Azure Reserved Instances (RIs) are the single most effective mechanism for reducing Azure compute costs — delivering 40–65% discounts versus pay-as-you-go pricing for workloads that run predictably. Despite this, most enterprise Azure deployments leave significant RI savings unrealised: either by under-purchasing reservations, purchasing the wrong scope, or failing to manage the RI portfolio as workloads evolve.
This guide covers how Azure Reserved Instances actually work, the economics of 1-year versus 3-year commitments, scope and flexibility management, the exchange and cancellation rules that govern your risk, and the buying strategy that maximises savings without creating stranded commitments.
How Azure Reserved Instances Work
An Azure Reserved Instance is a billing commitment, not a resource commitment. This distinction matters significantly for how you manage them. When you purchase a reservation:
- You commit to paying for a specific VM size (or resource type) in a specific region for 1 or 3 years
- Azure automatically applies the reservation discount to matching resources that run in your subscription(s)
- You are billed the reservation cost regardless of whether your VMs are running — idle VMs still consume the reservation but deliver no compute output
- The reservation discount applies as a billing discount, not as a capacity reservation — capacity availability is not guaranteed
Reserved Instances are a billing commitment, not a capacity guarantee. Purchasing a reservation does not guarantee VM availability in that region. For capacity reservations (guaranteed availability), Azure offers On-Demand Capacity Reservations as a separate, higher-cost product. Most enterprises do not need capacity reservations — RI is sufficient for predictable baseline workloads.
Discount Mechanics: Payment Options and Rates
Azure Reserved Instances offer three payment models, each with different discount rates and cash flow implications:
| Payment Option | 1-Year Discount | 3-Year Discount | Cash Flow | Best For |
|---|---|---|---|---|
| Upfront (All) | ~40% | ~60–65% | Full payment on purchase | Stable workloads, strong cash position |
| Monthly | ~35–38% | ~55–60% | Spread over term, no interest | Predictable workloads, cash flow managed |
| Partial Upfront | ~37–39% | ~57–62% | 50% upfront, remainder monthly | Compromise between cash and discount |
The discount difference between upfront and monthly for 1-year reservations is typically 2–4 percentage points. For most enterprises, monthly payment preserves cash flow at a minimal discount cost — the difference on a $10K/month commitment is approximately $3K over the year. Unless cash is constrained, monthly 1-year reservations offer the best balance of discount and flexibility.
Scope Options: Shared vs Single Subscription
When purchasing a reservation, you must specify the scope — which subscriptions the reservation applies to. This is one of the most commonly misconfigured aspects of Azure RI management.
Single Subscription Scope
The reservation discount applies only to the subscription specified at purchase. If that subscription's workloads change, the reservation may go unused while matching workloads in other subscriptions pay PAYG rates. Single subscription scope is appropriate for isolated workloads that will remain in a specific subscription long-term.
Shared Scope
The reservation applies to any subscription within the same billing account or management group. Shared scope maximises utilisation by automatically applying discounts across subscriptions — if one subscription's workloads scale down, the reservation applies to matching resources in other subscriptions. Shared scope is the recommended default for enterprise organisations with multiple subscriptions.
Management Group Scope
Introduced in 2023, management group scope allows reservations to apply to all subscriptions within a specific management group. Useful for organisations that have separated subscriptions by business unit but want RI benefits to flow within a unit without sharing across the entire tenant.
Default to Shared scope for all RI purchases unless there is a specific billing or chargeback reason to restrict scope. Shared scope significantly improves utilisation rates and reduces the risk of unused reservations as workloads shift between subscriptions.
Instance Size Flexibility
Azure's instance size flexibility feature allows a reservation to apply to different VM sizes within the same instance series family — without requiring an exchange. This is critical for managing reservations as workloads scale.
Instance size flexibility is enabled by default on most VM reservations. A reservation for one Standard_D4s_v5 (4 vCPUs) can apply to one Standard_D8s_v5 (8 vCPUs) at 50% utilisation, or two Standard_D2s_v5 (2 vCPUs) at 100% utilisation. The normalisation factor is based on vCPU count within the same series.
Instance size flexibility does not apply across series — a D-series reservation does not apply to E-series VMs. Understanding your VM series mix is important before committing reservations at scale. The RI vs Savings Plans comparison explains when Savings Plans (which are series-agnostic) are preferable to RIs.
Exchange and Cancellation Rules
The rules governing RI changes are commercially significant. Understanding them before purchasing prevents regret.
Self-Service Exchanges
You can exchange a reservation for a new reservation of equal or greater value at any time. The exchange is completed in the Azure Portal. Key rules:
- The new reservation must be of equal or greater total value (pro-rated remaining value applies)
- You can change region, VM size, series, term, and payment option in a single exchange
- Exchanges are not limited in frequency — you can exchange as many times as needed
- No penalty for exchanging, but you cannot exchange for a lower-value reservation and receive a refund for the difference
Cancellations (12% Penalty)
You can cancel a reservation, but Microsoft applies a 12% early cancellation fee on the remaining prepaid value. The refund cap is $50,000 per rolling 12-month period across your billing account. For monthly-payment reservations, you simply stop future monthly payments — the 12% penalty and cap apply to the remaining future payments.
Microsoft removed the unlimited self-service exchange policy in January 2024 for some reservation types. Virtual machine reservations retain self-service exchange, but Azure SQL Database, Azure Cosmos DB, Azure Synapse, and several other service reservations now require contacting Microsoft Support for exchanges. Verify the exchange policy before purchasing non-VM reservations.
Reserved Instances vs Azure Savings Plans
Azure Savings Plans (introduced in 2022) provide flexible hourly spend commitments that apply across VM families, regions, and operating systems. The tradeoff versus RIs:
| Dimension | Reserved Instances | Savings Plans |
|---|---|---|
| Discount level | Higher (40–65%) | Lower (15–45%) |
| Flexibility | Limited to instance series (with flexibility) | Applies to any VM, any region, any OS |
| Commitment basis | Specific VM type/region | Hourly spend amount |
| Best for | Stable, predictable workloads | Dynamic, cross-region workloads |
| Exchange/cancel | Exchange allowed, cancel with 12% fee | Cannot exchange or cancel |
The recommended approach for enterprise organisations: use RIs for the stable baseline (typically 60–70% of your compute footprint), Savings Plans for the dynamic workload layer (10–20%), and PAYG for truly variable/burst workloads (10–30%). This layered approach achieves 50–60% blended savings without the inflexibility risk of over-committing RIs.
Azure Services Covered by Reservations
RIs extend well beyond virtual machines. Understanding the full scope of reservable Azure services is often the difference between an adequate and an excellent reservation programme:
- Virtual Machines: All Windows and Linux VM series — the core RI use case
- SQL Database and Managed Instance: vCore-based SQL workloads see 40–55% discounts; SQL Managed Instance discounts are comparable
- Azure Cosmos DB: Request Unit (RU) commitments for consistent Cosmos throughput
- Azure Synapse Analytics: DWU commitments for data warehouse workloads
- Azure Cache for Redis: Cache instance reservations
- Azure App Service: Isolated plan stamp reservations
- Azure Dedicated Host: Physical host reservations for compliance-sensitive workloads
- Azure Disk Storage: Managed disk reservations
SQL Database and Managed Instance reservations often deliver the highest per-dollar impact in enterprise environments because SQL licensing costs on PAYG are elevated. Combining a SQL MI reservation with Azure Hybrid Benefit for SQL Server licences can reduce database costs by 60–75% versus full PAYG pricing.
Enterprise RI Buying Strategy
The systematic approach to building an enterprise RI portfolio:
Step 1: Identify Stable Workloads
Pull Azure utilisation data for the past 90 days. Identify VMs and services with consistent 70%+ utilisation. These are RI candidates. VMs with utilisation below 50% or highly variable patterns are better suited to Savings Plans or PAYG.
Step 2: Start with 1-Year Monthly for Core Workloads
For your first RI purchase, 1-year monthly payment provides strong discounts (35–38%) with flexibility to course-correct. Avoid committing to 3-year upfront on workloads you haven't observed running stably for at least 6 months.
Step 3: Migrate Proven Workloads to 3-Year
After 6–12 months, workloads that have run at consistent utilisation become candidates for 3-year upfront reservations. The additional discount (60–65% vs 35–38%) on stable workloads is material. At $10K/month PAYG, 3-year upfront ($43K) versus 1-year monthly ($77K/year equivalent) saves an additional $30K+ over 3 years.
Step 4: Use Shared Scope and Set Up Utilisation Monitoring
Enable shared scope on all RIs and configure Azure Cost Management utilisation alerts. Alert when any reservation drops below 80% utilisation — this is the trigger to review whether an exchange is warranted.
Step 5: Layer Savings Plans for Coverage Above RI Baseline
Once your RI portfolio covers your stable baseline, add a Savings Plan commitment to cover the dynamic workload layer. The combined approach achieves 50–55% blended savings across your total compute footprint.
The comprehensive framework for Azure cost optimisation — including RI strategy, Savings Plans, rightsizing, and governance — is in the Azure cost optimisation guide. For the MACC dimension of commitment purchasing, see the MACC leverage guide.
Common RI Mistakes to Avoid
- Single subscription scope: Limits utilisation and increases stranded reservation risk as workloads shift
- Over-committing stable workloads at 3-year before proving stability: Locks in commitments on workloads that may be decommissioned
- Ignoring SQL and PaaS service reservations: VM-only RI programmes leave significant database savings unrealised
- Purchasing without Azure Hybrid Benefit activated: AHB should always be activated before purchasing RIs — it changes the baseline PAYG price against which RIs are discounted
- No utilisation monitoring: Unused reservations deplete the savings ROI without triggering alerts
- Purchasing RIs before rightsizing: RI savings are applied to whatever size VM you're running — right-size first, then commit
For the complete Azure cost reduction framework including rightsizing, MACC, and governance, our Azure Cost Management advisory service covers all dimensions as a structured engagement.