AWS compute is sold through five pricing models, each suited to a different demand shape. On demand is the flexible full price baseline you pay with no commitment. Spot is deeply discounted spare capacity that AWS can reclaim, so it fits interruptible work only. Savings Plans give a flexible commitment discount across instance families and regions, while Reserved Instances give a specific commitment discount for a fixed configuration. The Enterprise Discount Program layers an account wide discount tier on top in exchange for a multi year spend commitment. There is no single cheapest model: the cheapest is the one matched to the workload. The skill is layering them so commitments cover only the steady demand you can defend, Spot carries the interruptible work, and on demand absorbs the variable peak.
Here is what each model is, when it wins, and how they combine into one purchasing strategy.
What are the five models?
| Model | What it is | When it wins |
|---|---|---|
| On demand | Full price, no commitment, pay per second or hour | Spiky or unpredictable demand, and the variable peak above your committed floor |
| Spot | Spare capacity at a steep discount, reclaimable with short notice | Fault tolerant, interruptible work such as batch, CI, and stateless processing |
| Savings Plans | Flexible commitment discount across families, sizes, regions, and EC2, Fargate, Lambda | The steady compute floor when the fleet mix may shift |
| Reserved Instances | Specific commitment discount for a fixed family and region | Stable, specific workloads such as a long lived database |
| Enterprise Discount Program | Account wide discount tier for a multi year spend commitment | Large, predictable total spend where you can commit a forecast volume |
Savings Plans and Reserved Instances discount roughly 20 to 72 percent against on demand in exchange for utilization risk the buyer carries; Spot can discount more but with reclaim risk. All percentages are indicative and depend on term, payment option, instance family, and region, so verify against current AWS pricing.
Why is there no single cheapest model?
Because each discount is paid for with a constraint. Spot is cheapest per hour but only works if your workload tolerates being interrupted. Reserved Instances are deep but only if the configuration never changes. Savings Plans are flexible but slightly less deep than an equivalent Reserved Instance. On demand has no discount but no constraint either. The cost question is therefore not which model is cheapest in the abstract but which constraint your workload can absorb. A batch pipeline that can restart should run on Spot; a steady web tier whose mix evolves belongs under Savings Plans; a fixed database fits a Reserved Instance. Matching the constraint to the workload is the entire skill.
How do you layer them into one strategy?
Think of demand as a stack. At the bottom is the steady floor that runs every hour of the year; cover it with Compute Savings Plans, plus Reserved Instances where a workload is stable and specific. Interruptible work comes off that floor and runs on Spot. Above the floor sits the variable peak, the demand that comes and goes; leave it on demand so you never pay a commitment for capacity you do not always use. The Enterprise Discount Program then applies its account wide tier across the whole stack. The single most common mistake is sizing the committed floor too high, covering demand that is not actually steady, which strands spend on capacity you stop using. Size the floor to a forecast you can defend, and never commit against a fleet you have not yet right sized.
A worked example
A scaling fintech paid almost entirely on demand, including for a large steady baseline and a nightly batch job. Mapping demand to models changed the bill without changing the architecture much. The steady floor moved to Compute Savings Plans, the stable primary database to a Reserved Instance, and the interruptible batch job to Spot with a restart safe design, while the genuinely variable daytime peak stayed on demand. Layered this way, the committed coverage matched real steady demand rather than the whole fleet, so nothing was stranded, and the blended rate fell sharply. The purchasing redesign was part of the program that left the company 41 percent lighter on cloud spend. Figures are verified against billing data and anonymised.
Frequently asked questions
What are the AWS pricing models?
Which AWS pricing model is cheapest?
How do these models combine?
Match your purchasing to your demand
We help enterprises map AWS demand to the right pricing models and size commitments to a forecast they can defend, as an independent advisory that takes zero provider commissions and answers only to you. Our guarantee: we reduce your cloud spend or we reimburse our service fee, on a Fixed Fee or a no risk Gainshare basis. Download the AWS Savings Plan kit, read the deeper AWS cost optimization guide, and see how AWS Savings Plans work for buyers.
Put a defensible number on your cloud spend.
No provider in the room, no published price list. Tell us your footprint and we will scope the savings against your billing data — we reduce your cloud spend or we reimburse our service fee.
The Cloud Spend Navigator: what changed in cloud pricing, commitments, and FinOps — no vendor spin.