① Journal · Cost trends
A few percent of extra capacity a year, with no project behind it, is the quietest line item on the mainframe. By 2026 it compounds against consumption pricing and Tailored Fit baselines that lock the higher number in. Here is where the creep comes from and how the buyer side stops it billing.
Nobody approves MIPS creep. Everybody pays for it.
Every mainframe estate has a number that goes up on its own. New application releases run a little heavier than the ones they replace. Batch windows stretch as data volumes grow. A reporting job that used to finish before the morning peak now overlaps it. None of these has a business case, a budget line, or an approval. Added together across a year they lift consumed capacity by a few percent, and on the mainframe that capacity is what the software is billed on.
This is MIPS creep, and in 2026 it bites harder than it used to. Under sub-capacity pricing the recurring monthly license charge follows the rolling four hour average peak, so creep that raises the peak raises the bill directly. Under Tailored Fit Pricing the consumption baseline is set from recent usage, so a baseline inflated by creep gets written into the agreement and into every annual reconciliation that follows. The creep does not just cost more this year. It resets the floor for every year after. Read this alongside our sub-capacity vs full capacity explainer and our cost optimization service.
Common sources of unmanaged capacity growth · none carry a project or a budget line
| Source | How it shows up | Why it bills |
|---|---|---|
| Software upgrades | New releases of middleware and applications run heavier than the versions they replace | More service units consumed for the same business work |
| Data growth | Tables, indexes, and files grow, so the same jobs process more rows | Longer, heavier jobs lift the peak window |
| Batch overlap | Jobs that once finished before the online peak now run into it | Two demands stack inside the same rolling four hour average |
| Inefficient code | Unoptimized SQL, loops, and sorts ship without a capacity review | Pure waste consumed at full meter rate |
| New interfaces | APIs and real time feeds add transaction volume to the online day | Higher sustained R4HA across more of the day |
| Defensive capacity | Headroom added after one bad month and never given back | Installed capacity and baselines ratchet up, never down |
Growth rates vary widely by estate and workload mix; the pattern, not a fixed percentage, is the point. Where creep runs unmeasured, a few percent a year compounds into a structurally higher baseline within a renewal cycle.
The danger is not the year. It is the compounding.
Treat the peak that sets your bill as a baseline of 100 units. Let unmanaged creep add a modest seven percent a year. After one year the baseline is 107. After three it is roughly 122. After a five year renewal cycle it is around 140, a forty percent lift in the number your software is billed on, with not one approved growth project behind it. The same compounding that makes the mainframe reliable makes its cost base drift if nobody is watching the meter.
The reason this hides so well is that no single month looks alarming. The invoice rises a little, the capacity report nods along, and the increase is absorbed as the cost of doing business. It only becomes visible at the renewal, when the vendor anchors the new term to a baseline that years of creep have quietly inflated, then negotiates an uplift on top of it. By then you are arguing about a number you helped raise. The discipline is to manage the meter every year, not to discover it at renewal. Our capacity planning with software cost guide puts the cost back into the model.
The whole estate average is the wrong number. Find the few peak hours that set the rolling four hour average each month and attribute that consumption to named workloads. Capacity you cannot attribute is capacity you cannot defend or reduce.
You cannot manage a meter you do not read.
Where the peak is driven by deferrable work, defined and group capacity limits hold the R4HA where it pays to without starving the online day. Capping is a buyer side cost control, not a performance compromise, when it is aimed at the right window.
Hold the peak the bill is calculated on.
Batch that runs into the online peak stacks two demands inside the same four hour window. Move the deferrable jobs out of it and the peak falls without removing a single unit of real work. The cheapest MSU is the one you never put on the peak.
Spread the demand, lower the meter.
Never let a renewal or a Tailored Fit baseline be set from a peak that years of creep inflated. Reduce the peak first, then anchor the agreement to the cleaned number. A baseline signed dirty is overpaid for the life of the term.
Reduce first, then lock the number.
⑤ The discipline that pays
Creep is invisible until the renewal prices it. By then you are paying for capacity nobody chose. Read the meter every year, not at the table.
Typical reduction negotiated on renewal spend
Mainframe spend negotiated on the buyer side
Engagements delivered since 2019
The steady rise in consumed capacity that happens with no growth project behind it. Heavier releases, longer batch, inefficient code, and added headroom that never comes back lift the rolling four hour average a few percent a year. Nobody approved it, but the bill follows it.
Sub-capacity bills on the R4HA peak, so creep that raises the peak raises the charge directly. Under Tailored Fit Pricing the baseline is set from recent usage, so creep inflates the committed number and every annual reconciliation after it.
At a modest seven percent a year, a baseline of 100 reaches around 140 over five years, a forty percent lift with no approved project. The compounding, not any single month, is what makes creep expensive.
Measure the peak hours that set the charge, soft cap the costly window, reschedule deferrable batch off the peak, and clean the baseline before signing any renewal or Tailored Fit commitment. See our cost optimization service.
Related: soft capping wins · sub-capacity vs full capacity · capacity planning with software cost · cost optimization service
Audit notice or renewal under 18 months out? We mobilize within 48 hours.
Get expert help →