EC2 Pricing Options

On-Demand, Savings Plans, Reserved Instances, Spot, Dedicated Hosts and Capacity Reservations - and when each fits.

What is it?

The same instance can be bought several ways. You trade commitment and flexibility for price.

  • On-Demand: pay per second (or hour, depending on OS) with no commitment. Highest unit price, maximum flexibility. Ideal for unpredictable or short-term workloads.
  • Savings Plans: commit to a consistent amount of usage (measured in dollars per hour) for one or three years in exchange for lower rates. Compute Savings Plans are the most flexible, applying across instance families, Regions, and even Fargate and Lambda; EC2 Instance Savings Plans give deeper discounts for a chosen family in a Region.
  • Reserved Instances (RIs): an older billing discount tied to specific instance attributes for one or three years, with Standard and Convertible flavors. Often used for steady-state databases and servers.
  • Spot Instances: spare capacity at a steep discount, but AWS can reclaim it with a short warning. Perfect for fault-tolerant, interruptible work.
  • Dedicated Hosts: a whole physical server for you, useful for per-socket or per-core licensing and compliance. Dedicated Instances run on hardware dedicated to you but with less visibility into the host.
  • Capacity Reservations: reserve capacity in a specific AZ for any duration, without a billing discount by themselves (combine with Savings Plans).

Payment options for commitments typically include all upfront, partial upfront, and no upfront; paying more earlier gives a bigger discount.

Reservation details: Standard Reserved Instances give the biggest discount for a fixed instance family; Convertible RIs let you change family, OS or tenancy for a somewhat smaller discount; both come in 1- or 3-year terms with no, partial or all upfront payment. Reserved Instances and Savings Plans discounts can be shared across the accounts in an AWS Organization with consolidated billing, and unused RI capacity can be sold on the Reserved Instance Marketplace (Standard only). Compute Savings Plans are the most flexible, applying across instance families, Regions, Fargate and Lambda.

Tenancy: Dedicated Instances run on hardware used only by your account but AWS picks the host and you pay a per-instance premium; Dedicated Hosts give a whole host you control, which matters for server-bound licenses. Capacity Reservations guarantee capacity in an AZ and can be combined with Savings Plans or RIs for the discount.

Explain like I'm 10

Think of a gym. Dropping in each time costs the most but needs no promise (On-Demand). A one- or three-year membership is cheaper per visit if you really go (Savings Plans/RIs). Off-peak standby slots are nearly free but the trainer may ask you to leave (Spot). A private studio only you can enter is the Dedicated Host.

Examples

Choosing a model

Workload                                   Best fit
-----------------------------------------  -----------------------------
Dev server used a few hours, then deleted  On-Demand
Always-on production API for 3 years       Savings Plan / Reserved
Nightly image-processing jobs, retry-safe  Spot
Database licensed per physical core        Dedicated Host
Must have capacity in one AZ for an event  Capacity Reservation
Unknown, changing compute mix              Compute Savings Plan

Requesting Spot capacity

aws ec2 run-instances \
  --image-id ami-0123456789abcdef0 \
  --instance-type c6i.large \
  --instance-market-options 'MarketType=spot,SpotOptions={SpotInstanceType=one-time}'

# Inspect recent Spot price history
aws ec2 describe-spot-price-history \
  --instance-types c6i.large \
  --product-descriptions "Linux/UNIX" --max-items 3

Spot suits work that can checkpoint and restart. Do not run a single critical server on it.

How it works

AWS bills each running instance by its usage. Commitment discounts (Savings Plans, RIs) are applied automatically to matching usage on your bill; you do not 'assign' them to a particular server. Spot capacity is auctioned from spare pools; when AWS needs the capacity back it sends a two-minute interruption notice before reclaiming it.

A sensible strategy mixes models: commit to your steady baseline, use On-Demand for the unpredictable layer on top, and use Spot for flexible background work.

  Usage over time
  |            ____ On-Demand (spikes)
  |       ____|    |____
  |  ____|              |____    Spot (batch work)
  | [==== Savings Plan / Reserved baseline ====]
  +---------------------------------------------> time

Why does it exist?

AWS has spare capacity at some moments and needs predictable revenue at others. Discounts reward customers who give predictability, while On-Demand keeps the door open for those who cannot. It lets different workloads pay fairly for what they need.

When to use it

Commit after you have watched real usage for a while and know your baseline. Use Spot for stateless, interruptible, or batch jobs. Use Dedicated Hosts when licenses or regulations require physical server visibility.

When not to use it

Do not buy a three-year commitment for a project that may be cancelled. Do not run databases or single-instance critical services on Spot. Do not choose Dedicated Hosts without a licensing or compliance reason - they are expensive.

Common mistakes

  • Committing before understanding actual usage, then paying for idle commitments.

  • Running stateful services on Spot without handling interruptions.

  • Believing a Capacity Reservation gives a discount on its own.

  • Mixing up Dedicated Hosts (a whole physical server) and Dedicated Instances (instances on single-tenant hardware).

  • Forgetting that stopped instances still incur EBS storage charges.

  • Confusing Dedicated Instances (isolated hardware, no host control) with Dedicated Hosts (host-level visibility and license reuse).

Practice exercises

  1. Easy:

    Match each scenario to a pricing option: a weekend load test, a 24/7 production DB, a fault-tolerant video transcoder.

  2. Medium:

    Explain how Compute Savings Plans differ from EC2 Instance Savings Plans in flexibility and discount.

  3. Medium:

    Design a purchasing mix for a web app with a steady baseline of 10 servers, daily peaks of 25, and a nightly batch job.

  4. Hard:

    Spot capacity is interrupted with short notice. Describe an architecture (queue, checkpoints, auto scaling) that tolerates this safely.

Interview questions

When would you use Spot Instances?

For fault-tolerant, flexible, or batch workloads that can handle interruption, such as rendering, big data processing, and CI jobs.

Why might a company choose Dedicated Hosts?

To use existing per-socket or per-core software licenses or to meet compliance requirements that need a physical server dedicated to them.

What is the difference between Savings Plans and On-Demand?

Savings Plans trade a one- or three-year spend commitment for a lower rate; On-Demand has no commitment and the highest rate.

Exam-style: A steady-state application will run for three years. Which option is the most cost-effective? (A) On-Demand (B) Spot (C) Savings Plans or Reserved Instances (D) Dedicated Hosts

C.

Exam-style: Which option offers the largest discount but may be interrupted?

Spot Instances.

Which Reserved Instance type allows changing instance family during the term?

Convertible Reserved Instances.

Exam-style: Which option guarantees EC2 capacity in an AZ without a long-term commitment?

On-Demand Capacity Reservations.