Azure reservations explained: how to cut compute cost with commitments

Quick answer

Azure reservations let you commit to a service (VMs, SQL Database, Cosmos DB, and more) for one or three years in exchange for a large discount versus pay-as-you-go, commonly in the range of a substantial double-digit to over 50 percent for three-year terms. The key concepts are scope (single subscription vs shared across your enrollment), flexibility (instance size flexibility lets one reservation cover different sizes in a family), and exchange or refund rules. Buy reservations for steady, predictable baseline usage; keep spiky or uncertain usage on pay-as-you-go.

If you run steady workloads on Azure, pay-as-you-go pricing leaves money on the table. Azure reservations trade a one or three year commitment for a large discount, and for baseline capacity that runs all year the savings are substantial. The art is committing to what you will genuinely use while keeping flexibility for the parts of your usage that vary.

What reservations discount

Service	Reservation covers
Virtual Machines	Compute cost of a VM family in a region
SQL Database / Managed Instance	Compute (vCore) capacity
Cosmos DB	Provisioned throughput
Other services	App Service, storage, and more, depending on offering

A reservation is a billing construct, not a capacity guarantee by default: you commit to spend on a service for the term and get a discounted rate applied to matching usage. The discount grows with term length, a three-year reservation saves more than a one-year, in exchange for a longer commitment.

Scope and flexibility

Scope controls which usage a reservation applies to: single-subscription scope applies the discount only within one subscription, while shared scope spreads it across all subscriptions in your billing account, maximizing utilization. Instance size flexibility (for supported VM families) lets one reservation cover different sizes within the same family, so a reservation for a mid-size VM can also discount two smaller ones or half a larger one. This flexibility reduces the risk of a reservation going unused if your exact sizes change.

Reservations versus savings plans

Azure also offers savings plans for compute, which commit to an hourly spend amount rather than a specific VM family, trading a slightly smaller discount for more flexibility across VM types and regions. Reservations give the deepest discount for a known, stable family; savings plans give broader coverage when your compute mix shifts. The trade-off mirrors the AWS reserved-instances-versus-savings-plans choice, so match the commitment type to how predictable your usage is.

Avoiding overcommitment

Buy reservations for the steady baseline you are confident you will run for the full term, and keep spiky, seasonal, or uncertain usage on pay-as-you-go. Track reservation utilization (an underused reservation wastes the commitment) and use exchange or refund options if your needs change. Set budget alerts and treat commitment coverage as a KPI. Because a poor reservation decision locks in cost for years, it is worth modeling first: pricing your steady infrastructure against the resource catalog and comparing it to reserved rates shows how much a commitment would save before you make it. Reservations are one of the biggest levers on the largest cost drivers, which are usually compute and databases.

FAQ

What is an Azure reservation?

A commitment to use a service (VMs, SQL Database, Cosmos DB, and others) for one or three years in exchange for a large discount versus pay-as-you-go, commonly a substantial double-digit percentage and more for three-year terms. It is a billing construct that applies a discounted rate to matching usage over the term, best suited to steady, predictable baseline workloads.

How much do Azure reservations save?

The discount depends on the service and term, but three-year reservations commonly save a large fraction versus pay-as-you-go, often reaching over half for some services, while one-year reservations save less. The longer the commitment, the deeper the discount. For baseline capacity that runs all year, the savings are substantial and compound over the term.

What is the difference between single and shared reservation scope?

Single-subscription scope applies the reservation discount only to matching usage within one subscription. Shared scope spreads the discount across all subscriptions in your billing account, so any matching usage anywhere can absorb it. Shared scope usually maximizes utilization because it is less likely a reservation goes unused, which is why many organizations default to it.

What is instance size flexibility in Azure reservations?

For supported VM families, instance size flexibility lets one reservation cover different sizes within the same family. A reservation sized for a mid-size VM can instead discount two smaller VMs or part of a larger one in that family. This reduces the risk of a reservation going unused if your exact instance sizes change during the term.

Should I use Azure reservations or a savings plan?

Use reservations for a known, stable VM family to get the deepest discount. Use a savings plan for compute when your compute mix shifts across VM types or regions, since it commits to an hourly spend amount rather than a specific family, trading a slightly smaller discount for broader flexibility. Match the commitment type to how predictable your usage is.

Does C3X help decide on Azure reservations?

C3X prices your steady infrastructure against a live catalog before deploy, so you can see baseline pay-as-you-go cost and compare it to reserved rates to judge how much a commitment would save. Because a reservation locks in cost for years, modeling the baseline first helps you commit to what you will genuinely use rather than overcommitting.

What to do next

Model your baseline cost before committing to a reservation. C3X prices your Terraform against a live resource catalog.