① Journal · Benchmarks
A typical estate runs 600 to 2,500 MSUs at peak and pays $15,000 to $80,000 or more a month in MLC charges. But cost per MSU is not a list price, it is a curve. Here is how the number actually breaks down, and what moves it.
There is no single price per MSU. There is a curve, and your position on it is negotiable.
Buyers ask for a benchmark figure and the honest answer is a range with a shape. A typical enterprise estate consumes somewhere between 600 and 2,500 MSUs at its measured peak, and the resulting Monthly License Charges land roughly between $15,000 and $80,000 or more, before the IPLA and one time products that sit alongside them. The reason the range is so wide is that MSU cost is not a list price multiplied by a count. It is the product of the rolling four hour average peak, the declining tier curve that prices each MLC product, the specific mix of products in the LPAR, the sub-capacity discipline applied, and the discount negotiated at the last renewal.
That structure has a useful consequence: the effective cost per MSU falls as volume rises, because most MLC products price on tiers that get cheaper at the margin. A large estate pays a lower marginal rate than a small one, and the same estate pays less per MSU when workloads are consolidated into the lower tiers than when they are scattered. So the benchmark that matters is not a dollar figure copied from someone else's contract, it is the shape of your own curve. Read this with our explainer on MIPS and MSU and our note on MIPS creep in 2026.
Directional ranges buyers report · 2026 · MLC software, excluding one time IPLA
| Estate size at peak | Typical monthly MLC band | What drives the spread |
|---|---|---|
| Small · under 600 MSU | Lower volume, higher marginal rate per MSU | Workloads sit in the costly upper tiers of the curve |
| Mid · 600 to 1,200 MSU | Roughly $15K to $40K range | Product mix and SCRT discipline swing the figure most |
| Large · 1,200 to 2,500 MSU | Roughly $40K to $80K+ range | Tier curve and negotiated discount dominate |
| Very large · 2,500 MSU and up | Lowest marginal rate, largest absolute bill | Consumption model terms and the floor become the question |
Directional ranges drawn from publicly reported figures and patterns we commonly observe, not a price list. Bands overlap and depend on machine generation, product mix, sub-capacity position, and negotiated terms. Your SCRT data and contract govern the real number.
The rolling four hour average peak drives the bill, so every MSU you keep out of that peak is one you do not pay for. Soft capping and workload timing pull non urgent batch and reporting jobs off the peak window. This is the single largest lever, and it works before any renewal conversation begins.
You pay for the peak, so flatten the peak.
Because the tier curve gets cheaper at the margin, scattered workloads pay more per MSU than the same workloads consolidated. Reviewing how LPARs and products map to the tier structure often recovers cost with no reduction in capacity, simply by landing volume where the marginal rate is lower.
Same MSUs, lower tier, smaller bill.
You should never pay on a peak you have not independently checked. SCRT output drives the charge, and small measurement errors compound across a large estate. Validating the report each month removes inflated peaks and gives you defensible data for the renewal, where the per MSU rate and any consumption baseline are actually set.
Never pay on an unverified peak.
④ Where the per MSU number is won
Cost per MSU is not a price, it is a curve. Lower the peak, consolidate the tiers, verify the data. Then negotiate from where you actually sit on it.
Typical reduction negotiated on renewal spend
Mainframe spend negotiated on the buyer side
Engagements delivered since 2019
No single number. A typical estate runs 600 to 2,500 MSUs at peak and pays roughly $15,000 to $80,000 or more a month in MLC charges, depending on machine generation, product mix, and terms. The effective rate falls as volume rises because MLC products price on declining tiers. Any benchmark is directional; your contract and consumption govern.
Because it is the product of the measured peak, the tier curve per product, the product mix, sub-capacity discipline, and the negotiated discount. Two estates of identical MSU size can pay very different effective rates. The useful benchmark is the shape of the curve, not a borrowed dollar figure. See MIPS explained.
Reducing the measured peak through soft capping and workload timing, then consolidating workloads into the lower marginal tiers and validating SCRT so you never pay on an inflated peak. Hardware refresh helps only when paired with a software conversation. See MSU optimization quick wins.
Only on a baseline built from real consumption. Consumption pricing can lower the marginal rate for a growing estate but commits you to a floor. Model it on validated data, not the vendor projection. Our cost optimization service builds the curve and our license negotiation sets the rate.
Related: mainframe cost optimization · MIPS explained · MSU optimization quick wins · the hidden cost of MIPS creep · IBM publisher hub
Audit notice or renewal under 18 months out? We mobilize within 48 hours.
Get expert help →