TL
The short answer

On AWS, a three year Savings Plan or Reserved Instance discounts deeper, roughly up to 72 percent against on demand versus closer to 66 percent for a one year term, but it locks you in for three times as long with no way to cancel. The right choice is not the deepest discount, it is the longest term your forecast can defend: commit three year terms only to the steady floor of demand you are confident will still run in three years, use one year terms for demand that is steady but less certain, and leave variable peaks on demand. Because the unused portion of any commitment is still owed, term length is a utilization risk you carry, so the buyer goal is high utilization across a laddered portfolio rather than the maximum headline discount.

Here is the real tradeoff between the two terms, the math that decides it, and a laddering approach that captures most of the discount without betting three years on a forecast you cannot stand behind.

What do you actually gain by going from one year to three?

The three year term buys a deeper discount in exchange for a longer lock. Indicative depths run to roughly up to 72 percent against on demand for three year all upfront Savings Plans and Reserved Instances, against roughly up to 66 percent for the equivalent one year term, with the exact gap depending on instance family, region, and payment option. That marginal six or so points sounds small, but on a large steady base it compounds. The catch is that the deeper rate only pays off if the committed capacity stays utilized for all thirty six months. A three year commitment that goes underutilized in month twenty after a rearchitecting or a migration can easily give back more than the extra discount it bought. Verify current depths against the AWS pricing pages, since published rates change.

One year or three year: which fits which workload?

DimensionOne year termThree year term
Indicative discountRoughly up to 66 percent vs on demandRoughly up to 72 percent vs on demand
Lock in12 months, refreshes annually36 months, no cancellation
Forecast confidence neededModerate, one year horizonHigh, three year horizon
Best forGrowing or shifting fleets, newer workloadsMature, stable demand floors and steady databases
Main riskSmaller discount left on the tableStranded commitment if the workload moves or shrinks

The deciding question is not how much discount you can get, it is how confident you are the demand will still be there. Figures are indicative; confirm against current AWS pricing.

How do you ladder the two terms together?

Treat coverage as layers against a demand profile. Identify the floor that runs every hour through troughs and quiet seasons, then split it by confidence. The deepest, most certain part of the floor, a stable production database or a baseline service that has run flat for years, carries three year commitments. The next layer, steady but subject to a possible rearchitecting or growth, carries one year commitments that you renew or adjust annually. The variable peak above the floor stays on demand. Laddering staggers your renewal dates too, so you are never forced to recommit the whole portfolio in a single quarter against one point in time forecast. Aim for utilization in the high nineties on every layer rather than the highest coverage percentage.

A worked example

Worked example

A scaling fintech was about to put its entire steady compute base on three year all upfront Savings Plans to chase the deepest rate. Reviewing the demand profile showed that only about 60 percent of that base was genuinely durable production demand; the rest was a platform mid rearchitecting that would change instance families within a year. We put the durable 60 percent on three year terms, the uncertain portion on one year terms, and left the variable peak on demand. When the rearchitecting landed, the one year layer simply was not renewed against the old families, avoiding a stranded three year lock, while the three year base kept compounding its deeper discount. The laddered structure 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

Is a three year AWS Savings Plan worth it?
A three year plan is worth it for the steady floor of demand you are confident will still run in three years, because it discounts roughly up to 72 percent against on demand versus closer to 66 percent for a one year term. The extra discount is only real if utilization stays high for the full term, so commit three year terms only to demand with a defensible long horizon forecast.
Can you cancel an AWS Savings Plan or Reserved Instance early?
No. AWS Savings Plans cannot be cancelled, and Standard Reserved Instances cannot be cancelled either, though Convertible Reserved Instances can be exchanged. The unused commitment is still owed, which is why term length is a risk decision, not just a discount decision.
How do you split coverage between one year and three year terms?
Layer the terms to match forecast confidence: put three year commitments only on the demand floor you are sure persists, one year commitments on demand that is steady but less certain, and leave variable peaks on demand. This laddered structure captures most of the discount while keeping the portfolio able to flex as the fleet changes.

Get your AWS commitment term right

We size and ladder AWS Savings Plan and Reserved Instance portfolios to a forecast you 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 compare instruments in Reserved Instances versus Savings Plans.

Independent · buyer-side

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.

Buyer-side intelligence, monthly.

The Cloud Spend Navigator: what changed in cloud pricing, commitments, and FinOps — no vendor spin.