
Spotto surfaces the whole picture for a client in one place. It finds the underutilised resources worth right-sizing, the steady baselines worth reserving, and the coverage gaps where a discount is sitting unused, across a client's subscriptions. For each opportunity, your team gets:
- The numbers and the term options. Current spend, the reserved rate, the monthly saving, current coverage, and a one-year versus three-year comparison, so you have the numbers to build the business case.
- A realistic effort estimate. A number on the work, so you can plan and price it rather than guess.
- The considerations, already listed. Is the baseline real production demand or inflated by bursts, will seasonal peaks move the workload, who owns the saving, is the same demand already covered elsewhere.
You take those inputs and turn them into a scoped, costed piece of work to put in front of the client: the scope, the effort and the saving, ready to price and deliver.
What does a reservation not cover?
A reserved VM instance covers compute only, not the Windows licence, networking or storage. For Windows VMs and SQL, Microsoft points to Azure Hybrid Benefit to cover the licence portion, and an Azure Savings Plan covers the flexible compute a reservation does not. Stack all three, reservation, savings plan and Hybrid Benefit, and you optimise the whole picture rather than one layer. (Source: Microsoft.)
Can you exchange or refund an Azure reservation?
A reservation is not a trap. You have two ways out: exchange it for another reservation in the same product family, or refund and cancel one you no longer need, up to 50,000 US dollars of cancelled commitment in a rolling 12-month window. Microsoft is not currently charging any early-termination fee. Its terms note that a 12 percent fee may apply to future cancellations, but Microsoft states it has no date set to enable it. So the commitment is real, and the exit today is cheaper than most assume. An Azure Savings Plan, by contrast, cannot be cancelled or refunded at all, one more reason to reserve the firm baseline and only flex the remainder. (Source: Microsoft, Exchange and refund Azure reservations.)
Why reservations are billable advisory work
For a managed services provider, a reservation strategy is a high-value piece of work a client needs and will pay to have done well. Right-sizing an estate, modelling the baseline, choosing between a reservation and a savings plan, applying Hybrid Benefit where it is eligible, and managing exchanges over a three-year term is genuine advisory work.
What the client is buying is a provider who walks into the next review already knowing what matters, what to do about it, and what it is worth. In the client's own language: less waste, fewer invoice surprises, a cloud that's looked after. That's what makes them trust you with more of their estate.
Spotting the opportunity early, before it becomes a wasted year of spend, is the work that is easy to lose in a busy week. Spotto surfaces it and hands your team the finding, the numbers and the considerations, so you can scope, cost and package it for the client.
Take a free look
Pick 3 customers. Take a free look.
You do not have to take our word for it. Choose a few of your customers and run Spotto across their live Azure, read-only, to see the reservable baseline, the coverage sitting unused, and the saving already there. Nothing is changed. You simply see what has been hard to see, and take the work to the client.
Start with your own customers on the platform.
Common questions
What is an Azure Reserved Instance?
A one-year or three-year commitment to a specific Azure resource in exchange for a discount of up to 72 percent off pay-as-you-go. It is a billing discount and does not change how the resource runs.
How much do Azure Reserved Instances save?
Up to 72 percent versus pay-as-you-go, and up to around 80 percent for Windows Server (up to 85 percent for SQL Server) when combined with Azure Hybrid Benefit on eligible workloads. Those are Microsoft's stated ceilings; what a client realises depends on how much of the estate is steady, reservable compute.
What is the difference between Reserved Instances and Savings Plans?
Reserved Instances commit to a specific resource for a deeper discount, up to 72 percent. Savings plans commit to an hourly spend across compute for more flexibility, up to around 65 percent, and the discount follows your usage as the mix shifts. Many estates use both: a reservation for the firm baseline, a savings plan for the flexible remainder.
Do Reserved Instances, Savings Plans and Azure Hybrid Benefit work together?
Yes. They are three separate levers, each attacking a different part of the bill, and the biggest saving comes from combining them. Reserve the stable baseline with Reserved Instances (up to 72 percent off compute), cover the flexible remainder with an Azure Savings Plan (up to 65 percent), and stack Azure Hybrid Benefit, which lets a client reuse on-premises Windows Server or SQL Server licences with Software Assurance, to drop the licence charge on top. Microsoft puts the combined saving at up to 80 percent for Windows Server and up to 85 percent for SQL Server versus pay-as-you-go.
Should I right-size before buying a reservation?
Yes. Reserving an oversized resource locks in the wrong size for years. Right-size using real usage data first, then reserve the true baseline. This is Microsoft's recommended approach.
Can you cancel or refund an Azure reservation?
Yes. You can exchange it for another reservation in the same product family, or refund and cancel it, up to 50,000 US dollars of cancelled commitment in a rolling 12-month window. Microsoft is not currently charging an early-termination fee; its terms note a 12 percent fee may apply to future cancellations, but no date has been set to enable it. Azure savings plans, by contrast, cannot be cancelled or refunded at all.
Who owns an Azure reservation if the client leaves?
Whoever owns the billing scope it was purchased under. In a CSP model the partner usually holds and manages it, and a CSP customer cannot cancel or exchange a reservation themselves; they ask their partner to do it. If a clean break matters, keep the client's Azure in their own name so the reservation and its commitment leave with them. If you reserve on their behalf under your own billing, you carry the term. Decide who owns the reservation up front, and price that risk in.
Can a Reserved Instance cost an MSP more than it saves the client?
Yes, if you reserve the wrong thing. Reserving a temporary spike as if it were a baseline, or reserving before right-sizing, locks in three years of the wrong size. And if you hold the commitment in your own billing and the client's usage drops or they leave, you keep paying. The discipline that avoids it is always the same: right-size from real usage first, reserve only the firm baseline, flex the rest with a savings plan or on-demand, and be explicit about who owns the commitment.
About Spotto
Spotto is an AI-Assisted Cloud Operations platform for Cloud MSPs and cloud teams. It investigates a client's Azure estate and surfaces the work hiding inside it, from reservable baselines and resources worth right-sizing to coverage gaps, with the numbers, term options and effort estimate attached, so an MSP can scope, price and take the work to the client.