IBM's own 2026 pricing deck tells sellers increases are 'often achieved through a combination of list price increases and renewal uplifts', which is why a single-trigger 5% cap leaves roughly 11% of annual increase unbound
A cap that only binds the renewal percentage does nothing when IBM moves the 6% global list price underneath it, and IBM knows this because its own harmonization materials describe the two levers as a combination. Buyers running best-in-class deals are landing 0 to 3% annual caps bound to both triggers, against IBM's working default of 5 to 7%. The difference on a $1M S&S line is roughly $210,000 over three years, and it is decided by about forty words of clause language you either write now or pay for later.
Prepared by Redress Compliance · August 25, 2026 · IBM advisory. ELA, Passport Advantage, and Cloud Pak renewals, 2024 to 2026.
Executive summary
The single most valuable sentence in this negotiation is IBM's own: increases are "often achieved through a combination of list price increases and renewal uplifts", language circulated to client teams in the 2026 Price Harmonization materials.
Read to a rep, it collapses the standard objection that a 5% renewal cap already protects you, because the January 2026 harmonization moved list roughly 6% globally (IBM i SWMA at 10%) entirely outside that cap.
IBM's working default is the maximum permitted increase every year, commonly 5 to 7% on S&S renewals, which compounds to about 16% at 5% and about 23% at 7% across three years on flat usage.
On a $1,000,000 support line that is $160,000 to $230,000 of pure escalation before a single new workload lands, and benchmarked uncapped renewals have been arriving at 8 to 12%.
Best-in-class outcomes are 0 to 3% annual caps, with some strategic accounts holding flat 0% for the first two to three years, and 0 to 4% is the realistic ELA band.
The gap between IBM's opening 7% and a negotiated 3% is worth roughly $125,000 over three years per $1M of annual support, which is usually larger than the license discount concession the account team is offering you instead.
When IBM refuses a hard cap, the tradeable fallbacks are worth more than the cap itself in many estates: true-up at contracted discount rather than spot list, swap rights with a defined PVU exchange ratio, and true-down at renewal to certified deployment.
Without true-down, IBM renews against committed volume rather than measured usage, so an inflated baseline carries forward and the uplift is applied to entitlement you never deployed.
Support fees are frequently calculated on original list price rather than the discounted price you actually paid, which is why a 40% license discount can coexist with an S&S line that behaves as though you paid list.
Any cap must therefore name the fee basis explicitly, or IBM's 20%-of-SRP support trajectory quietly erases the discount you negotiated at signature.
How IBM delivers an increase: two levers, one cap, and the gap between them
The reason most negotiated caps underdeliver is that they bind the wrong number. A renewal uplift cap governs the percentage IBM applies to your prior-term fee. It says nothing about the price IBM starts from.
When IBM moved global list roughly 6% effective January 2026 (10% on IBM i SWMA, 12% on selected appliances, 3% rising toward 7.5% on Cloud services), it moved the base under every cap in the installed base without touching a single contract.
Layer in Red Hat's 10% April 2025 increase on EUR and GBP subscriptions and the OpenShift substrate under your Cloud Paks moved too.
Meanwhile the fee basis itself is working against you: S&S typically runs 15 to 22% of license value, most commonly 17 to 20%, and it is frequently calculated on original list price rather than the discounted price you actually paid.
A 40% license discount that leaves support pegged to list gives IBM a compounding line it fully controls. The notice window is the last piece: roughly three months, and in the AD24-2253 cycle customers got the email on October 29 for a letter dated September 3.
That is not enough runway to build an alternative, which is precisely why the clause has to exist before the letter arrives.
| Lever | What it moves | 2026 observed | Does a standard uplift cap bind it? | Clause element required |
|---|---|---|---|---|
| Global list harmonization | The base your uplift multiplies | ~6% most software, 10% IBM i SWMA, 12% appliances | No | Cap the aggregate fee payable, not the percentage |
| Renewal uplift | The percentage applied at renewal | 5 to 7% default, 8 to 12% uncapped | Yes, if drafted | Explicit 0 to 3% annual ceiling |
| S&S fee basis | Ratio of support to license | 17 to 20% typical, 22 to 25% premium | No | Basis fixed to price actually paid in prior term |
| SKU reclassification | Which price list applies | Ongoing across PA catalog | No | Successor and renamed SKU binding |
| Notice timing | Your ability to respond | ~3 months, sometimes less in practice | No | Notice floor plus post-expiry extension right |
The table cannot show the multiplication. The two levers do not add, they compound.
A 5% single-trigger cap sitting on top of a 6% list move produces an effective increase of roughly 11.3% on the line, and if your S&S is pegged to list rather than net, the list move hits the full undiscounted number while your cap only governs the smaller one.
Over three years at flat usage, that is the difference between a 16% cumulative increase and something in the mid-30s.
IBM's own guidance to sellers describes the two as a combination, which means the gap is not an accident of drafting. It is the design.
Anyone who has negotiated the same clause with the same account team across multiple cycles will recognize the pattern: IBM concedes the visible number and keeps the invisible one.
The clause language that actually binds, word by word
Start by discarding the wording most buyers arrive with.
"Support fees shall not increase by more than 5% annually in years 2 and 3" fails on four counts: it caps a percentage rather than an amount, it does not name the fee basis, it dies at renewal boundary, and it says nothing about SKUs IBM renames.
The construction that holds reads closer to this: the aggregate fees payable by Customer for Subscription and Support in each renewal term shall not exceed the aggregate fees payable in the immediately preceding term by more than three percent (3%), regardless of any change to IBM's list.
Suggested retail, or published pricing, and regardless of any reclassification, renaming, bundling, or succession of the underlying Part Numbers.
Four phrases carry the load. "Aggregate fees payable" caps a dollar figure, so a list move underneath produces no effect.
"Fees payable in the immediately preceding term" names the price you actually paid, closing the list-versus-net basis gap that IBM otherwise uses to charge support on the undiscounted number. "Regardless of any change to IBM's list" is the dual-trigger sentence and the one IBM will fight hardest.
"Reclassification, renaming, bundling, or succession" stops the oldest workaround in the book, where the capped Part Number is retired and its successor arrives uncapped at full price.
Two structural additions matter as much as the percentage.
First, close the year-four hole with a post-expiry extension right at predefined pricing, otherwise your carefully drafted three-year cap simply expires into IBM's then-current pricing and the whole exercise buys you a delay rather than a ceiling.
Second, consider a multi-year aggregate formulation, "no more than 10% total across the three-year renewal term," which is often easier for IBM to approve than a hard annual number because it lets the seller shape the internal forecast, and it removes the compounding entirely.
On a $1M S&S line, 10% aggregate beats 5% annual by roughly $60,000 across the term.
Treat any CPI tie as a concession, not a win. CPI clauses fail in two directions: they are almost always unbounded on the upside, and IBM keeps list-price movement outside the index, so you have accepted an inflation-linked uplift stacked on top of a harmonization increase.
If IBM insists on CPI, bound it at CPI or 3%, whichever is lower, and pull the dual-trigger sentence forward so the index applies to the aggregate.
The broader clause set that surrounds this (true-up terms, swap rights, exit pricing) is covered in our work on the IBM ELA redlines that decide the next three years, and the cap should be negotiated as part of that package rather than in isolation, because IBM will trade against it.
Cut your IBM ELA renewal with 8 buyer side levers
Eight buyer side levers that cut an IBM ELA renewal: the baseline reset, true forward exposure, product rationalization, and the Cloud Pak shift.
Get the white paper →Why IBM defends the uplift harder than the discount, and how to use that
Sit across from an IBM client team long enough and you notice a pattern that has nothing to do with the products: the rep will fight you for hours over three points of license discount and hand it over in the last week.
Then dig in over two points of renewal cap as if the account's existence depends on it.
It does. Signature-year discount is a quota problem, absorbed once, forgotten by the next fiscal year. The uplift is where the multi-year value of the account is manufactured, and it is the number the territory plan, the account forecast, and the seller's own compensation on future S&S all rest on.
When a rep tells you the cap is "not something we do," what they mean is the cap is the only concession that costs them beyond this quarter.
Run the arithmetic and it is obvious why. On a $1M S&S line, IBM's working default of 5 to 7% annually compounds to roughly 16% over three years at the low end and about 23% at the high end, which is $160,000 to $230,000 of unbudgeted spend at completely flat usage.
Compare that to what an extra three points of license discount is worth on a one-time $2M license purchase: $60,000, once. The uplift is almost always the larger number, and it is the number nobody in procurement has modeled because it does not appear on the signature page.
IBM's team has modeled it. That asymmetry of preparation is the whole negotiation.
Read IBM's stated trajectory as the admission it is.
The direction of travel documented since 2020 is raising S&S by up to 10% annually until support equals 20% of SRP list price, alongside the Passport Advantage change that put over five thousand on-premises SKUs at SRP with S&S pegged at 20% of list. That is not a support-cost recovery story.
That is a business model where the compounding line is the product and the license transaction is the customer acquisition cost.
Once you accept that framing, the negotiation stops being about what you pay to acquire IBM software and becomes about what you pay to keep it, which is the only number that recurs.
The 2026 harmonization materials confirm this is a managed campaign rather than a market condition.
Sellers are told increases are "often achieved through a combination of list price increases and renewal uplifts," given scripted objection handling including the "other companies are raising prices" line.
And coached to run urgency plays so clients buy "this year at lower prices" before the increases take effect.
Market conditions do not come with talking points and FAQs. A campaign does. That distinction matters at the table because it tells you the 6% global increase and the 5 to 7% uplift demand are policy positions with internal exception paths, not physics, and exception paths exist to be used.
So reframe. Walk in and make the cap the headline concession you are asking for and the discount the giveback you are offering.
Say plainly that you will accept a lower license discount, three to five points below your opening ask, in exchange for a dual-trigger cap at 0 to 3% bound to both list movement and renewal percentage. The rep hears a deal that protects signature-year revenue, which is what they are measured on.
You hear a deal that protects the compounding line, which is what actually costs you. Both sides can sell that internally, and the trade only works if you name it before the discount conversation closes, because after signature the discount is spent and you have nothing left to trade.
The buyers who lose this are the ones who treat the cap as a legal cleanup item raised in the final redline pass. By then IBM has banked the discount concession, the quarter is closing, and the only lever left is your signature.
The cap belongs in the first commercial conversation, priced explicitly, next to the discount ask, so IBM's team has to choose which one they want. They will choose the discount, every time. That is the outcome you wanted.
For the wider set of terms that move with it, the IBM ELA clauses that decide the next three years operate as a package, not as isolated redlines.
Fallbacks that outperform a cap when IBM says no
When IBM refuses a hard dual-trigger cap, do not settle for a softer cap. Trade for structural terms that reduce the base the uplift applies to, because a smaller base beats a smaller percentage in most over-entitled estates.
The sequence that works: true-up at the contracted discount band rather than spot or list, swap and substitution rights with a written PVU exchange ratio, true-down at renewal to ILMT-certified deployment.
And a fixed exit or extension price so year four does not become "then-current pricing." Understand the trap first.
IBM renews against committed volume, not measured usage, and unused entitlement rolls forward as billable baseline. In an estate carrying 20 to 30% shelfware, which is common in our engagements, a true-down right removes more cost in one renewal than a 3% cap saves across three years.
| Fallback term | What it binds | Value on a $1M S&S line |
|---|---|---|
| True-down to ILMT-certified deployment | Renewal base, not the percentage | $200,000 to $300,000 at 20 to 30% shelfware |
| True-up at contracted discount band | Mid-term growth priced off your discount, not list | Avoids list-price delta on every addition |
| Swap rights with fixed PVU exchange ratio | Migration cost between products and Cloud Paks | Prevents re-purchase at 2026 harmonized list |
| Fixed extension or exit price | Year four, the uncapped year | Removes the post-term jump entirely |
| 3% annual cap (for comparison) | Percentage only, base untouched | Roughly $90,000 over three years |
The table makes the case that IBM's own sellers understand: capping the rate while leaving the base uncontested protects the smaller number. A 3% cap on $1M saves roughly $90,000 across three years.
A true-down right on the same line, in an estate running 25% shelfware, removes $250,000 from the base before any percentage is applied, and the saving compounds because every future uplift is calculated on the reduced figure.
Ask for all four, then concede the ones you can live without.
Swap rights matter most if a Cloud Pak consolidation is likely, and the exchange ratio must be written now, not referenced as "IBM's then-current conversion table." The forthcoming Redress guidance on Cloud Pak swap rights and on termination for convenience covers those two mechanics in detail.
Bring your ILMT data to the first meeting: without certified deployment numbers, true-down is an argument you cannot win.
Program-level resets that defeat a well-drafted cap
A cap that reads perfectly can still be defeated by machinery that lives outside the contract, and IBM knows exactly which levers those are. The first is the Passport Advantage RSVP level, recalculated at each anniversary against the trailing twelve months of point accumulation.
A quiet year of purchasing, a divested business unit, a paused project, and the account drops a tier.
Depending on where you sit in the band structure, a tier crossing is worth roughly 5 to 15 discount points on subsequent orders, which means a 3% capped renewal can land as a double-digit effective increase on the invoice while IBM correctly insists the cap was honored.
The cap governed the uplift percentage. It never governed the discount level the uplift was applied to. Second lever: lapsed S&S reinstatement, where IBM charges back the gap period plus a penalty, and reinstated lines re-enter at current list rather than your legacy entitlement.
Third, and increasingly the largest, is the Red Hat OpenShift subscription sitting underneath every Cloud Pak. That layer prices independently, and Red Hat pushed 10% on EUR and GBP subscriptions in April 2025 on FX grounds. Your Cloud Pak cap does not reach it.
The fix is one clause and about twenty words: the cap must reference the entitled discount level, RSVP tier, and Suggested Volume Price band in force at signature.
Held for the term regardless of subsequent point accumulation, with any component priced by a third party (Red Hat included) either brought inside the cap or carved out with its own ceiling.
Expect IBM to resist the RSVP hold hardest, because that is the mechanism that quietly re-prices flat estates. Fold this into your broader ELA redline set rather than raising it as a standalone ask.
What the 2024 to 2026 cycle actually shows
AD24-2253 carried 6% worldwide while IBM claimed most of 31,053 line items moved within plus or minus 5%.
Power IBM i SWMA at 10%, selected appliances at 12%, Red Hat EUR and GBP at 10%, all under a "6%" headline.
The pattern across two harmonization cycles is consistent enough to negotiate against.
The headline number is real but it is an average, and the average conceals a wide distribution that IBM itself acknowledged in 2024 when it claimed most of 31,053 line items would fall within plus or minus 5% of the stated 6%.
January 2026 repeated the structure: 6% across Passport Advantage distributed software, SaaS, and mainframe MLC and OTC, but Power IBM i SWMA at 10%, PowerVS at 2%, IBM Cloud services at 3% with published further movement to 7.5%, selected appliances at 12%.
And Red Hat at 10% for euro and sterling buyers.
If your estate is Power heavy or appliance heavy, the number that matters to you is not 6%. Price your actual line items against the letter, not the summary slide the account team brings.
Two timing facts change your negotiation posture. IBM typically gives around three months of notice, and in the 2024 cycle the customer email landed October 29 against a letter dated September 3, which means the practical window is shorter than the formal one.
Second, sellers are explicitly coached to run pre-increase urgency plays, encouraging purchases to close before the effective date. Treat every urgency claim as a testable assertion.
The March 2026 report of a 25% across-the-board increase on hardware, software, and services was flagged as low credence by the publication that carried it and remains unconfirmed.
If an IBM rep cites a number, ask for the announcement letter identifier and the line-item page for your SKUs before you move a signature date by so much as a week.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
Your first five moves
- Model the compounded exposure before you take the meeting. Run your actual S&S base at 7%, 5%, and 3% annually: 5% compounds to roughly 16% over three years and 7% to about 23%, so on a $1M line you walk in knowing the spread between IBM's default and your target is around $210,000, not an abstraction.
- Put IBM's own harmonization language into the negotiation record. Retrieve the 2026 Price Harmonization materials, read the sentence stating increases are "often achieved through a combination of list price increases and renewal uplifts" back to the account team in writing, and ask them to confirm or deny it, because once it is on the record a single-trigger cap is no longer defensible as standard commercial terms.
- Open with dual-trigger cap wording, not a percentage debate. Table the redline at 0 to 3% annual, binding both the renewal uplift and any change in the underlying list price or S&S calculation basis, and hold the drafting itself as the opening position so IBM has to negotiate against your paper rather than its own.
- Price the fallback bundle so the cap is tradeable. Value the true-down right, the swap ratio, and true-up at contracted discount in dollars before IBM refuses the hard cap, so you can concede two points of uplift for a package you have already scored, a discipline covered further in the IBM ELA clause redlines work.
- Set your signature date ninety days clear of the notice window. IBM typically gives about three months' notice on increases and its sellers are coached to run buy-now urgency plays, so a date with slack removes the only lever they have left.
Frequently asked questions
What annual uplift cap should I actually target with IBM?
Target 0 to 3% for enterprise deals and treat 0 to 4% as the realistic ELA band. IBM's working default is 5 to 7% and benchmarked uncapped renewals have arrived at 8 to 12%, so the opening number is a starting position rather than a policy.
Some strategic accounts hold flat 0% for the first two or three years, which is worth asking for explicitly before conceding to a percentage.
Why does a 5% cap on renewal increases not protect me?
Because it binds only one of two levers. IBM's own 2026 harmonization materials describe increases as achieved through a combination of list price increases and renewal uplifts, and the January 2026 harmonization moved list roughly 6% globally.
If your cap governs the uplift percentage but the list price underneath it moves independently, your effective increase can land near 11% while IBM remains contractually compliant.
Does a big license discount protect me from support increases?
Usually not. Support fees are frequently calculated on original list price rather than the discounted price you paid, and IBM's stated direction has been to move S&S toward 20% of SRP list.
A 40% license discount can therefore sit alongside an S&S line that behaves as if you paid list, which is why the cap clause must name the fee basis as the amount actually paid in the prior term.
What is the year-four hole in a standard cap clause?
Most drafted caps cover years two and three of a three-year term and then expire, leaving the next renewal exposed to IBM's then-current pricing.
Close it by adding a post-expiry extension option at a predefined price, typically the final capped year plus the same cap percentage, exercisable at your election. Without it, the entire benefit of a three-year cap can be reversed in a single renewal cycle.
What should I ask for if IBM refuses a hard cap?
Trade for true-down at renewal to certified deployment, true-up at your contracted discount band rather than spot or list price, and swap rights with a defined PVU exchange ratio.
In an over-entitled estate, true-down alone often beats a 3% cap, because IBM renews against committed volume rather than measured usage and unused entitlement otherwise rolls into the next term.
How much notice does IBM give before a price increase?
Roughly three months is typical, though the practical notice can be shorter than the letter date suggests. In the AD24-2253 cycle, customers received email notification on October 29 of a letter dated September 3.
Build the notice period into your clause explicitly and set your signature date at least ninety days clear of any announced effective date so the urgency argument has no grip.
Should I tie the cap to CPI instead of a fixed percentage?
Only with a hard ceiling on top. An unbounded CPI tie transfers macro risk to you and leaves list price movement outside the mechanism entirely. If IBM pushes CPI, accept it as the lower of CPI or a fixed 3%, and make the clause bind the aggregate fee payable rather than the uplift percentage.
