Reducing cloud costs is this year’s top priority for organizations, according to this year’s State of FinOps. And with more than 50 percent of AWS cost coming from the EC2 computing service, it’s a prime target for cost reduction. 

There are three primary options to reduce and optimize EC2 costs: 

  • Using a different service/resource plan (usage optimization)
  • Committing to a financial plan for a discount (rate optimization)
  • Reducing the size of instances in use (rightsizing)

While all three strategies play a role, many teams struggle with the sequencing—especially when it comes to rate optimization vs. rightsizing. If you commit to a 1- or 3-year plan before verifying that your instances are right-sized, you risk locking in inefficient spend. On the other hand, delaying commitment may mean missing out on valuable discounts.

So what should you do first—reserve or rightsize?

In this blog, we’ll explore both approaches, weighing the pros and cons of each. Let’s break it down.

What is Rightsizing?

Rightsizing is the process of analyzing your EC2 instances to see if they’re too large or too small for the workload they’re running. It involves checking actual usage metrics—like average CPU and memory—and deciding whether to downsize, upscale, or switch instance families. For example, if an m5.2xlarge regularly uses only 15% of its CPU, moving to an m5.large could save money without performance loss. Rightsizing can also uncover idle instances that can be shut off. It’s typically done using cost and usage data, CloudWatch metrics, or tools like AWS Compute Optimizer.

What is Reserving?

Reserving EC2 means committing to use specific compute resources over a 1- or 3-year period in exchange for a discounted hourly rate. This can be done through Reserved Instances, which lock in specific instance types and regions, or Savings Plans, which are based on committing to a certain amount of dollar per hour spend. It’s simplest to do with workloads that run consistently over time. For example, a Reserved Instance for an m5.large in us-east-1 can offer up to 70% off the On-Demand price. You have to pay for the commitment whether you use the capacity or not, so accurate forecasting is key.

Reserving First, Rightsizing Next

The strategy of reserving, then rightsizing can create some challenges. Why?

If you purchase reservations for all consistently running instances, it’s likely that your actual usage will not match the commitment. For example, if you reserve 20 r6g.4xlarge instances but only need 15, the unused capacity still incurs charges—resulting in unnecessary spend. Without accurate usage data, overcommitting can lead to long-term inefficiencies instead of savings.

View underutilized commitments in nOps

This leaves organizations with a difficult decision: if you don’t rightsize the 5 machines, you’ll overpay for the instances. If you do rightsize the EC2 instances, you’ll overpay on their commitment. Neither way is optimal.

Partial Solution: Size Flexibility

One way to mitigate overcommitment is by using regional Reserved Instances with size flexibility. This allows the reservation discount to apply across any instance size within the same family—for example, from r6g.4xlarge down to r6g.2xlarge. While you’re still financially committed to a specific amount of capacity, size flexibility helps absorb changes in instance sizing needs without wasting the reservation entirely. It doesn’t eliminate the risk of overcommitment but can reduce its impact.

Partial Solution: Convertible RIs  

Convertible Reserved Instances offer more flexibility than Standard RIs, allowing organizations to exchange their existing reservations for different instance types, families, or regions as needs evolve. They provide more control than Savings Plans and offer better discounts than standard Savings Plans—though typically lower than Standard RIs. In theory, this makes CRIs a strong tool for adapting to changing workloads. In practice, managing exchanges manually is complex and time-consuming, often requiring ongoing analysis and forecasting. Without automation, the operational overhead will likely outweigh the benefit.

Rightsizing First, Reserving Next

In theory, rightsizing before reserving makes sense — it ensures you commit only to what you actually need. But according to AWS, you should only downsize your instances if your maximum CPU and memory usage is less than 40% over a four-week period, which can take at least four weeks to determine. 

This means that in practice, rightsizing initiatives can take six months or more—leaving you stuck paying On-Demand prices and missing out on reservation savings in the meantime. This creates a dilemma: delay reservations and lose potential savings, or commit early and risk locking in inefficiencies.

Related Content

The Definitive Guide to AWS Rightsizing

Cut cloud waste and pay only for what you need
Download Now

Partial Solution: Selling on the Marketplace

One workaround is to purchase reservations during the rightsizing period and sell unused ones later on the AWS Marketplace. This lets you capture savings upfront while still adjusting your environment. However, not all reservations are eligible for resale—only Standard RIs in specific configurations and regions can be sold. Even when eligible, managing Marketplace transactions at scale is complex, especially with staggered purchase times and mixed instance types. For many organizations, it’s not a viable long-term strategy without automation.

Key Difference Between Rightsizing and Reserving

Let’s summarize the main differences between rightsizing vs reserving in a quick table. 

Metric Rightsizing Reserving
Goal Eliminate waste by matching instance size to usage Reduce cost by committing to long-term usage
Focus Usage efficiency Pricing discount
Timing Done before committing to instances Done after usage is stable and predictable
Commitment No commitment; changes can be made anytime 1- or 3-year financial commitment
Cost Saves money by reducing overprovisioning Saves money through discounted hourly rates

When to Use Which: Rightsizing vs. Reserving?

The most effective approach isn’t choosing one or the other—it’s sequencing them intelligently. Start with a rapid baseline rightsizing pass using available metrics and tools to eliminate obvious waste. This gives you a clearer picture of steady-state usage without delaying too long. Then, reserve capacity for the workloads that remain stable post-rightsizing.

Avoid overcommitting early, but don’t wait for a perfect environment either—rightsizing is ongoing. Consider partial reservations or automation tools that can dynamically adjust commitments over time. The goal is to strike a balance: reduce waste first, then lock in savings where usage is predictable.

Manage Both Rightsizing and Reserving with nOps

Rightsizing and commitment management work best together—but doing both manually is time-consuming, error-prone, and nearly impossible to scale. That’s where nOps comes in.

nOps automates rightsizing based on real usage patterns, and continuously manages Reserved Instances and Savings Plans to align with your evolving workloads. We offer a 100% utilization guarantee for your commitments — so you never have to worry about underutilization again. 

Here are the benefits of letting nOps manage your cloud optimization:

  • More Cost Savings: Intelligently analyzes usage patterns and selects the optimal rightsizing and commitment strategies based on workload changes, ensuring you’re always leveraging the best pricing model for your needs.
  • Less Management Overhead: By integrating with AWS APIs, nOps fully automates the rightsizing and commitment process—reducing human error, guesswork, and the time spent juggling spreadsheets.
  • 100% Visibility: Understand exactly how much you’re saving with nOps, with real-time insights into both cost avoidance from rightsizing and realized RI/SP savings across your environment.

nOps was built to make it easy for engineers to save on cloud costs, so they can focus on building and innovating. We manage $2B in cloud spend and are rated 5 stars on G2 — sign up to get started for free or book a personalized demo with one of our cloud optimization experts.   

Frequently Asked Questions

When it comes to rightsizing and reserved instances, here are the most commonly asked questions:

What are the benefits of right sizing?

Right sizing helps cut cloud waste by matching resource capacity to actual usage. You eliminate oversized instances, underused storage, or idle services—resulting in lower bills without impacting performance. It also improves system efficiency and stability by reducing overprovisioned resources that drive up cost or underperforming resources that lead to bottlenecks. When done regularly, right sizing keeps your cloud environment cost-effective and production-ready.

What is rightsizing in cloud cost optimization?

AWS rightsizing means analyzing actual resource usage and adjusting the instance type, size, or configuration to better align with real demand. For example, switching an EC2 m5.4xlarge to m5.2xlarge if CPU usage is consistently low. It applies to compute, memory, storage, and even container workloads. In cloud cost optimization, rightsizing is one of the fastest ways to reduce costs while maintaining performance and availability.

What is the difference between rightsizing and downsizing?

Rightsizing is about matching resources to actual need—scaling up or down based on usage. Downsizing implies only scaling down, often reactively or for cost-cutting without considering performance impact. When it comes to rightsizing vs downsizing, rightsizing is data-driven and balanced; downsizing can risk underprovisioning if not carefully managed. Rightsizing also includes scaling up when usage demands it—so the goal is efficiency, not just reduction.

What are Azure Reservations?

Azure Reservations let you prepay for virtual machines, SQL databases, and other services over a one- or three-year term in exchange for a discount—up to 72% compared to pay-as-you-go rates. They are similar to AWS Reserved Instances. Azure Reservtions are ideal for predictable, steady workloads. Reservations apply to specific scopes (subscription or shared) and can be exchanged or refunded (with limits). Unlike auto-scaling savings plans, reservations require more manual planning and upfront commitment.

What are right sizing resources?

Right sizing resources means adjusting cloud configurations—instance type, size, storage class, or container specs—based on observed usage patterns. For compute, it could mean reducing vCPU or RAM; for storage, switching to a lower-cost tier. Tools like nOps or AWS Compute Optimizer recommend rightsizing actions using telemetry data. Then once you rightsize, nOps helps you reserve the right instances so you can save on both usage and pricing.