How Much More Discount Does AWS Give for Three Years Instead of One?

The term premium on an AWS Private Pricing Agreement is roughly five points, not the fifteen your account team implies when it compares against Savings Plans math. This page prices duration in isolation, shows where the premium is real, and shows the four clauses that decide whether three years costs you money.

The Term Premium Is Five Points, Not Fifteen

Hold the annual commit constant and the term premium collapses to roughly five points. At $1M a year, a one-year Private Pricing Agreement lands near 10 percent and a three-year near 15 percent, and independent benchmarking puts the three-year advantage at 5 to 8 points over annual at identical spend, with the marginal step priced at 3 to 5 points per additional year. That is the whole prize. It is not trivial, but it is nowhere near what your account team implies when it walks you through Savings Plans math on a whiteboard. At the SKU layer, where AWS publishes rates, the gap between a one-year and three-year EC2 Savings Plan runs 12 to 22 percentage points depending on payment option. Duration is priced roughly three times more aggressively inside the product than inside the contract, and the reason is simple: a Savings Plan locks an instance family and an hourly rate, while a PPA locks a dollar figure you can spend on almost anything. AWS knows the difference. The blur is deliberate, because a buyer who imports SKU-layer term economics into a contract-layer negotiation will pay four extra years of lock-in for five points he could have had for two. The tell is a rep saying "three years is worth a lot more." Ask which layer that number comes from. If the answer wanders toward Savings Plans, you have just been quoted the wrong instrument. Our discount bands by spend tier benchmark gives you the rate level; this page isolates what duration alone buys.

Duration is priced roughly three times more aggressively inside the product than inside the contract, and AWS is happy for you to confuse the two.

Why AWS Rations the One-Year Option (and What That Tells You)

Advisors disagree about which terms AWS actually sells. Some report one, two, and three years as standard; others report one through five; one reports that three and five year terms are the defaults and one-year deals surface only for accounts under $1M annual commit, priced at a token 3 to 5 percent. Treat that disagreement as the finding rather than noise. The one-year PPA exists. It is rationed, and it is rationed to the exact population least able to push back. That is how a five-point term premium gets presented as fifteen: AWS is not comparing one year at 10 percent to three years at 15 percent, it is comparing a deliberately starved one-year quote at 4 percent to a three-year at 15 percent, and letting you draw the arithmetic yourself. The driver sits in account team compensation, not in AWS's cost of capital. Multi-year total contract value is what gets recognized against quota, and booked revenue predictability is what the field organization is measured on. Nobody on the AWS side is rewarded for handing you a clean, competitively priced twelve-month commit. This is where the two-year term becomes the most useful ask on the curve: it is the least defended point, it rarely has a published internal floor, and it hands the rep something to bring back to pricing that still looks like multi-year TCV. In our experience across cloud commit negotiations, two years at 12 to 13 percent on a $3M commit is a realistic landing zone and preserves a renegotiation window before your architecture changes underneath you, the same logic buyers apply to Google Cloud committed use discounts. Demand priced one-year, two-year, and three-year quotes in writing, side by side, at the same annual dollar. AWS will resist producing the short-term quotes, and the resistance is the point: a refusal to price a term they claim to offer is material you can escalate, and it converts their strongest talking point into your evidence.

Pricing the Flexibility You Surrender

The five points you gain on rate are the easy half of the arithmetic. The hard half is what a wrong forecast costs you across 36 months instead of 12, and AWS has already published the answer at the SKU layer. Commit to M5 capacity on a three-year Savings Plan in January 2024 and AWS ships M7i with roughly 15 percent better price-performance two months later: on a one-year term you rebalance in January 2025, a 10-month lag. On three years you wait 33 months. That lag is not theoretical spend, it is the delta between what you pay and what the market rate would have been for two full years. Usage.ai's worked case makes the number concrete: a three-year commitment priced 15 percent deeper delivers $13,140 of rate advantage and then surrenders $52,560 in unused commitment, a net loss of $39,420. The deeper rate did not save the customer. It financed the mistake.

Translate that into a decision rule you can defend in front of a CFO. The term premium is roughly five points of rate. Your exposure is the probability-weighted shortfall on the commit. If your historical forecast error on annual AWS spend sits above roughly 12 to 15 percent in either direction, the expected shortfall exceeds the five points you are buying, and one or two years wins on expected value alone. In my experience across cloud commits, the accounts that get hurt are not the ones with volatile spend. They are the ones that never measured their own forecast error before signing, so they had no basis to price the risk they were accepting. Pull three years of actual versus budgeted AWS spend before you take a term position. That single number determines whether this is a rate negotiation or a risk negotiation.

The deeper three-year rate did not save that customer money. It financed the mistake.

Instrument the tripwire before signature, not after. Utilization below 80 percent for two consecutive months is the published signal that you have overcommitted by 20 percent or more, and on a $200,000 commitment that is $40,000 a year burning quietly inside a discount you were told was a win. Build the alert into your FinOps reporting on day one of the term, with a named owner and a standing quarterly review against the commit ramp. If your spend trajectory is genuinely predictable, benchmark the rate you are being offered against the AWS private pricing discount bands by spend tier before conceding duration, because term length should never be the first lever you give away.

The Four Clauses That Decide Whether Three Years Costs You Money

Term length is half the trade. The other half is what happens when the forecast is wrong, and that lives in four clauses AWS would prefer you skim. Start with the measurement base, because it is the highest-leverage sentence in the document. AWS measures your commitment against gross, pre-discount spend. Commit $5M at a 10 percent discount and your real obligation is $5M of list-price consumption producing a $4.5M invoice. The discount you just negotiated silently raised the consumption target it was supposed to reward. Read the shortfall definition first, before the rate table.

Carry-forward is where a hard penalty becomes a timing question. Ten to fifteen percent of the annual minimum, rolled into the following year, costs AWS almost nothing in a growing account because the spend arrives anyway, just later. Ask for it explicitly and separately from rate. Early termination exposure is the clause that ends careers: unused commit liability that can run past 75 percent of what remains. On a three-year deal that is a number you cannot pay in cash and cannot argue away in year two. Cap it, or accept a shorter term. Finally, the uplift schedule. A 20 percent annual growth demand repriced the whole agreement while you were reading the discount percentage. The same commit-shortfall structure shows up across hyperscalers, which is why the Google Cloud committed use discount shortfall trap is worth reading as a cross-check on AWS language.

Say the conclusion out loud in the room: a three-year at 15 percent with gross measurement, no carry-forward, and 20 percent uplift is a worse deal than a one-year at 10 percent with clean terms. Price the clauses, not the headline.

What AWS Will Do When You Ask for One Year

Ask for a one-year Private Pricing Agreement and you will trigger a rehearsed sequence. The first move is the rate cliff: the account team returns a one-year number in the 3 to 5 percent range, sometimes 4 to 6, against a three-year offer at 12 to 15 on the same annual dollar. That is not a market price for one year of predictability, it is a deterrent price engineered so the three-year looks compulsory. The counter is to make them defend the gap in writing. Ask for the one-year rate, the two-year rate, and the three-year rate on the same commit, side by side, in the same document. AWS publishes 1, 2, and 3 year terms as available structures, so a two-year quote is a legitimate request and it collapses the cliff into a curve. In my experience the two-year number usually lands within 2 to 3 points of the three-year, which exposes the one-year rate as artificial rather than actuarial.

The second move is tier gating: the claim that the deeper discount band "requires" a three or five year term. Sometimes true at the very top of the curve, usually not at $3M to $10M. Push back by asking which specific approval body imposes the term condition and whether an exception has been granted in the last two quarters. The third move is the sweetener bundle, migration funding, POC credits, marketplace incentives, arriving in the final two weeks of an AWS quarter. Price those credits at zero unless they are unconditional and non-expiring; most are neither. The fourth move is escalation to a regional or worldwide sales leader for a "strategic partnership" conversation, which is a framing exercise, not a pricing one. The fifth is the Savings Plans conflation: quoting the 12 to 22 point SKU-layer term gap as if it applied to the contract layer.

Here is what actually matters. Under term pressure, AWS defends rate hard and concedes language easily. The concessions that move fastest are flexibility clauses: shortfall cure windows, mid-term recommit rights, growth-triggered rate step-ups, and co-termination of acquired entities. A buyer who genuinely needs a shorter horizon should stop fighting for the last point of rate and take three years with one-year risk characteristics instead.

Modeling the Break-Even: When Three Years Actually Pays

Run the arithmetic before the meeting, not in it. On a $5M annual commit, five points is $250K a year and $750K over the term. That is the entire prize. Now price the downside. Because shortfall is measured against gross, pre-discount consumption, a 20 percent miss in year three is $1M of unmet commit at list, and AWS will invoice the gap or roll it forward on its terms. One bad year erases the full three-year premium and then some. The threshold is not close: you need roughly 75 percent confidence that year three consumption lands within 5 percent of plan before the premium is rational.

One bad year erases the full three-year premium and then some.

Three profiles genuinely win on the longer term. First, contracted growth already underwritten: a signed migration plan with named workloads, dated cutovers, and a business case someone has already funded. Not a slide, a plan. Second, workloads with no near-term architectural change, steady-state ERP, data warehouses, regulated systems that will not move for reasons unrelated to cost. Third, accounts already above the $5M threshold, where dedicated account attention arrives and service-level adders stack on top of the cross-service rate, compounding the term benefit in a way it never does at $1.5M. Below that, the premium is thin and the flexibility cost is high.

Three adjacent variables change the answer, and each is priced separately. Where you sit on the spend curve determines the base, covered in our benchmark of AWS private pricing discount bands by spend tier. Service-specific concessions sit on top and are negotiated on their own axis, benchmarked in what service-level adders you can get above the cross-service rate. Our sibling pages on annual commit uplift caps and effective rate after Savings Plans layering complete the model. Compute all four before you accept a term, because a three-year deal with an uncapped year-two uplift is not a discount, it is a schedule.

What To Do First

Do the arithmetic before you open a conversation. Pull 24 months of billing-record spend at gross list, not invoiced net, and compute your actual forecast error month over month. If your 12-month-ahead forecast has been off by more than 15 percent in either direction, you do not have a three-year deal problem, you have a commit-sizing problem, and AWS will happily price your uncertainty as a term premium. Usage.ai's worked case is the cautionary number: a 15 percent deeper discount worth $13,140 erased by $52,560 of unused commitment, a net loss of roughly $39,000. That is the shape of a bad three-year yes.

Then ask for one, two, and three year quotes in a single document, same annual dollar, same service mix, same measurement base. AWS resists this because side by side comparison is where the five-point spread becomes visible instead of the fifteen-point story built from Savings Plans math. Set your three-year ask at the one-year rate plus five points as the floor, not the target. Benchmarks put the real spread at 5 to 8 points, so a three-year quote that only clears one-year by four points is below market. Cross-check the absolute level against the discount bands for your spend tier before you decide whether the term step is worth anything at all.

Two clauses come before term concession, not after: unused commit carry-forward between years, and measurement against discounted spend rather than gross list. Concede term only once both are in writing. Time the close into AWS quarter end, with fiscal year end the stronger of the two, and give the account team a reason to escalate internally rather than to you.

Your walk-away is a two-year deal at a thinner rate with clean flexibility language. A shorter term with carry-forward and net measurement beats a headline number wrapped in gross-spend shortfall math every time. Ask for the three-quote document this week.

Related advisory: the full topic guide.

Frequently asked questions

What discount should I expect on a three-year AWS PPA versus a one-year at the same annual commit?

At roughly $1M annual commit, expect about 10 percent on a one-year and about 15 percent on a three-year, a five-point spread. Independent benchmarks put the term premium in the 5 to 8 point range, with each additional committed year worth roughly 3 to 5 points. If AWS is showing you a spread wider than 8 points, they are almost certainly holding back the one-year rate to make the three-year look compulsory.

Does AWS actually offer a one-year Private Pricing Agreement?

Yes, but it is rationed. Advisors report one, two, three, and five year terms in circulation, while others describe three and five years as the default with one-year offered mainly below $1M commit at only 3 to 5 percent. Ask for all three terms priced in the same document; the refusal to quote a one-year is itself a negotiation signal.

Why is the term gap on Savings Plans so much bigger than on a PPA?

Because they price different things. The published one-year versus three-year gap on EC2 Savings Plans runs 12 to 22 percentage points depending on payment option, since you are locking a specific hourly usage shape. A PPA locks aggregate dollars, which is far less risky for AWS, so duration is worth roughly a third as much at the contract layer. Do not let a rep import Savings Plans math into a PPA conversation.

What is the single most expensive clause in a three-year AWS commit?

How shortfall is measured. AWS defaults to measuring your commitment against gross, pre-discount spend, so a $5M commit at 10 percent means you must generate $5M of list-price consumption while only paying about $4.5M. Negotiating measurement against discounted spend is worth more than a point or two of headline rate, and AWS does concede it for accounts with leverage.

How do I know if my forecast is stable enough for three years?

Compute your absolute forecast error against actual billing-record spend for the last eight quarters. If the error routinely exceeds 12 to 15 percent, the expected shortfall cost outweighs the five-point term premium, and one or two years plus carry-forward is the better deal. Also instrument utilization: two consecutive months below 80 percent is an early overcommit warning.

Is a five-year PPA ever worth it?

Rarely on rate alone. Moving from three to five years typically adds only 4 to 6 points, worth $2.0M to $3.0M cumulative at $10M annual commit, against exit exposure that can exceed 75 percent of remaining unused commit. Buyers who renegotiated at the end of year three have reached similar total spend while keeping the option to reprice, which is usually the stronger position.