AWS Private Pricing Discount Bands by Spend Tier: What a Good Number Looks Like

Your AWS discount is determined by your commit tier, your term, and how credible your alternative looks, not by the largest number an advisory firm has ever published. This guide maps the defensible bands at each tier, names the marketing inflation that ruins buyer targets, and shows you how to hold a number your account team will actually clear internally.

The Number You Should Be Chasing Is Set by Your Tier, Not by Someone Else's Deal

The fastest way to lose an AWS negotiation is to walk in with a number you borrowed. I have watched buyers at $1.2M of annual commit open at 25% because a conference slide or a vendor blog told them that was the market, and then spend three months trying to recover credibility with an account team that stopped taking their target seriously in week two. The account manager does not argue with you. He nods, escalates nothing, and waits. Meanwhile your renewal clock runs down and your only remaining lever is time pressure, which is AWS's lever, not yours. The uncomfortable truth is that at $1.2M you are in a band that credible advisory observation puts around 8% to 12%, with AWS's standard opening offer sitting inside that band, and the 25% you read about belongs to a company committing twenty times what you commit. Importing that number did not make you ambitious. It made you cheap to dismiss.

The arithmetic of an AWS private pricing outcome is more boring and more useful than the headline chase. Your tier sets the range. Term moves you roughly five points inside it: at a fixed $1M commit, observed outcomes run near 10% on a one-year deal against roughly 15% on three years, which is the single most reusable ratio in this whole exercise. Credible competitive tension from Azure or Google adds two to four points, and only if it is real enough to survive one question about which workloads move and when. Everything else you win is clause value, not percentage: shortfall cure periods, Marketplace offset treatment, ratchet relief, co-termination, exit rights. Buyers who chase the last two points of headline discount routinely give away clauses worth more than those points, because the discount is the number the CFO reads and the clauses are the ones that bill you.

Importing someone else's discount number did not make you ambitious, it made you cheap to dismiss.

There are two very different readers here and confusing them is expensive. If you sit between roughly $500K and $2M of annual commit, you are signing a template. AWS suggests $1M as the practical entry point, some buyers get in nearer $500K on a high-growth story, and dedicated account attention generally does not arrive until closer to $5M. Your leverage is not the discount percentage, which is largely pre-baked; it is term length, growth assumptions, the Enterprise Support attach, and whether the shortfall clause has a cure period. Fight there and you will get a defensible deal. Fight for 20% and you will get a polite no and a signature deadline. If you are at $10M or above, you have something structurally different: an approval chain. Points above the standard band at your tier exist because a regional leader can sign them, and that person responds to a documented competitive alternative, a growth commitment they can put in a forecast, and a deal that closes inside their quarter. That is where 14% to 20% becomes reachable and where our published AWS EDP discount benchmarks should be read as a floor for your ask rather than a ceiling.

So the first move is not to name a number. It is to establish which of those two negotiations you are actually in, then set a target that a regional leader could clear without an exception memo. A target AWS can approve is worth more than a target AWS admires.

PPA and EDP Are the Same Vehicle: Why the Label Change Matters to Your Leverage

AWS retired the Enterprise Discount Program label and now papers new agreements as Private Pricing Agreements. The commitment structure, the tiering logic, the Marketplace offset rules, the Enterprise Support attach, all of it carries over. Your account team will use both terms in the same call, sometimes in the same sentence, and neither word signals a different economic construct. Treat a mid-negotiation switch from "your EDP" to "the PPA framework" as vocabulary, not news. Where buyers get hurt is when the relabeling is used to reset expectations: a fresh paper name invites the suggestion that prior benchmarks, prior terms, and even your own prior agreement's concessions are no longer the reference point. They are. If your last agreement carried a 60-day shortfall cure and Marketplace counted toward commit, those are precedents you already own, and the burden of explaining why they disappeared under a new document title sits with AWS.

The deeper leverage point is that AWS has never published a discount rate card for either label. There is no list price for private pricing. Every band circulating in the market, including the ones in this article, is advisory observation drawn from deals people saw, and the sample sizes are small. That cuts both ways and most buyers only notice the side that hurts them. Yes, it means your benchmark is contestable. It also means the account team's assertion that "14% is the maximum at your tier" is equally unsourced, and you are entitled to ask what it is based on. In twenty five years across this table, the answer has never been a document. It is a desk guideline, an approval threshold, or a quota position. Naming that out loud changes the conversation from you defending a number to both sides discussing what an approver can actually sign, which is the only conversation that produces movement.

Practical use of the asymmetry: never present your target as a market fact. Present it as a range with a stated basis and an internal decision attached, the same discipline that works on Cisco ELA spend tiers and every other unpublished-rate vehicle. Say what your tier and term support, say what the competitive alternative is worth, and say what happens on the date if the number is not there. Then make AWS argue against a structure rather than against a percentage. Silence about method invites dismissal; a stated method forces a counter, and a counter is progress.

The Credible Discount Band at Each Spend Tier

Here is the honest starting point: AWS publishes no discount rate card, so every band you will ever see (including this one) is advisory observation assembled from deals people have sat in. That cuts both ways. It means your account manager cannot show you a table proving your number is impossible, and it means you cannot show them a table proving your number is standard. What you can do is arrive with a band that is internally consistent, tier-appropriate, and net of the things AWS strips out of its own math. The single most damaging thing a buyer brings to the table is a target lifted from a tier above their own. A $2M commit chasing 20% will burn three months of calendar and land at 9%, because the account team knows the deal desk will never route that approval, and they will happily spend your time proving it to you.

Two published views sit on either side of that table and you should use them differently. The conservative view (roughly 5% at entry scaling to 15 to 20% at the largest commitments) is what you plan against and what you put in your approval memo, because it is the view that survives contact with a finance business partner who will ask what happens if you miss. The aggressive view drawn from contract samples (12 to 18% at $5M to $20M, 18 to 28% above $20M) is what you cite across the table, because it is sourced to a body of contracts rather than one advisory's client base, and because an account manager cannot dismiss it without conceding they have no rate card either. Plan against the floor. Negotiate against the ceiling. Never confuse which document you are writing.

Plan against the floor, negotiate against the ceiling, and never confuse which document you are writing.

Notice what the table does not do: it does not scale smoothly. That is deliberate and it is the most exploitable feature of AWS private pricing. Observed breakpoints cluster at $1.5M, $2M and $5M, and moving from $1.5M to $1.6M of commit has been documented to add several percentage points, not several basis points. If your forecast lands at $1.4M or $4.6M, you are leaving points on the table for the sake of a rounding error in a spreadsheet you built yourself. Model the commit at the next breakpoint, price the shortfall risk on the incremental dollars, and compare the two. In most cases the incremental exposure is a fraction of the discount gain, because the discount applies to the entire commit while the risk applies only to the delta. Our broader work on AWS EDP discount benchmarks and what good looks like covers how to structure that math for an approval committee.

What AWS will do in response is predictable enough to script. They will open in the 8 to 10% zone regardless of your tier, because that number clears without escalation and roughly 40% of buyers accept it. They will attribute the gap between your target and their offer to your growth profile rather than to their approval structure, which reframes a pricing conversation as a commitment conversation. They will offer a five-year term as the bridge, which is worth roughly 4 to 6 points and almost never worth two extra years of exposure. And if you hold, they will escalate, because at $3M and above there is a regional approval path that exists precisely for this purpose. From market experience across these negotiations, the deals that land at the top of their band share three traits: a documented alternative that a competitor has actually priced, a commit set at a breakpoint rather than at forecast, and a buyer who never once quoted a percentage from a tier they do not occupy.

• Below $1M, stop negotiating the percentage and negotiate the terms. You are in template territory and the points are not there; the flexibility clauses are.
• At $1M to $3M, treat 8 to 10% as the opening, not the answer. The gap between opening and outcome at this tier is 2 to 4 points and it is entirely a function of whether you push.
• At $3M to $10M, a landing below 12% means the deal was not competitive. Document a priced alternative before you name a number.
• Above $10M, insist on knowing which approval level your number requires. If your account manager cannot answer, you are talking to the wrong person.
• Above $25M, the cross-service rate stops being the whole conversation. Service-level adders on heavy compute are where the remaining upside sits.

State Your Band Net of Enterprise Support or You Are Lying to Your Own CFO

Enterprise Support is a mandatory attach on private pricing. You cannot sign the commit without it, and it runs at roughly 3% of monthly usage for large accounts with a $15,000 per month floor. That is not a footnote, it is a third of your discount at the mid tiers. A 10% headline on a $2M commit delivers $200K of nominal benefit and consumes roughly $60K in support fees on the same usage base, netting closer to 7%. At $1M commit, the $15,000 monthly floor is $180K a year against a $60K to $90K discount, which means the entry tier can be net negative on paper before you count the operational value of the support tier itself. AWS account teams will never present it this way and they have no reason to. Their compensation and their internal approval both reference the gross percentage.

The second layer is the Savings Plans and Reserved Instance stack. Commitment discounts apply first, then the private pricing percentage applies to what remains. So if 60% of your compute is already under a three-year Savings Plan, the headline percentage is operating on a much smaller base than your gross AWS bill, and any benefit projection built off total spend is inflated. Then there is the burn-down asymmetry: the discount itself does not retire commit. Spend $2M, receive $200K of benefit, and only $1.8M counts toward the commitment. You are structurally short by the exact amount of your own discount, every year, which is a shortfall exposure most buyers discover in month nine.

Do this before your next internal review. Rebuild every AWS proposal in one column: gross benefit, minus Enterprise Support at 3% or the floor (whichever binds), minus the portion of spend already covered by Savings Plans, plus the commit gap created by the discount itself. Present that net number to your CFO alongside the headline, and label the headline as the vendor figure. This is the same discipline that separates real from nominal in Cisco ELA tier benchmarking, and it does two things at once: it protects you when the savings do not materialize as forecast, and it hands you a legitimate argument for more points, because AWS cannot dispute the arithmetic of its own mandatory attach. From market experience, buyers who put the net table in front of their account team recover 1 to 2 points on average, purely because the gap becomes visible to the person who has to justify the approval.

Term: The 3-Year Versus 5-Year Trade Is Worth About 5 Points and Rarely Worth Taking

Term is the concession AWS gives away most easily, which tells you exactly how much it values your signature on year four and year five. At a fixed $1M commit, a one-year private pricing agreement lands near 10% while a three-year at the same spend level lands near 15%. That five point spread is not a reward for volume, it is a reward for duration, and it is the cleanest arbitrage available at the entry tiers. Push further and the curve flattens: moving from a three-year to a five-year term typically adds 4 to 6 points, which at $10M of annual commit is worth roughly $2.0M to $3.0M cumulative. Those are real numbers and your CFO will like them on a slide. They are also the numbers AWS will lead with the moment you signal that budget is the constraint, because extending term costs AWS nothing today and locks in your consumption while your alternatives are still theoretical.

Interpret the trade properly. Four to six points on a five-year term buys you two additional years of exposure to three risks you do not control: AWS list price movement (the discount is a percentage off a list you are not allowed to fix), service deprecation and re-architecture forced by AWS roadmap decisions, and your own business change including divestiture, acquisition, or a platform strategy shift you cannot currently see. A five-year term also destroys your renewal leverage twice over: you give up the natural renegotiation event at month 36, which is the single moment when your account team is most motivated, and you arrive at the eventual renewal with five years of consumption history proving you are not going anywhere. In market experience the buyers who take the five-year deal do so for budget optics rather than economics, and they discover in year four that they are paying a three-year rate on a five-year workload profile that no longer matches the commit.

There are legitimate cases for the longer term. If your workloads are stable, regulated, and effectively immovable (core banking, clinical systems, sovereign or FedRAMP-constrained estates), if you have no migration optionality worth preserving because the exit cost dwarfs any discount delta, and if your consumption forecast has been accurate to within roughly ten percent for the last two years, then the extra points are close to free money. But the five-year should never travel alone. Minimum protections: annual commit fixed in nominal dollars with no ratchet beyond what you signed, a mid-term reset or re-baseline right at month 30 or 36 tied to a defined trigger (M and A, divestiture, or a variance threshold), the discount schedule locked so the percentage cannot be re-tiered downward if your spend mix shifts, a price protection clause covering list increases on your top ten services by revenue, and a service continuity commitment so a deprecation event that forces re-platforming gives you a commit adjustment rather than a shortfall bill. If AWS will not write those, the five-year term is not worth 6 points, it is worth walking away from.

The tactical play is to sell duration late and cheaply. Open on a three-year at the top of your tier band, work the service-level adders and the clause package first, and hold the five-year in reserve as the closing concession when your account team needs a few more points to clear internal approval on something you actually care about. Never volunteer it in the first meeting. If AWS opens by pricing only the five-year, ask for the three-year and the one-year quoted side by side on the same commit, in writing. That comparison exposes the term premium as a discrete line item rather than letting it hide inside a single blended number, and it is the same discipline that produces defensible targets across the tiers documented in the AWS EDP discount benchmark bands.

Tier Breakpoints: Where an Extra $100K of Commit Buys Several Points

The discount curve is a staircase, not a ramp, and buyers who negotiate as though it were a ramp leave points on the table for no reason. Observed pricing breaks sit near $1.5M, $2M and $5M of annual commit, and at those thresholds a minor increase in committed volume yields a disproportionate jump. Moving a $1.45M commit to $1.55M can add several points. Moving $2.5M to $3.0M, an increase of half a million dollars, often adds close to nothing because both numbers sit inside the same band. That asymmetry is the most underused lever available to mid-tier buyers, and it costs you nothing to exploit because the incremental commit is spend you were going to make anyway.

The failure mode is predictable. A buyer forecasts $1.4M, adds a safety margin down rather than up because shortfall is billable, commits $1.3M, and signs a rate that a $1.55M commit would have beaten by three or four points on the entire base. That is the wrong risk calculation. The correct sequence is to model your commit in $100K increments against the actual discount schedule before you agree a number, then decide whether the incremental commit is real consumption or a gamble. If your run rate genuinely supports $1.55M, crossing the breakpoint is the cheapest points you will ever buy. If it does not, you are trading a shortfall exposure for a rate improvement, and that trade only works if you also secure a cure period and Marketplace offset headroom (Marketplace ISV purchases can retire up to 25% of the annual commitment, which is real ballast against a stretch number).

Do not guess at where the steps sit. Make AWS show you. The ask is specific and reasonable: quote the same three-year term at $1.3M, $1.4M, $1.5M, $1.6M, $1.8M and $2.0M of annual commit, on one page. Account teams resist this because it exposes the staircase and hands you the arithmetic, and they will counter by offering a single blended number and calling the schedule confidential. Push back once, in writing, and escalate to the regional leadership approval level that has to sign anything above the standard opening band anyway. In market experience the request gets answered roughly half the time at the $1M to $5M tiers, and even a partial answer tells you which side of the step your current forecast sits on. The cluster analysis on commit tier breaks goes deeper into how to reconstruct the schedule from partial quotes when AWS refuses the full grid.

Extending term costs AWS nothing today, which is precisely why it is priced as generously as it is.

First action: pull your last twelve months of actual spend, project the next twelve at your current growth rate, and identify the nearest breakpoint above that number. Then decide, before any AWS meeting, the maximum commit you would sign and the discount that would justify crossing the step. Bring both numbers. Negotiate term last.

How AWS Inflated Bands Get Manufactured, and How to Spot Them in a Deck

Nobody in this market publishes a rate card, so every number a buyer walks in with came from someone's inference. That is fine. What is not fine is inference dressed as observation, and there are exactly three mechanisms that produce it. The first is blended outcomes. When you see "40%+ savings" in an advisory deck, read the footnote: that figure is achieved "when combined with aggressive AWS rate optimization that targets specific service types." Translated, it means the private pricing percentage plus Savings Plans coverage plus rightsizing plus Graviton migration plus storage tiering, all rolled into one headline. Those are real savings, but three quarters of them come from work your engineering team does, not from a signature your account team clears. Bring a 40% number to a private pricing negotiation and you have asked AWS to pay for optimization you were going to do anyway. The second mechanism is category confusion, and it is the single most common reason buyers arrive with an impossible target. Commitment discounts versus on-demand run 30 to 40% on one-year terms and 50 to 65% on three-year Savings Plans and Reserved Instances, with another 5 to 15% for upfront payment. Those are resource-level discounts on a specific instance family. The cross-service private pricing percentage is a different animal, applying across more than 200 services, and it stacks on top of the commitment discount rather than competing with it. Benchmark guides that publish a "10 to 55%" EDP or MACC range are almost always mixing the two, sometimes with Azure MACC mechanics stirred in for good measure. The third is small-n self-reporting. The most aggressive stacking claims in circulation, service-specific adders of 11 to 19 points above the underlying cross-service rate, rest on roughly 25 to 35 negotiations run in 2024 and 2025 with no contract disclosure. That is a plausible dataset from real work, and the same source concedes that combined outcomes rarely exceed 35% total, which is the honest part. But 30 undisclosed deals cannot establish a band, and AWS knows it.

An impossible target does not make your account team work harder; it makes them stop working the deal at all.

Here is why this is not an academic point. The cost of an inflated target lands on you, not on the advisory firm that published it. When a buyer opens at 30% cross-service on a $4M commit, the account manager does the arithmetic in real time, concludes the deal is unwinnable, and reallocates the quarter to accounts that will close. Escalation to regional leadership requires the account team to sponsor your case, and no rep sponsors a case they expect to lose. In practice you get slow-walked: fewer meetings, a solutions architect instead of a decision maker, and a first paper that arrives three weeks before your renewal date with the standard opening offer on it. The deal you eventually sign is frequently worse than the disciplined ask would have produced, because you burned the calendar and you are now negotiating against your own expiry. In market experience the pattern is consistent: buyers who open two to four points above their tier band land inside it, and buyers who open fifteen points above it land at or below the standard opening. Discipline is not modesty. It is the mechanism by which your number gets carried into an internal approval process. Three tests to run on any band before you build a target from it: name the denominator (is the percentage off list, off on-demand, or off already-discounted commitment rates), name the numerator (is it cross-service private pricing alone or a blended savings outcome), and name the sample (how many contracts, at what tiers, in what years). If a source cannot answer all three, use it as color and set your target from your own tier and term. Our AWS EDP discount benchmark work exists precisely to give you the defensible spine rather than the best-ever anecdote.

Service-Level Adders: Where the Real Upside Above the Cross-Service Rate Sits

Once you accept that the cross-service percentage is capped by your tier, the interesting question becomes where the incremental points actually live. They are not in a bigger blanket number. AWS does not hand out a 24% cross-service rate on a $6M commit because a buyer asked twice. What AWS will do, and does routinely, is price specific services below the blanket rate when it can see the workload and wants to protect it. That is the entire game above your tier band: fix the cross-service floor, then attack your three or four concentrated spend lines individually with usage data on the table. Heavy EC2 and Fargate, Lambda at scale, and the managed data services (RDS, Aurora, DynamoDB, OpenSearch, Redshift) are where the concentration usually sits, and they are also the lines AWS is most defensive about because they are the ones a competitor can quote against. The published stacking claims of 11 to 19 points above the underlying cross-service rate on heavy compute are, as noted above, thinly sourced, but the direction is right and the ceiling is consistent across sources: combined cross-service plus service-specific outcomes rarely clear the mid-30s in total, and mid-30s is a large, concentrated, multi-year deal with a genuine alternative in the room, not a $3M renewal.

Sequence matters more than ambition here. If you open on service-specific pricing before the cross-service rate is settled, AWS will trade you an adder in exchange for a lower blanket number, and you will lose the trade because the blanket number applies to everything. Settle the floor, get it in writing, then present the concentration. The data you need is unglamorous: twelve months of Cost and Usage Report detail by service, the growth curve on each of your top four lines, and a clear statement of what is portable and what is not. AWS prices adders against a workload it can see and a threat it believes. A Lambda-heavy platform with no realistic exit gets a courtesy point. The same platform with a documented containerization path and a competing quote gets meaningfully more. Two practical constraints to build into your model: adders are typically drafted as service-specific rate cards with their own review or true-up language, so read the term alignment carefully, and adders do not usually change the burn-down arithmetic, meaning a deeper rate on EC2 retires your commitment more slowly and quietly pushes you toward shortfall risk. Model the commitment coverage at the adder rate, not the list rate, before you sign.

• Rank your top four services by twelve-month spend and by projected growth; anything under 8% of total spend is not worth a separate negotiation track.
• Fix the cross-service percentage in writing first, then table the concentration; never let AWS trade a blanket point for a service point.
• Attach a portability story to each concentrated line (containerization, managed-service equivalence, data egress cost) because the adder is priced against the credibility of that story.
• Re-run your commitment burn-down at the adder rates to confirm you still clear the annual number, then negotiate the shortfall cure period accordingly.

First move this week: pull the twelve-month service-level spend breakdown and mark which lines a competitor could genuinely quote against. That single sheet determines how much upside exists above your tier band, and it is the only document that makes an adder conversation productive rather than aspirational.

The Clauses That Are Worth More Than Two Extra Points

Every buyer I have sat next to walks into the room with a discount target and walks out having lost more money on paper terms than the discount fight was ever worth. The arithmetic is not subtle. At a $3M annual commit, two additional points of cross-service discount is $60K a year, $180K over a three-year term. A single shortfall event on a commit you oversized by 12% is $360K, billable, in one invoice, and it refunds nothing. AWS knows this asymmetry better than you do. The account team will trade paper terms for discount points all day because the discount comes out of a governed rate card their manager has to approve, while the shortfall clause, the true-up cadence, and the excluded-services definition come out of the standard template that nobody senior reviews. That is where your leverage is cheapest and your return is highest. Below $10M annual commit, I would take a soft ramp and a 90-day cure over one more point every time, and I have never regretted it.

The shortfall clause is the fight. AWS's standard construction requires payment of the difference between committed and actual consumption at the end of the measurement period, and the default posture is that this is simply billed. Achievable enhancements observed across advisory practice cluster around a cure period of 30 to 90 days after the measurement date, during which you can consume your way out of the gap rather than write a check. Push for 90. AWS will open at zero, then concede 30 as a "process accommodation," then land at 60 if you hold. The second-order ask is carry-forward of unused commit into the following contract year, which AWS resists harder because it breaks their revenue recognition, but which is worth more than the cure period because it converts a hard cliff into a rolling balance. Ask for both. Settle for cure plus a reshaped ramp.

Ramp shape is the single most underused lever at the $1M to $5M tiers. AWS defaults to a flat or front-loaded annual commit because it de-risks their year one. Your migration timeline almost never matches that. A back-loaded ramp on a $9M three-year total (say $2.4M, $3.0M, $3.6M rather than $3M flat) costs AWS nothing in total contract value, preserves your discount tier because they price on the aggregate, and removes the year-one shortfall exposure that kills most first-time PPA buyers. Say it plainly in the room: total commit unchanged, shape changed. That framing gets signed. A demand to reduce the total does not.

Two extra discount points is $60K a year at a $3M commit; one shortfall event on a commit you oversized by 12% is $360K, billable, in a single invoice.

Three of those rows deserve a note on how AWS responds. On excluded services, the account team will tell you the list is standard and non-negotiable. It is standard. It is not non-negotiable, and more importantly the exposure is not the current list, it is AWS's ability to add to it during the term. Get "no unilateral additions without written consent" and you have neutralized most of the risk without arguing about any individual service. On the price file freeze, expect real resistance, because AWS reprices publicly and often. What you can usually win is a freeze on your top five services by spend, which is where 70% or more of the exposure sits. On mid-term renegotiation, AWS's counter is a "commercial review" with no obligation attached. Refuse that language. A review you cannot act on is theater. Tie the right to specific, objectively verifiable events: a divestiture over a defined revenue threshold, an acquisition, or a documented workload migration off the platform. AWS will accept event-triggered language far more readily than an open right to reopen, because it bounds their exposure. That is the trade. Take it.

Sequence matters. Land the paper terms while the discount number is still open, not after. Once you have agreed a rate, your leverage on the template collapses because AWS has already booked the concession internally and every subsequent ask looks like re-trading. Bring the clause list into the same conversation as the number, and price them explicitly against each other: "we will accept your 12% if we get a 90-day cure, a back-loaded ramp, and a closed excluded-services list." That trade is one an account manager can approve without escalation, which is exactly why it closes. If you are calibrating what the underlying rate should be before you spend your terms leverage, the tier-by-tier work in the AWS EDP discount benchmark ranges is the reference point to argue from.

The Ratchet and the 20% Growth Ask: What Uplift Numbers Buyers Are Actually Signing

AWS's standard construction does not permit the annual commitment to fall below the prior year's. A $2M year one cannot become $1.75M in year two. That is the ratchet, and it is not negotiable in most templates. What is negotiable, and what almost nobody negotiates hard enough, is the growth expectation layered on top of it. AWS account teams routinely frame the ask as roughly 20% year-over-year growth, and CloudKeeper's own program description cites approximately 20% YoY growth as a typical condition of entry. Buyers hear "20%" and mentally price it as a modest stretch. It is not. It compounds.

Run the arithmetic before you agree to anything. A $3M year one at 20% annual uplift is $3.6M in year two and $4.32M in year three: $10.92M total commit against a headline that everyone in the room has been calling a "$3M deal." That is 21% more total obligation than a flat $9M, and it is 21% more shortfall exposure. Cap the uplift at 8% and the same deal is $9.74M. The delta between an 8% cap and a 20% ask is roughly $1.18M of committed spend over three years. At a 14% discount, one extra point of rate on that deal is worth about $97K over the term. The uplift cap is worth twelve times more than the point. This is the trade buyers get backwards most often, and it is the single clearest example of why the discount headline is the wrong thing to optimize.

The stronger move is to reject percentage-based growth entirely and tie uplift to defined workload migration events instead. Frame it this way in the room: "we will commit to a step-up of $X when the ERP workload lands, and a further step-up of $Y when the data platform migrates, both subject to a 90-day grace after cutover." AWS will push back initially because event triggers are harder to forecast against, but they accept them more often than buyers expect, because the alternative is a buyer who refuses to sign any growth at all. What you are doing is converting an obligation you cannot control into one you trigger yourself. If AWS insists on a percentage, cap it. Market experience across enterprise PPA negotiations puts credible caps in the 5% to 10% band for stable estates and 12% to 15% for genuinely high-growth ones. Anything above 15% should require AWS to show you the migration pipeline that justifies it.

Two defensive asks belong in the same conversation. First, the cap must be a ceiling and not a floor: language that says "up to" rather than "no less than." Second, the ratchet should not survive a material contraction event. If you divest a business unit representing 20% of your consumption, the year-two floor should reset to reflect it. AWS resists this hard, and you will not always get it, but raising it establishes that you understand what you are signing and changes the tone of every subsequent ask. The uplift cap benchmark work elsewhere in this cluster goes deeper on where those caps actually land by tier.

Do this first: before your next call with the account team, build the three-year total commit table at AWS's proposed uplift and at your target cap, and put both numbers in front of your CFO. The conversation stops being about a discount percentage and starts being about a multi-million dollar obligation delta, which is the only frame in which you will get the mandate to hold the line.

Burn-Down Mechanics That Quietly Cost You a Point or Two

The two clauses that erode your effective discount are not in the pricing exhibit, they are in the definition of what counts as consumption against your commitment. First trap: the private pricing benefit does not retire commit. If you sign a $2M annual commit and consume $2M at list, your 10% benefit returns $200K, which means only $1.8M of gross spend actually burns down against the number you signed. You are $200K short on the very run rate you sized the deal around. Buyers who model commit off last year's net invoice rather than gross list consumption walk into a manufactured shortfall in month ten, and AWS is not obliged to warn them. The correction is arithmetic, not negotiation: size the commit off gross pre-discount consumption, then apply your growth assumption, then hold back a buffer. At a 14% discount on a $10M commit, that mechanic alone consumes roughly $1.4M of headroom per year. On a three-year deal that is a $4M gap between what you thought you promised and what you actually have to spend.

Second trap: Marketplace. AWS Marketplace ISV purchases can offset up to 25% of the annual commitment, which is a genuinely useful safety valve, but the software you route through it usually does not carry your private pricing rate. So Marketplace spend counts toward the commitment while earning you nothing on the discount line. That is a deliberate design: AWS gets credit for the transaction, the ISV gets the margin, and your blended effective rate quietly drops. A buyer at $10M commit who parks $2.5M through Marketplace and holds a 14% cross-service rate is actually earning 14% on $7.5M, or about 10.5% blended. Nobody in the room will call that out for you.

None of this argues against Marketplace. It argues for pricing it deliberately. If you are genuinely at risk of shortfall, routing existing ISV renewals (observability, security tooling, data platforms) through Marketplace is the cheapest cure available, far cheaper than writing a shortfall check for spend you never consumed. In our experience across these deals, buyers who plan Marketplace routing into year one negotiate better ISV terms too, because the ISV sees a channel deal rather than a rescue. What you should refuse is a clause that caps the offset below 25%, or that gives AWS discretion over which listings qualify. Get the 25% written as a floor entitlement, get the qualifying categories defined broadly, and get confirmation in the pricing exhibit that Marketplace dollars count at gross transaction value, not at some net figure.

• Model commit against gross list consumption, then subtract the expected benefit dollars from your burn-down capacity before you sign anything.
• Ask AWS in writing which services in your top ten by spend carry a reduced or zero private pricing rate; the answer usually reveals two or three.
• Fix the Marketplace offset at 25% of annual commit as a contractual right, with categories defined by you, not by AWS discretion.
• Pair the offset with a 30 to 90 day shortfall cure period so Marketplace routing is a live remedy rather than a theoretical one.

Proving Your Benchmark Without Handing AWS a Reason to Dismiss It

AWS has a well-drilled response to benchmark claims, and it runs in three moves. First, they ask for the source. Second, they question comparability: different tier, different term, different service mix, different region, different growth profile, different year. Third, if you survive both, the request escalates to a pricing desk that will not clear an exception against an anonymous number. That last point is the one buyers underestimate. The account manager is not your obstacle. The account manager is trying to build an internal case, and an unsourced "we hear people get 20%" gives them nothing to submit. If you want points, you have to hand your rep the paperwork that lets someone two levels up sign off without personal risk.

So present the band the way a pricing desk reads it. Tier-matched: state your own committed dollar band, not the band above it. Term-matched: a three-year number and a five-year number are different arguments and you should not blur them. Net of Enterprise Support: at roughly 3% of usage with a $15,000 monthly floor, support consumes a meaningful slice of a mid-tier discount, and quoting gross invites an easy rebuttal. Stated as a range rather than a single figure, because a range reads as observed data and a point estimate reads as a demand. And tied to a decision you are actually prepared to make. Our guidance on building an AWS discount benchmark you can defend goes deeper on constructing the evidence pack, and the cluster piece on proving a benchmark without naming the source covers how to describe comparable deals without breaching confidentiality obligations you may be carrying from prior employers or advisors.

Competitive tension is worth 2 to 4 points at the mid and upper tiers, but only in a specific form. A vague statement that you are "also talking to Azure" is worth zero and both sides know it. What moves the pricing desk is documented and time-bound: a named workload with a dollar value, a target migration window, an executed proof of concept or a signed Azure or GCP commitment covering part of your estate, and a decision date that sits before your AWS renewal. In market experience, the tension only converts when AWS can see the revenue at risk quantified. Tell them the workload is $2.4M of annual EC2 and RDS spend, that a landing zone exists on the alternative platform, and that the architecture review closes in six weeks. That is a number the desk can weigh against a discount concession. "We have options" is not.

Do this first: write a one-page benchmark memo before your next call. Line one, your gross committed spend and term. Line two, your target band as a range, net of support, tier-matched. Line three, the two or three service adders you are pursuing above the cross-service rate. Line four, the specific alternative, its dollar value, and its decision date. Line five, the three clauses (cure period, Marketplace offset, no downward ratchet on commit) you will trade rate points to secure. Send it to your account team and ask them what they need to clear it internally. You will learn more from that answer than from any published band.

Frequently asked questions

What discount should I get from AWS at $1M annual commit?

Plan for 6 to 12% cross-service, with AWS typically opening near 8 to 10% on a three-year term and closer to 10% on a one year. At this tier you are signing a largely templated agreement, so the realistic upside is in ramp shape, shortfall cure and uplift caps rather than in an extra five points of headline discount. State the number net of Enterprise Support before you take it to your CFO.

Is a 25% or 30% AWS discount realistic?

Not as a cross-service private pricing rate below roughly $25M to $50M annual commit. Figures in the 30 to 65% range almost always describe Savings Plans or Reserved Instance discounts against on-demand, which is a different mechanism that stacks underneath private pricing. Combined outcomes including service-specific adders rarely clear the mid-30s in total, and only on concentrated compute-heavy estates.

How much more discount does AWS give for five years instead of three?

Typically 4 to 6 points, which at a $10M annual commit is worth roughly $2.0M to $3.0M cumulative. That is real money, but it is also the cheapest concession AWS has, and you are buying it with two extra years of exposure to list price changes, service deprecations and your own architectural decisions. Most buyers should hold at three years unless the workload is genuinely static.

Does the AWS discount reduce my commitment burn-down?

No. The discount benefit itself does not retire commit, so a $2M spend that generates $200K of benefit only burns $1.8M against your commitment. Size your commit against net burn, not gross spend, or you will be short at true-up through arithmetic alone.

Can Marketplace spend help me hit my AWS commitment?

Yes, up to roughly 25% of the annual commitment in most agreements, and that makes Marketplace routing a real lever for a buyer at shortfall risk. The catch is that Marketplace spend usually contributes at face value without earning the private pricing rate, so it protects you from a shortfall bill but does not improve your effective discount.

When should I start the AWS private pricing negotiation?

120 to 180 days before the term date. AWS discount approvals above the standard band need regional leadership sign-off, which takes weeks, and a buyer who is already inside 30 days or already carrying shortfall exposure has almost no leverage regardless of spend tier.