alphaus cloud logo

Reserved Instances

Reserved Instances (RIs) are a cloud pricing model where users commit to long-term usage (1-3 years) for discounted rates, but specifics vary by provider: AWS requires fixed instance types/regions for up to 72% discounts.

Reserved Instances (Commitment-Based Discounts)

Overview

Reserved Instances (RIs) are a cloud pricing model where users commit to long-term usage (1-3 years) for discounted rates, but specifics vary by provider: AWS requires fixed instance types/regions for up to 72% discounts; Azure’s Reservations allow flexibility within VM families (e.g., resizing instances) for similar discounts; Google Cloud’s Committed Use Discounts apply to vCPU/memory resources rather than instance types, offering ~57% discounts with broader flexibility. While the core concept—committing for savings vs. on-demand pricing—is consistent, AWS is more rigid, whereas Azure and Google Cloud prioritize adaptability in their reservation models.

Brief History

AWS pioneered the concept of Reserved Instances (RIs) in 2009, offering customers up to 72% savings in exchange for 1 or 3-year commitments to specific EC2 instances. Over time , AWS introduced flexibility with Convertible RIs (2016), instance size flexibility (2017), and the RI Marketplace (2012) resale. Google Cloud initially took a different approach which led to the launching of Sustained Use Discounts (2014) that automatically reduced pricing based on monthly usage, eliminating upfront commitments. 

In 2017, Google introduced Committed Use Discounts (CUDs), allowing customers to commit to a certain amount of vCPU and RAM for deeper savings while retaining instance-type flexibility. 

Microsoft Azure entered the space later, launching Azure Reserved VM Instances (2017) —now called Reservations in the current branding — with built-in options for exchanges and cancellations, offering more management flexibility than AWS’s initial RIs.

Key Features and How They Work

Regardless whether you are using AWS, Azure, or Google Cloud, Reserved Instances (or equivalent commitment-base pricing models) share  several common features:

Long-Term Commitment for Cost Savings

  • Customers commit to using cloud resources for a fixed term (typically 1 or 3 years) in exchange for significant discounts (up to 72% off compared to On-Demand pricing), but in  Google Cloud’s case up to maximum of 57%..

Upfront or Periodic Payments

  • Different payment options are available: All Upfront (largest savings), Partial Upfront, or No Upfront (smaller savings, but paid monthly). Some providers, like Azure which allows monthly payments with no extra fees. For Google Cloud, however, there are no upfront payments —discounts apply automatically to usage over the commitment term.

Instance or Resource-Based Commitment

Cloud providers have introduced flexibility over time:

  • AWS offers two flexible RI models: (1) Convertible RIs, which allow exchanges for different instance types/families (with value constraints) at lower discounts; and (2) Regional RIs, which automatically apply to any size within the same instance family in a region. While Convertible RIs enable broader configuration changes, Regional RIs provide seamless size flexibility without manual intervention.
  • Azure: Azure allows unlimited penalty-free exchanges for reservations of equal or greater value (e.g., changing VM types, regions, or terms), but cancellations incur a 12% early termination fee on the remaining commitment balance, followed by a prorated refund of the adjusted amount. 
  • Google Cloud: Google Cloud’s Committed Use Discounts require committing to vCPU/memory resources (not instance types) for 1 or 3 years, with no cancellations. While you cannot reduce the commitment itself, lowering usage avoids overpaying for unused resources. Discounts apply automatically across all projects under the billing account.

Automatic vs. Manual Cost Savings

  • Google Cloud provides Sustained Use Discounts that automatically apply when using resources for longer periods, without requiring a commitment.
  • AWS and Azure require manual RI/Reservation purchases. Azure, however, has no equivalent automatic discount model.

Reserved Capacity vs. Pure Discount

  • AWS Standard Reserved Instances (RIs) provide discounts (up to 72%) without capacity guarantees, while Capacity Reservations—a separate product—ensure resource availability in specific Availability Zones but incur additional On-Demand costs. 
  • Similarly, Azure Reservations and Google Cloud’s Committed Use Discounts (CUDs) are purely discount-focused models (up to 72% for Azure, ~57% for Google) with no capacity guarantees. For capacity assurance, Azure offers Reserved Capacity and Google Cloud provides Compute Engine Reservations, both requiring separate purchases and costs. 
  • Across all providers, cost-saving commitments (RIs/Reservations/CUDs) are distinct from capacity guarantees, which demand extra expenditure and tailored products.

Types of Reserved Instances

Type Commitment Flexibility Discount Range Cloud Providers
Standard Reserved Instances 1-3 years Fixed instance type/region (AWS); VM family (Azure) (~72%) AWS, Azure
Convertible/Flexible RIs 1-3 years Can change instance type, OS, region, family (~54%) AWS
Resource/Spend-Based Commitments 1-3 years AWS/Azure Savings Plans: Apply to usage dynamically
Google CUDs: vCPU/memory (any instance type)
AWS/Azure:(~72%)
Google Cloud: (~57%)
AWS, Azure, Google Cloud
Sustained Use Discounts None (automatic) Fully flexible (~30%) Google Cloud
Scheduled Reserved Instances (deprecated) Depreciated Fixed schedule, (discontinued in 2022) (~5-10%) AWS (discontinued)

Benefits and Drawbacks

Aspect Benefits Drawbacks
Cost Savings Up to 72% discount (AWS/Azure), ~57% (Google CUDs) Requires long-term commitment (1-3 years); overcommitment risks paying unused resources.
Budgeting Predictable pricing with lock-in rates Overprovisioning wastes spend; careful forecasting required.
Capacity Reservation AWS Capacity Reservations (separate product) guarantee resources Standard RIs (AWS/Azure) and CUDs (Google Cloud) do not reserve capacity.
Payment Options Options: All Upfront, Partial, No Upfront (AWS/Azure); No Upfront, discounts accrue (Google Cloud) Highest savings require All Upfront payments (AWS/Azure)
Flexibility AWS Convertible RIs, Azure Exchangeable RIs Limited changes allowed in Standard RIs; Google CUDs cannot be cancelled/modified.
Best for Stable, predictable workloads Not ideal for variable or unpredictable workloads

Reserved Instances are a great cost-saving tool for predictable workloads, but they require careful planning to avoid overcommitting. If workloads are dynamic, more flexible alternatives like Savings Plans or Sustained Use Discounts might be better options. 

Conclusion

Reserved Instances (RIs) remain a cornerstone of cloud cost optimization, offering discounts of up to 72% (AWS/Azure) or 57% (Google Cloud) for predictable workloads. However, their long-term commitments, rigid instance requirements (AWS), and lack of built-in capacity guarantees demand meticulous planning to avoid overprovisioning. Providers like Azure and Google Cloud have introduced flexibility through exchanges (Azure) or resource-based commitments (Google CUDs), while AWS offers Convertible/Regional RIs for limited adaptability. For dynamic workloads, spend-based models (Savings Plans) or automated discounts (Google’s Sustained Use) provide better agility. 

To maximize savings without the complexity of manual RI management, Octo by Alphaus streamlines cost optimization by automating Reserved Instance purchases, tracking usage patterns, and recommending flexible commitments tailored to your cloud environment. Visit Octo to simplify your cloud cost strategy and turn reserved capacity into predictable savings. 

Simplified Cloud Cost Management by Alphaus

Learn how we help over 3000+ users, companies and enterprises to visualize, understand and optimize their cloud costs.
Learn More