Aiden Tran

Jan 12, 2026

6 min read

AWS EC2 Instance Purchasing Options: The Art of Negotiating Price

When getting started with EC2, many people only know about On-Demand Instances — the simplest approach of “spin up a machine and pay.” But in reality, AWS offers up to 6 different purchasing options, each with its own pros and cons suited to specific scenarios.

Understanding these options not only helps you save significant costs (up to 90%) but is also essential knowledge for passing the AWS Certified Solutions Architect Associate (SAA) exam.

Let’s explore the “art of negotiating price” in the EC2 world!

1. On-Demand Instances

This is the most basic usage model, like walking into a hotel and paying the listed room rate.

How It Works

Billing: Pay per second (Linux/Windows) or per hour (other OSes) for what you actually use.
Billing details:
- For Linux or Windows: Billing per second, after the first minute.
- For other OSes: Billing per hour.

Advantages

No long-term commitment, no upfront payment.
Maximum flexibility — start/stop anytime.
No risk of sudden interruption.

Disadvantages

Highest cost among all instance purchasing options.

Best Suited For

Short-term workloads that cannot be interrupted.
Newly deployed applications with unpredictable load patterns.
Development and testing environments.
Applications with unpredictable, variable traffic.

2. Reserved Instances (RI)

Like signing a 1 or 3-year hotel room lease to get a big discount.

How It Works

Commitment: You commit to using a specific instance type for 1 year or 3 years.
Discount: Up to 72% off On-Demand pricing.

Payment Options

AWS offers 3 payment methods, and the more you pay upfront, the deeper the discount:

All Upfront: Highest discount.
Partial Upfront: Balance between discount and cash flow.
No Upfront: Most flexible but lowest discount.

Types of Reserved Instances

Standard RI

Characteristics: Highest discount (up to 72%).
Limitation: Less flexible — cannot change instance type, OS, or tenancy after purchase.
Best for: Applications with fixed requirements that won’t change over time.

Convertible RI

Characteristics: Allows changing instance type, OS, and tenancy throughout the commitment period.
Discount: Lower than Standard RI (around 66%).
Best for: Applications that may need configuration changes in the future but still want cost savings.

Best Suited For

Applications running steadily and continuously (steady-state workloads).
Database servers.
Web servers with stable traffic.
Any application you’re certain will run for at least 1-3 years.

3. Savings Plans

A more modern model than RI, focusing on spending amount rather than specific server types.

How It Works

Commitment: You commit to spending a certain amount (e.g., $10/hour) for 1 year or 3 years.
Scope: Applies to any instance in the same Region, as long as it falls within the committed spend.

Key Advantages

Extremely flexible: You can:
- Change instance size (e.g., from m5.xlarge to m5.2xlarge).
- Change OS (e.g., from Linux to Windows).
- Change instance family (e.g., from General Purpose to Compute Optimized).
- Change Region (with some conditions).
Discount: Comparable to RI (around 72%).
Auto-applied: AWS automatically applies the Savings Plan to eligible instances — no manual management needed.

Types of Savings Plans

EC2 Instance Savings Plans

Applies to EC2 Instances.
Flexible across instance family, size, OS, tenancy, and Region.

Compute Savings Plans

Applies more broadly: EC2, Lambda, Fargate.
Most flexible — can switch between services.

Best Suited For

Systems that need infrastructure flexibility but have stable long-term spend.
Organizations with multiple teams using EC2 with different requirements.
When you want to save costs but aren’t sure about the specific instance configuration in the future.

4. Spot Instances

This is about leveraging AWS’s unused capacity at “clearance” prices — one of the most effective cost-saving methods.

How It Works

Bidding: Price is determined based on supply and demand of available resources in AWS.
Dynamic pricing: Price can change in real time, but you only pay the price at the time the instance is running.
Discount: Up to 90% off On-Demand (the cheapest of all options).
Max Price: You set a maximum price you’re willing to pay. If the Spot price exceeds your max price, the instance will be terminated.

Important Risks

Spot Instance Interruption: AWS can reclaim the instance at any time if:
- An On-Demand or Reserved Instance customer needs those resources.
- The Spot price exceeds your max price.
Advance notice: AWS will warn you 2 minutes before reclaiming the instance (Spot Instance Interruption Notice).

Spot Fleet

Definition: A Spot Fleet is a set of Spot Instances, optionally combined with On-Demand Instances.
Advantage: Automatically distributes workload across multiple instance types and Availability Zones to optimize cost and reduce interruption risk.
Allocation Strategies: LowestPrice (cost optimization), Diversified (risk reduction), CapacityOptimized (capacity optimization).

Important Notes

Spot Block is no longer supported by AWS: The Spot Block strategy (which allowed users to lock a Spot Instance for 1-6 hours without interruption) is no longer supported. Keep this in mind when designing your architecture.

Not suitable for Stateful Applications: Spot Instances are not suitable for stateful applications like database primary nodes, as interruptions can cause data loss.

Best Suited For

Fault-tolerant workloads.
Batch processing jobs.
Big data analytics.
Testing and development environments.
Containerized workloads.
CI/CD pipelines.
Machine Learning training and inference.
Web crawling and data scraping.
Any workload that can recover from a checkpoint.

5. Dedicated Hosts & Dedicated Instances

For special hardware and security requirements.

Dedicated Host

Definition: You rent an entire physical server from AWS.
Control:
- Control over instance placement.
- Deep control over the physical server.
Most important feature: Supports software licenses based on core or socket counts — Bring Your Own License (BYOL).
Cost: Most expensive among all instance purchasing options.

Dedicated Instance

Definition: An instance running on hardware dedicated to you.
Difference from Dedicated Host:
- You don’t have deep control over the physical server like with Dedicated Host.
- AWS manages placement for you.
Guarantee: Isolation from other customers on the same hardware.

Best Suited For

Dedicated Host:
- Applications requiring per-socket/per-core software licensing (e.g., SQL Server, Windows Server with Software Assurance).
- Strict compliance requirements.
- Need for full placement control.
Dedicated Instance:
- Isolation requirements from other customers.
- Compliance requirements without needing placement control.
- Applications that don’t need BYOL but require dedicated hardware.

6. Capacity Reservations

Not a discount model, but a reservation model.

How It Works

Purpose: You reserve a quantity of instances in a specific Availability Zone (AZ) to ensure capacity is available when you need it.
Cost: Charged at On-Demand rates whether you use the instances or not.
No discount: This is not a cost-saving tool — it’s a capacity assurance tool.

Important Notes

Combined usage: Typically used in combination with RI or Savings Plans to get both reservation and discount benefits.
Regional vs Zonal:
- Zonal Capacity Reservation: Reserved for a specific AZ.
- Regional Capacity Reservation: More flexible, can be used in any AZ within the Region.

Best Suited For

Mission-critical applications that need guaranteed capacity at a specific time.
Compliance requirements demanding capacity guarantees.
Combined with RI/Savings Plans to optimize both cost and capacity.

Quick Comparison Summary

Option	Discount	Term	Key Characteristic	Best Suited For
On-Demand	0%	None	Most flexible, pay as you go	New apps, unpredictable load
Reserved Instances	Up to 72%	1 or 3 years	Best for steady-state workloads	Databases, stable web servers
Savings Plans	Up to 72%	1 or 3 years	Flexible config, commit by dollar amount ($)	Systems needing flexibility with stable spend
Spot Instances	Up to 90%	None	Cheapest, but can be interrupted anytime	Fault-tolerant, batch jobs, testing
Dedicated Host	Low (can be more expensive)	Optional	Rent entire physical server, for special licensing	BYOL requirements, strict compliance
Dedicated Instance	Low	Optional	Instance on dedicated hardware	Need isolation, no BYOL needed
Capacity Reservations	0%	Optional	Reservation only, ensures capacity when needed	Mission-critical, combined with RI/SP

Hybrid Strategy (Combined Approach)

In practice, you don’t need to pick just one type. AWS encourages using a combined strategy to optimize both cost and performance:

Typical Strategy Example

Base workload (60-70%): Reserved Instances or Savings Plans for stable, predictable applications.
Variable workload (20-30%): On-Demand Instances for traffic spikes and newly deployed applications.
Fault-tolerant workload (10-20%): Spot Instances for batch jobs, data processing, and testing.