AWS opens 2026 term sheets at a 30 to 40 percent annual commit step-up, but mature estates are signing 10 to 20 percent

The illustrative ramp AWS account teams put in front of buyers, $5M growing to $7M, $9M, $11M and $13M across five years, is a 30 to 40 percent annual escalation dressed up as a discount vehicle. Advisory files show opening commitments landing 15 to 30 percent above the customer's realistic spend curve, which converts a 12 to 18 percent discount into a shortfall true-up. Deciding your defensible growth number before the first pricing call is what determines whether the discount survives the term.

Prepared by Redress Compliance · August 19, 2026 · AWS private pricing advisory. 20 to 25 EDP and PPA negotiations benchmarked 2024 to 2026.

The uplift exposure in an AWS Private Pricing Agreement is not created by the discount percentage. It is created by one sentence of standard language: each year's minimum commitment cannot be lower than the prior year's. That is a ratchet, not a forecast.

There is no downward flex clause in the default paper, no mid-term reset tied to divestiture or workload migration, and no forgiveness of unspent commitment.

Shortfall is settled as a true-up, usually at term end, which means the money you did not spend on compute you still pay for as a line item with no asset attached.

Read against the 5 to 20 percent discount range that AWS actually pays across the observable spend tiers, the arithmetic is unforgiving: a 15 percent discount on a five-year term is fully erased by roughly 15 percent cumulative unconsumed commitment.

Sign a ramp you cannot fill and you have paid list price with extra paperwork.

The correct read is that commit growth and discount rate are separate negotiations that AWS deliberately bundles. Your leverage on the growth line is highest before you have named a Year 1 number, because everything after that is anchored arithmetic.

Ask for the shortfall mechanics in writing early: is unspent commitment invoiced, credited toward the following year, or carried to a renewal? A carry-forward provision, even a partial one.

Is worth more than two points of headline discount on any estate whose forecast confidence sits below 90 percent.

The opening paper and the signed paper are two different documents, and the gap is wider on growth than on discount. AWS account teams are still presenting the illustrative five-year ramp, $5M growing through $7M, $9M, $11M to $13M, which is a 30 to 40 percent annual step-up.

Across advisory files the opening commitment lands 15 to 30 percent above the customer's realistic spend curve. Mature estates, meaning organizations past the initial lift-and-shift phase with a functioning FinOps practice, sign 10 to 20 percent.

That is the landing zone, and it is defensible because it matches what those estates actually consume. Where the estate is stabilized, already cost-optimized, or facing a Savings Plans re-baseline that will mechanically reduce on-demand spend in Year 1, 5 to 10 percent holds.

In market experience, the buyers who land at the low end are the ones who bring a service-level consumption model to the first pricing call rather than a total spend number.

Here is the part AWS will not volunteer: accepting a steeper ramp buys almost no incremental discount.

The discount bands stay inside 5 to 20 percent regardless of how aggressive the growth curve is, because the bands are set by commitment thresholds, and the offer itself is a shifting mix of credits and rate discounts whose composition changes quarter to quarter.

Committing to $45M over five years instead of $34M does not move you into a new discount universe. It moves you one threshold, worth perhaps two or three points, in exchange for $11M of additional unhedged obligation.

That is a bad trade at any ratio, and it is the single most common error in the files.

Expect the account team to respond with three moves. First, the aggregate growth narrative: AWS grew 36.7 percent year over year in Q2 2026, so your 12 percent looks conservative. That number is AI and new-logo weighted and has nothing to do with a mature enterprise estate.

Second, term extension, because moving three years to five is worth 4 to 6 points and is the cheapest concession AWS can make while buying two extra years of ratchet. Third, a marketplace pitch, noting that qualifying purchases can contribute up to 25 percent toward commit.

Useful, but that is a separate mechanism from the discount and it should be priced as a shortfall hedge, not counted as growth capacity. Compare the arithmetic against the effective rate once Savings Plans are layered in before you concede anything on the growth line.

Start with the uncomfortable premise: your account team is not asking for 30 to 40 percent annual commit growth because they believe your estate will grow that fast. They are asking because the internal approval form pays them to ask.

An AWS private pricing proposal gets scored on three inputs: total committed dollars, year-over-year growth of the commitment, and term length. The discount authority the team can release expands as those three numbers rise.

So the first paper you see is not a forecast, it is a maximization exercise across all three levers at once, and the growth line is the one the account team can move without asking anyone's permission.

That is why the illustrative ramp lands in front of you before anyone has asked for your capacity plan, your application retirement schedule, or your reserved instance coverage.

Once you see the uplift ask as a scoring artefact, the pricing conversation changes shape.

The relevant question is no longer "is 30 percent growth reasonable," it is "what does each concession cost the seller." Term is cheap for AWS to reward: three years to five years is worth roughly four to six points of discount and costs the account team nothing except patience.

A flat commitment is expensive to them because it drags their growth score down. Shortfall protections cost them close to nothing when your consumption record is clean, because the risk they are underwriting is a risk they have already modeled as near zero.

The buyer who understands that ranking stops trading the wrong currency.

The second thing worth interpreting is the narrative AWS brings to the table in 2026. Amazon reported AWS growth of 36.7 percent year over year in Q2 2026, the fastest in eighteen quarters, on a run rate around $169 billion.

That number will be quoted at you, explicitly or by implication, as evidence that a 20 percent commit step-up is conservative. It is not evidence of anything about your estate. Aggregate AWS growth in this cycle is AI workload and new-logo weighted.

A mature enterprise estate with a stable application portfolio, active rightsizing, and Savings Plans already in place grows organically at 10 to 20 percent, and advisory files put that squarely in the defensible range.

Borrowing the vendor's aggregate curve as your commitment floor is how a 12 to 18 percent discount turns into a shortfall true-up at term end.

Note what AWS actually delivers on discount too: the observed band across benchmarked negotiations runs 5 to 20 percent, never above 20, and the offer is a mix of credits and discounts whose composition shifts quarter to quarter.

So a "richer" number late in the quarter often just means more credit and less rate.

Test that against the discount bands by spend tier before you accept a growth ramp as the price of a better headline.

Now the capacity story, because it will come up. With 2027 capacity largely reserved and 2028 partly spoken for, the account team has a ready-made urgency line: commit now or take your chances on access. Read the same fact from the other side.

If capacity is genuinely sold forward, AWS does not need your growth commitment to fill racks; it needs your growth commitment to firm up a forecast it is already building against. That is an appetite, not a constraint on you.

The urgency framing is weak, but the appetite is real and unusually strong right now, which cuts both ways. You should not panic-sign, and you should absolutely charge for what you are being asked to underwrite.

Which brings us to the asymmetry that actually matters. A growth commitment is a one-way obligation. If your consumption overshoots the ramp, you pay for the overage at your negotiated rate and AWS books it.

If your consumption undershoots, you pay the shortfall anyway, and the effective discount you bragged about in the business case inverts. For AWS, the same line is a forecast credit with no downside. There is no version of this contract where the ramp helps you and hurts them.

So price the ramp as debt, not as a discount input. Every dollar of Year 3, 4 and 5 commitment above your realistic curve is unsecured borrowing at 100 cents on the dollar, and it should be modeled that way in the approval memo you take to your CFO.

When the account team offers two more points for a steeper ramp, run the arithmetic: two points on a $30 million total is $600,000, and $4 million of unsupported commitment is $4 million of exposure. That trade loses.

Once you have that number in front of you, the growth negotiation becomes a credit decision rather than a percentage argument, and you stop conceding the one term that has no ceiling.

The counter is not "we want a flat commit." It is a specific structure that gives AWS the total dollars and the term it scores on while pulling your exposure out of the early years, where forecast error is highest and your negotiating position is weakest. Take a $30 million three-year deal.

A flat $10 million per year puts $10 million at risk in Year 1, before your migration has landed and before you know whether the AI pilot converts.

A back-weighted $7 million, $10 million, $13 million ramp delivers the identical $30 million total and the identical headline discount rate, and cuts roughly $3 million of first-year exposure. AWS's growth score actually improves under the back-weighted version.

This is the rare structural change that scores better for the seller and costs the buyer less, which is exactly why it should be your opening counter rather than your fallback.

Then attach the language that makes the ramp survivable. Ask for a reforecast right in Year 2 or Year 3 tied to defined material business change, with a floor rather than an unlimited reduction, so AWS's approvers see bounded risk.

Carve out divestitures and carve-ins explicitly: if you sell a division that carries 15 percent of your run rate, the commitment should fall with it, and if you acquire, the acquired spend should count toward commit rather than sitting outside it.

Get the marketplace mechanism straight while you are drafting, because up to 25 percent of your commit can be satisfied by marketplace purchases, and that is a separate lever from the discount, not a substitute for it.

Add shortfall grace: rollover of an underspend into the following year, or a capped true-up percentage instead of dollar-for-dollar recovery.

The trade sequence matters more than the individual asks. Give term, not growth. Moving three years to five is worth four to six points to you and is inexpensive for AWS to approve, so spend that concession first and hold the ramp.

Give commit consolidation across business units, which raises total dollars without raising your growth obligation. Give a reference or a case study if legal will allow it.

Do not give a steeper ramp for a rate improvement, and do not give up the right to reforecast in exchange for credits, since credits expire and the commitment does not.

Where you want an above-band number, understand in advance what AWS charges in concessions for above-band discounts, because the currency they ask for is almost always growth or term.

Timing is the last lever. Shortfall protections, reforecast rights and M&A carve-outs cost AWS almost nothing when your consumption record is clean, which means they are essentially free before signature and functionally unavailable after.

Once you have signed, the same request becomes an amendment, and an amendment gets priced.

Bring the structure to the first pricing call in writing, with the three-year ramp already drawn and the reforecast clause already drafted. Account teams negotiate against proposals, not against objections.

The numbers in this piece come from three sources that do not agree on much but converge on the uplift question.

First, 20 to 25 EDP and PPA negotiations benchmarked between 2024 and 2026 across advisory files, where observed discounts ran roughly 5 to 20 percent and the commitment ramp, not the discount, was where the money moved.

Second, a third-party dataset covering 145 negotiated contracts and $18.7B in aggregate committed spend, which is the only place the shape of the discount curve becomes visible.

Third, AWS's own earnings disclosure: 36.7 percent year-over-year growth in Q2 2026, the fastest in eighteen quarters, on a $169B run rate. That third number is what account teams carry into the room.

It is also the least relevant to a mature estate, because AWS aggregate growth is AI workload and new logo weighted, and a seven-year-old production estate does not organically grow at the same rate as the vendor's overall book.

Four patterns repeat across nearly every file. The opening commit is anchored above the customer's own trendline, not below it, which means the buyer is negotiating down from a fiction rather than up from a fact.

The non-linear curve is deployed as a growth argument: because the marginal point of discount gets cheaper for AWS at scale, the account team can honestly say a bigger commit unlocks more, and dishonestly imply you can reach that tier.

Mandatory Enterprise Support, roughly $100K per year at $1M of spend, inflates the apparent base the ramp is calculated from, so a percentage step-up applies to a number that includes a cost you did not choose.

And account teams are instructed not to share benchmark data, which is the whole game: the buyer who does not bring an independent number negotiates against the seller's number.

If you need to use market data without exposing where it came from, there is a workable method for proving a benchmark to your account team without naming the source.

The tier logic matters too, since a small commit increase can cross a threshold while a larger one does not, which is the point behind finding the commit tier break.

Related advisory: the full topic guide.

What annual commit growth rate does AWS actually expect in an EDP or PPA?

AWS's illustrative ramps in 2026 term sheets commonly step up 30 to 40 percent per year, for example a $5M Year 1 commit growing to $7M, $9M, $11M and $13M over five years for $45M total. That is an opening position tied to internal proposal scoring, not a requirement.

Mature estates routinely negotiate down to 10 to 20 percent annual growth, and stabilised or cost-optimised estates can defend less.

Can I negotiate a completely flat annual commitment?

Standard PPA language requires each year's commitment to be no lower than the prior year, so a declining ramp is very hard to get. Genuinely flat is achievable but expensive, because a flat commitment is the concession AWS's scoring model penalises most heavily.

The more efficient play is a modest total growth number delivered through a back-weighted curve rather than fighting for a literal flat line.

Does accepting a steeper ramp actually get me a bigger discount?

Only marginally, and rarely enough to justify the exposure. Observed EDP and PPA discounts run roughly 5 to 20 percent, stepped by commitment thresholds, and no ramp shape pushes you above that ceiling.

Term length is the cheaper lever: moving from three years to five typically adds around 4 to 6 points, which is more discount than a steeper growth curve will buy you.

What happens if I do not hit the committed spend in a given year?

The shortfall does not disappear. It is settled as a true-up, generally at term end, which means an over-forecast commit curve converts your negotiated discount into a net cost.

This is why opening commitments sitting 15 to 30 percent above your realistic spend curve are the single largest financial risk in the deal, larger than a two-point discount difference.

How do I justify a lower growth number when AWS points to its own 37 percent growth?

Separate AWS's aggregate revenue growth from your estate's organic growth. AWS grew 36.7 percent year over year in Q2 2026 to a $169B run rate, but that acceleration is weighted toward AI workloads and new logos.

Bring a 24-month consumption trendline for your own account, net of one-off migration waves and Savings Plans re-baselining, and negotiate against that curve rather than against a quota narrative.

What is a back-weighted ramp and why does it matter more than the headline rate?

A back-weighted ramp keeps the same total commitment and the same headline discount but shifts the dollars into later years. On a $30M three-year deal, $7M / $10M / $13M instead of $10M / $10M / $10M leaves roughly $3M less at risk in Year 1, which is the year your forecast is least reliable.

It costs AWS almost nothing to agree pre-signature and is close to impossible to retrofit later.

When is the right time to raise the growth rate in negotiation?

Before the first pricing conversation, in writing, with your own trendline attached. Once AWS has anchored an illustrative ramp and built internal approvals around it, you are negotiating down from their number rather than up from yours.

Establish your Year 1 ceiling and your annual growth range as inputs to the discount discussion, not as concessions extracted from it.