EC2 purchase options compared: on-demand, Spot, Savings Plans, Reserved

Quick answer

EC2 has four main purchase options. On-demand pays full price per hour with no commitment or risk, best for unpredictable, short-lived, or spiky needs. Spot runs on spare capacity for up to 90 percent off but can be reclaimed with a two-minute warning, best for interruptible, fault-tolerant work. Savings Plans give up to about 72 percent off in exchange for a 1 or 3 year dollars-per-hour commitment, best for a steady baseline, with Compute Savings Plans more flexible and EC2 Instance Savings Plans deeper. Reserved Instances offer similar discounts with capacity reservation for specific instance types. The cheapest fleet layers them: commit the baseline, run elastic work on Spot, and leave a thin on-demand buffer.

EC2 gives you four ways to pay for the same compute, and the difference between using them well and running everything on-demand is enormous, often 50 to 80 percent of the bill. Each option trades discount against commitment and risk. The skill is not picking one, it is matching each part of a workload to the option built for it.

The four options at a glance

Option	Discount	Commitment	Risk	Best for
On-demand	None (baseline)	None	None	Unpredictable, short-lived
Spot	Up to ~90%	None	Interruption	Fault-tolerant, elastic
Savings Plans	Up to ~72%	1 or 3 yr spend	Over-commit	Steady baseline
Reserved Instances	Up to ~72%	1 or 3 yr per type	Over-commit, less flexible	Steady, fixed instance types

On-demand: flexibility at full price

On-demand charges the standard hourly rate with no commitment and no interruption risk. It is the most expensive per hour, so it should cover only what the other options cannot: unpredictable spikes beyond your Spot capacity, brand-new workloads whose steady level is unknown, and short-lived needs not worth committing to. On-demand is the buffer, not the base.

Spot: the deepest discount, with a catch

Spot sells spare capacity for up to 90 percent off, with AWS able to reclaim it on a two-minute warning. It needs no commitment, so it is ideal for elastic and interruptible work: batch jobs, stateless services, workers, CI, and the scale-out portion of a fleet. Diversifying across instance types and Availability Zones keeps interruptions absorbable, the approach in the Spot diversification guide. Spot is cheapest of all for work that can tolerate being stopped.

Savings Plans and Reserved Instances: commit the baseline

Both give up to about 72 percent off for a 1 or 3 year commitment, ideal for the steady baseline that runs around the clock. Savings Plans commit to a dollars-per-hour amount and apply flexibly across instance families (Compute Savings Plans) or within a family for a deeper discount (EC2 Instance Savings Plans). Reserved Instances commit to specific instance types and can reserve capacity, but are less flexible if your instance mix changes. For most teams Savings Plans win on flexibility, the comparison in the Savings Plans versus Reserved guide. The risk with both is over-committing past the baseline, which drops utilization.

Layering them into one fleet

The lowest-cost architecture uses all four in layers. Commit Savings Plans (or Reserved Instances) to the always-on baseline for a guaranteed discount with no risk. Run the elastic, interruptible capacity on Spot for the deepest discount. Leave a thin on-demand layer for unpredictable spikes and short-lived needs. The result is that almost none of the fleet pays full on-demand price, the combined strategy that beats any single option.

Choosing on cost

Separate your workload into baseline (steady), elastic (variable and interruptible), and spike (unpredictable). Commit the baseline, Spot the elastic, and on-demand the spike, then reassess as the baseline grows. Compare Compute versus EC2 Instance Savings Plans on flexibility versus depth, and ladder commitments over time. Price each layer against the resource catalog so the blended rate across all four options is a concrete number before you deploy.

FAQ

What are the EC2 purchase options?

Four main ones: on-demand (full price, no commitment or risk, for unpredictable or short-lived needs), Spot (up to 90 percent off spare capacity that can be reclaimed on a two-minute warning, for interruptible work), Savings Plans (up to about 72 percent off for a 1 or 3 year dollars-per-hour commitment, for a steady baseline), and Reserved Instances (similar discount with capacity reservation for specific instance types). The cheapest fleets layer all four.

Which EC2 purchase option is cheapest?

Spot is cheapest per hour, up to 90 percent off, but only for interruptible, fault-tolerant work because AWS can reclaim it. For steady, always-on capacity that cannot be interrupted, Savings Plans or Reserved Instances are cheapest at up to about 72 percent off. On-demand is the most expensive and should cover only unpredictable spikes. No single option is cheapest for everything; matching each to the right workload is what minimizes total cost.

What is the difference between Savings Plans and Reserved Instances?

Both give up to about 72 percent off for a 1 or 3 year commitment. Savings Plans commit to a dollars-per-hour spend and apply flexibly across instance families (Compute Savings Plans) or within a family for a deeper discount (EC2 Instance Savings Plans). Reserved Instances commit to specific instance types and can reserve capacity but are less flexible if your instance mix changes. For most teams, Savings Plans win on flexibility.

How do I combine EC2 purchase options?

Layer them by workload type. Commit Savings Plans or Reserved Instances to the always-on baseline for a guaranteed discount with no interruption risk. Run the elastic, interruptible capacity on Spot for the deepest discount. Leave a thin on-demand layer for unpredictable spikes and short-lived needs. This ensures almost none of the fleet pays full on-demand price, beating any single option used alone.

How does C3X help choose EC2 purchase options?

C3X prices your EC2 compute from Terraform against a live catalog, so you can see the cost of the baseline, elastic, and spike portions of a workload before you deploy. That helps you size Savings Plan commitments to the baseline, identify capacity that should run on Spot, and understand the blended rate across all four purchase options as a concrete number in the pull request, rather than defaulting everything to on-demand.

What to do next

Match every workload to the right purchase option. C3X reads your Terraform and prices your resources against a live catalog. Start with the quickstart.