Azure’s pay-as-you-go pricing is straightforward, but it is also the most expensive way to run cloud infrastructure. Organizations that run production workloads around the clock on pay-as-you-go rates are paying a premium for flexibility they are not using — the ability to stop a resource at any moment without financial consequence. For workloads that run continuously for months or years, that flexibility has no practical value, but the premium is real: 30-72% more than committed pricing depending on the service and term length.
Microsoft offers two commitment-based pricing models to address this: Reserved Instances (RIs) and Savings Plans. Both require a 1-year or 3-year financial commitment in exchange for significant discounts. But they work differently, apply to different services, and offer different levels of flexibility. Choosing the wrong one — or using them incorrectly — can lock you into commitments that waste money instead of saving it.
This guide breaks down how each model works, when to use one over the other, and how to combine them for maximum savings. If you are looking for a broader view of cloud financial management, our guide on FinOps strategies covers the framework and practices that surround these purchasing decisions.
How Pay-As-You-Go Pricing Works
Before comparing RIs and Savings Plans, it helps to understand the baseline they are both discounting against.
Azure’s default pricing model is pay-as-you-go (PAYG). You consume resources — compute hours, storage, network bandwidth — and you are billed at the listed meter rate for each unit consumed. There are no upfront costs, no commitments, and no minimums. You pay for what you use, and you can stop at any time.
Why PAYG costs more:
- You are paying the full retail rate for every hour of compute
- There is no volume discount for sustained usage
- The rate does not decrease over time, regardless of how long you run the resource
- You are effectively subsidizing the flexibility that other customers use when they spin up short-lived resources
When PAYG makes sense:
- Short-term projects with a defined end date (less than 12 months)
- Highly variable workloads where demand is unpredictable
- Proof-of-concept environments that may not persist
- Development and testing environments that run only during business hours
For production workloads that run 24/7 with predictable resource needs, PAYG is almost always the most expensive option. The question is not whether to commit, but how.
What Are Azure Reserved Instances?
Azure Reserved Instances (RIs) are a commitment to use a specific type and quantity of Azure resource for a 1-year or 3-year term. In exchange for this commitment, Microsoft provides a discount off the PAYG rate. The discount is applied automatically to matching resources — you do not need to modify your deployments or change how you provision infrastructure.
How RIs Work
When you purchase an RI, you are committing to a specific:
- Service: Virtual Machines, SQL Database, Cosmos DB, Azure Dedicated Host, App Service, and many others
- Region: The Azure region where the resource runs (e.g., East US, West Europe)
- VM size or service tier: The specific size family or performance tier (e.g., D4s_v5, Standard S1)
- Quantity: The number of instances covered by the reservation
- Term: 1 year or 3 years
The discount applies automatically to running resources that match these attributes. If you purchase a 1-year reservation for two D4s_v5 VMs in East US and you have two D4s_v5 VMs running in East US, those VMs receive the discounted rate automatically. If you have only one matching VM, only one reservation is utilized and the other goes unused — you still pay for both reservations.
Scope Options
RIs can be scoped at different levels, which determines how broadly the discount is applied:
- Single subscription: The RI discount applies only to matching resources within a specific subscription
- Resource group: The discount is limited to resources within a specific resource group
- Shared: The discount applies across all subscriptions within a billing account or enrollment
- Management group: The discount applies across all subscriptions within a management group
Best practice: Use shared scope unless you have a specific reason not to. Shared scope maximizes the chance that your reservation is utilized because it can match resources in any subscription within your billing account. This is especially useful if VMs move between subscriptions during reorganizations.
Payment Options
RIs offer three payment options:
| Payment Option | Description | Discount Level |
|---|---|---|
| All upfront | Pay the entire reservation cost at purchase | Highest discount |
| Monthly (no upfront) | Pay in equal monthly installments over the term | Lowest discount (still significant vs. PAYG) |
| Partial upfront | Pay a portion at purchase, the rest monthly | Middle ground |
The difference between all-upfront and monthly payment is typically 1-5% in additional savings. For most organizations, the flexibility of monthly payments outweighs the marginal extra discount from paying upfront.
RI Discount Percentages
RI discounts vary by service, region, and term length. The following are representative ranges:
| Service | 1-Year RI Savings | 3-Year RI Savings |
|---|---|---|
| Virtual Machines | 30-40% | 55-72% |
| SQL Database | 25-35% | 50-65% |
| Cosmos DB | 20-30% | 40-60% |
| Azure Dedicated Host | 30-40% | 55-65% |
| App Service (Premium) | 30-40% | 50-55% |
| Azure Cache for Redis | 25-35% | 50-60% |
| Azure Synapse Analytics | 25-35% | 50-60% |
Exact percentages depend on the specific SKU and region. You can check current RI pricing on the Azure pricing calculator or within the Azure portal’s reservation purchase flow.
Instance Size Flexibility
One of the most important RI features is instance size flexibility for VMs. When enabled (it is on by default for most VM series), a reservation for a specific VM size automatically covers other sizes within the same VM series in the same region.
For example, if you purchase a reservation for a D4s_v5 (4 vCPUs), that reservation can also cover:
- Two D2s_v5 instances (2 vCPUs each)
- One D8s_v5 at half utilization (8 vCPUs)
- Four D1s_v5 instances (1 vCPU each)
Azure uses a normalized ratio system where each VM size in a series is expressed as a multiple of the smallest size. This means you do not need to perfectly match your reservation to your exact VM size — you need to match the total compute capacity within a series.
This flexibility is powerful but limited. It does not apply across regions, across VM series (D-series cannot cover E-series), or across services (a VM RI cannot cover SQL Database).
What Are Azure Savings Plans?
Azure Savings Plans are a newer commitment model introduced in late 2022. Instead of committing to a specific resource type and size, you commit to a specific dollar amount of compute spending per hour for a 1-year or 3-year term.
How Savings Plans Work
When you purchase a Savings Plan, you commit to spending a fixed amount per hour (e.g., $5.00/hour) on eligible compute services. Azure automatically applies the Savings Plan discount to your most expensive eligible usage first, maximizing your savings.
Key difference from RIs: you are not locking into a specific VM size, series, or region. You are locking into a spending level.
Two Types of Savings Plans
Azure offers two Savings Plan types:
Compute Savings Plan:
- Applies to Azure Virtual Machines, Azure Dedicated Host, Azure Container Instances, Azure App Service, and Azure Functions Premium
- Flexible across VM series, sizes, regions, and operating systems
- Provides the broadest flexibility
Azure Premium SaaS Savings Plan (formerly General Savings Plan):
- Applies to select Azure premium SaaS services
- Covers a narrower set of services with specific eligibility
For most organizations, the Compute Savings Plan is the relevant option.
Savings Plan Discount Percentages
Savings Plan discounts are slightly lower than RI discounts for the same term length and service. This is the trade-off for greater flexibility:
| Service | 1-Year SP Savings | 3-Year SP Savings |
|---|---|---|
| Virtual Machines | 15-33% | 40-65% |
| App Service (Premium) | 15-25% | 35-50% |
| Azure Container Instances | 15-25% | 35-50% |
| Azure Functions Premium | 15-25% | 35-50% |
The exact discount depends on the specific resource and region. You can model scenarios using the Azure pricing calculator.
Scope Options for Savings Plans
Savings Plans support the same scoping options as RIs:
- Single subscription
- Resource group
- Shared (billing account)
- Management group
As with RIs, shared scope is the recommended default to maximize utilization.
Head-to-Head Comparison: RIs vs Savings Plans
Here is a direct comparison across the dimensions that matter most:
| Dimension | Reserved Instances | Savings Plans |
|---|---|---|
| Commitment type | Specific resource (VM size, region, service) | Dollar amount per hour |
| Discount depth | Higher (up to 72% for 3-year) | Slightly lower (up to 65% for 3-year) |
| Term options | 1 year, 3 years | 1 year, 3 years |
| VM series flexibility | Within same series only (instance size flexibility) | Across all VM series |
| Region flexibility | No (locked to purchased region) | Yes (applies across all regions) |
| Service flexibility | No (VM RI only covers VMs) | Yes (covers VMs, ACI, App Service, Functions) |
| OS flexibility | Yes (Windows/Linux within same RI) | Yes |
| Best discount application | You manage which resources match | Azure automatically applies to highest-cost usage |
| Exchange/refund | Limited (exchange policy tightened in 2024) | No exchanges or refunds |
| Cancellation | Early termination fee applies | Not cancellable |
| Eligible services | VMs, SQL DB, Cosmos DB, Redis, Synapse, Dedicated Host, App Service, and others | Compute services (VMs, ACI, App Service, Functions Premium) |
| Non-compute coverage | Yes (databases, analytics, caching) | No |
When RIs Win
RIs provide deeper discounts and are the better choice when:
- You have stable, predictable workloads that will not change VM series, size, or region for the term length
- You are committing to non-compute services like SQL Database, Cosmos DB, or Azure Cache for Redis (Savings Plans do not cover these)
- You want the maximum possible discount and are willing to accept less flexibility
- Your infrastructure is well-established and changes infrequently
When Savings Plans Win
Savings Plans are the better choice when:
- Your compute workloads change VM series or sizes frequently
- You are planning or undergoing a migration between regions
- You want automatic discount application without managing individual reservations
- Your team is early in its cloud optimization journey and does not yet have stable, predictable usage patterns
- You want a simpler purchasing and management experience
Combining Reserved Instances and Savings Plans
RIs and Savings Plans are not mutually exclusive. They can be layered together, and Azure applies discounts in a specific priority order:
- Reserved Instances are applied first to matching resources
- Savings Plans are applied next to remaining eligible compute usage
- Pay-as-you-go rates apply to any remaining usage
This priority order means the optimal strategy for many organizations is:
- Purchase RIs for your most stable, predictable workloads (the VMs and databases you know will run unchanged for 1-3 years)
- Purchase Savings Plans to cover the remaining compute baseline — the spending you know will happen but where specific resources may change
- Leave truly variable and unpredictable workloads on PAYG
This layered approach captures the deepest discounts where certainty is highest (RIs) while still reducing costs on less predictable workloads (Savings Plans).
How to Analyze Your Usage Before Purchasing
Buying commitments without data is the most common and most expensive mistake. Before purchasing any RI or Savings Plan, analyze your actual usage patterns.
Azure Advisor Recommendations
Azure Advisor automatically analyzes your usage data and provides specific RI and Savings Plan recommendations. To access them:
- Navigate to Azure Advisor in the portal
- Select the Cost category
- Review the reservation and Savings Plan recommendations, which include estimated savings and recommended commitment amounts
Advisor bases its recommendations on your last 7, 30, or 60 days of usage data. For the most accurate recommendations, ensure you have at least 30 days of stable, representative usage before relying on these suggestions.
Azure Cost Management Analysis
For a deeper analysis, use Azure Cost Management to understand your spending patterns:
- Identify steady-state compute: Look at your compute spend over the past 3-6 months. What is the minimum baseline that has been consistent month over month?
- Find your commitment floor: The safest commitment amount is the lowest consistent compute spend you have observed. You will not over-commit if you stay at or below this level.
- Segment by stability: Categorize your workloads into stable (running the same VM consistently for 6+ months), somewhat stable (same workload but VM sizes change), and variable (unpredictable).
The Reservation Utilization Report
After purchasing RIs, monitor their utilization through the reservation utilization report in Cost Management. This report shows what percentage of your purchased reservations are being used. Target 95%+ utilization — anything below 80% indicates you may have over-committed or that the underlying workloads have changed.
Purchase Strategy: Where to Start
The question of whether to start with RIs or Savings Plans depends on your organization’s maturity and workload stability.
For Organizations New to Commitments
If you have never purchased RIs or Savings Plans, start with a conservative approach:
- Begin with Savings Plans at a conservative hourly commitment (60-70% of your current baseline compute spend)
- Run for 1-2 months and monitor utilization
- If utilization is consistently at 100%, consider adding more Savings Plan capacity or layering in RIs for your most stable workloads
- Use 1-year terms initially to limit risk while you learn your usage patterns
For Organizations with Established Workloads
If you have stable production infrastructure with 6+ months of consistent usage data:
- Start with RIs for your most predictable resources — production VMs that have not changed size or region in 6+ months, production SQL databases, Cosmos DB accounts with steady throughput
- Layer Savings Plans on top to cover the remaining compute baseline
- Consider 3-year terms for workloads where you have high confidence in long-term stability
- Leave 10-20% of compute on PAYG as a buffer for growth and changes
For Non-Compute Services
Savings Plans do not cover databases, caching, analytics, or other non-compute services. For these, RIs are the only commitment option:
- Azure SQL Database: RIs are available for vCore-based tiers (General Purpose, Business Critical, Hyperscale)
- Cosmos DB: RIs cover provisioned throughput (Request Units)
- Azure Cache for Redis: RIs are available for Premium and Enterprise tiers
- Azure Synapse Analytics: RIs cover dedicated SQL pool compute
- Azure Database for MySQL/PostgreSQL: RIs are available for vCore-based compute
If a significant portion of your cloud spend is on managed databases, RI purchases for these services can yield substantial savings that Savings Plans simply cannot address.
Common Mistakes to Avoid
Over-Committing
The most expensive mistake is purchasing more commitment than you can use. An unused reservation or an underutilized Savings Plan is money wasted with no return. Symptoms of over-commitment:
- Reservation utilization below 80%
- Savings Plan utilization consistently below 90%
- Total committed spend exceeds actual compute usage
How to avoid it: Start with a commitment level at 60-70% of your observed baseline. It is better to leave some spend at PAYG rates and add commitments incrementally than to over-commit upfront. You can always buy more. Review our guide on cloud cost optimization for additional strategies to pair with your commitment purchases.
Choosing the Wrong Scope
Setting an RI to single-subscription scope when you have multiple subscriptions means the reservation can only be used in that one subscription. If the matching resource moves or is deleted, the reservation goes unutilized.
How to avoid it: Default to shared scope at the billing account level. Only use narrower scopes if you have a specific financial reason (e.g., chargeback to a specific business unit).
Ignoring Utilization After Purchase
Many organizations purchase RIs and Savings Plans and then never look at them again. Workloads change, VMs get resized, applications move to containers. Without monitoring, commitments that were perfectly aligned at purchase can become wasteful within months.
How to avoid it: Set a monthly calendar reminder to review reservation and Savings Plan utilization in Cost Management. Configure alerts for utilization dropping below 90%.
Not Rightsizing Before Committing
Purchasing a 3-year RI for an oversized VM locks you into paying for capacity you do not need for three years. Always rightsize first, then commit.
How to avoid it: Review Azure Advisor rightsizing recommendations and act on them before making any commitment purchases. The sequence matters: rightsize first, then commit to the right-sized resource.
Overlooking the Exchange and Refund Policy
Microsoft’s reservation exchange policy changed in 2024. Exchanges for RIs purchased after January 1, 2024 are no longer permitted for most reservation types. Refunds are limited to a lifetime cap of $50,000 per billing account. Savings Plans cannot be exchanged or refunded at all.
How to avoid it: Understand the commitment is non-reversible (or nearly so). Only commit to what you are confident you will use for the full term.
Real-World Scenarios
Scenario 1: Stable Production VMs
Situation: A company runs 10 D4s_v5 VMs in East US for a production web application. These VMs have been running 24/7 for over a year with no plans to change.
Recommendation: Purchase 3-year RIs for all 10 VMs.
Why: The workloads are highly predictable, the VM size and region are stable, and the 3-year RI provides the deepest discount (up to 62% off PAYG for D4s_v5). At approximately $140/month per VM on PAYG, 10 VMs cost $1,400/month. With a 3-year RI at 62% savings, the effective cost drops to approximately $532/month — saving $10,416/year.
Scenario 2: Variable Compute with Frequent Changes
Situation: A software development company runs various VM sizes across multiple regions for CI/CD pipelines, staging environments, and microservices. Their total compute spend is relatively stable at $8,000-$10,000/month, but the specific VMs change frequently as they adopt new VM families and move workloads between regions.
Recommendation: Purchase a 1-year Compute Savings Plan at $10/hour (approximately $7,300/month, covering the lower bound of their spend).
Why: The specific VMs change too often for RIs to be practical. A Savings Plan provides 15-33% savings on the committed amount regardless of which VM series, size, or region is used. The 1-year term limits exposure while they stabilize their infrastructure. Even at a conservative 20% savings rate, that is approximately $1,460/month saved — $17,520/year.
Scenario 3: Mixed Workloads with Databases
Situation: A healthcare company runs production VMs, Azure SQL Database (Business Critical), and Cosmos DB. VM usage is moderately stable, SQL Database usage is very stable, and Cosmos DB throughput varies seasonally.
Recommendation:
- 3-year RIs for Azure SQL Database (the most stable, highest-cost non-compute service — Savings Plans cannot cover it)
- 1-year RIs for Cosmos DB at the minimum provisioned throughput level (seasonal variation makes 3-year risky)
- 1-year Compute Savings Plan for VMs at 70% of baseline compute spend (moderately stable but may change)
- PAYG for Cosmos DB throughput above the reserved baseline and any additional compute beyond the Savings Plan
Why: This layered approach matches commitment depth to workload predictability. The most stable resources get the deepest discounts. Variable capacity stays on flexible pricing.
Monitoring and Managing Your Commitments
After purchasing RIs and Savings Plans, ongoing management is critical. Set up the following:
Utilization Dashboards
Build a Cost Management view that tracks:
- RI utilization percentage (target: 95%+)
- Savings Plan utilization percentage (target: 95%+)
- Amortized cost vs. PAYG cost (shows actual savings achieved)
- Upcoming reservation expirations (plan renewals 60-90 days in advance)
For step-by-step guidance on building these views, see our Azure Cost Management dashboard setup guide.
Alerts
Configure budget alerts for:
- RI utilization dropping below 90%
- Savings Plan utilization dropping below 90%
- Total compute spend exceeding your committed amount by more than 20% (indicates you may need additional commitments)
Renewal Planning
RIs and Savings Plans do not auto-renew by default (though auto-renewal can be enabled). Set calendar reminders 90 days before expiration to:
- Review whether the commitment is still appropriate
- Evaluate whether to renew at the same level, increase, decrease, or switch between RI and Savings Plan
- Check for newer VM series that offer better price/performance (committing to an older series for another 3 years may not be optimal if a newer series offers more performance at the same price)
Choosing the Right Approach for Your Organization
There is no single correct answer to the RI vs. Savings Plan question. The right approach depends on your workload characteristics, organizational maturity, and risk tolerance:
| If you have… | Start with… |
|---|---|
| Stable, long-running VMs in fixed regions | Reserved Instances (3-year for deepest savings) |
| Managed databases (SQL DB, Cosmos DB, Redis) | Reserved Instances (only option) |
| Changing VM series/sizes but consistent total spend | Savings Plans |
| No commitment experience | Savings Plans (1-year, conservative amount) |
| Both stable and variable workloads | RIs for stable + Savings Plans for variable |
The overarching principle is this: commit to what you are confident about and leave room for uncertainty. A 90% utilized commitment is far more valuable than a 70% utilized one, even if the total committed amount is lower. Precision matters more than volume.
For organizations still evaluating which cloud platform to standardize on, our Azure vs AWS vs GCP comparison covers the broader platform decision. And if you need help building the financial governance practices around these purchasing decisions, our FinOps strategies guide provides the operational framework.
Next Steps
If you are ready to start optimizing your Azure spending with RIs and Savings Plans:
- Review Azure Advisor for immediate RI and Savings Plan recommendations based on your actual usage
- Analyze 30-90 days of usage in Cost Management to understand your compute baseline
- Rightsize first — act on rightsizing recommendations before committing to current resource sizes
- Start conservative — commit to 60-70% of your baseline and increase incrementally
- Monitor utilization monthly — treat commitments as ongoing management tasks, not one-time purchases
Cloud cost optimization is not a single decision — it is an ongoing practice. Reserved Instances and Savings Plans are powerful tools, but they deliver value only when they are right-sized to your actual usage and actively managed over time.