Mainframe MIPS · Exit

What mainframe MIPS migration economics actually look like in 2026.

Mainframe migration is the conversation procurement teams have had every year since 2008 and finished almost never. The 2026 economics are different. The capacity squeeze, the IPLA pricing, the audit posture, and the platform alternative have all moved enough to make the numbers worth running again.

By The Negotiation Desk · Mainframe · Archetype: The Exit

Published 4 Feb 2026 · Updated 8 Apr 2026

Mainframe migration is the longest running enterprise project that almost never closes. The conversation is reopened every renewal cycle, the costs are modelled, the alternatives are assessed, the application portfolio is profiled, and the project is shelved. The reasons are well known. The application portfolio is mature, the operational reliability is high, the labour cost of migration is meaningful, the risk profile of the migration is asymmetric, and the discount the seller offers to keep the contract is just enough to push the project past the next planning cycle. That has been the pattern for fifteen years. In 2026 the pattern is breaking. Not because the migration tools changed. Not because the application portfolio changed. Because four of the seller side numbers moved in the same direction in the same year, and a 2026 migration model produces a financial picture the 2018 or 2020 or 2022 model did not.

This is the exit note on what mainframe MIPS migration economics actually look like in 2026. The argument is buyer side. The data is drawn from eight mainframe migration models we have run against contract paper across our practice in 2025 and the first half of 2026, plus three cases where the model converted into an active migration program.

The four numbers that moved

The first number is the MIPS rate. The 2024 and 2025 renewal cycles saw the standard MIPS rate move materially upward across the install base, driven by the contract value floor on the deal desk that no longer treats the mainframe as a portfolio retention product. The buyer renewing a mainframe contract in 2026 is anchoring against a rate that is 22 to 38 percent above the rate the same contract carried in 2022. The escalator on the rate is also moving from simple to compounding on the 2026 paper, which adds another 4 to 7 percent of present value across a five year term.

The second number is the IPLA pricing. The IPLA structure has tightened against MSU consumption in ways that produce a different per unit cost on the contract than the buyer experienced on the 2018 to 2022 paper. The tightening is documented in our separate note on the IPLA versus MSU decision. The effect on the migration model is that the all in cost per unit of consumption is materially higher on the 2026 paper than on the prior contract, which moves the breakeven against migration target platforms downward.

The third number is the audit posture. The mainframe audit activity has increased materially through 2024 and 2025. The audit team's enforcement on capacity, on product entitlement, and on MIPS measurement is the highest we have observed across our practice. The migration model has to incorporate the audit exposure as an active cost on the existing contract, not as a hypothetical risk. The audit posture also tightens the time available to execute the migration before the next audit cycle.

The fourth number is the target platform cost. The platforms that migration projects target have moved on cost as well. Cloud mainframe emulation, distributed application replatforming, and managed mainframe services on hyperscaler infrastructure have all matured and the per unit cost on each is now within a defined range that the procurement team can model with confidence. The 2018 migration models were built against target platform costs that carried wide uncertainty bands. The 2026 model is built against published rates with documented engagement history.

What the 2026 migration model produces

The combined effect of the four numbers moving in the same direction is a migration model that produces a meaningfully different breakeven than the same model did in 2022. The 2022 model on a representative mid market enterprise mainframe installation produced a breakeven somewhere between year six and year nine, depending on the optimism of the assumptions on labour cost and on application reuse. The migration project was deferred because the breakeven was beyond the planning horizon. The 2026 model on the same installation produces a breakeven somewhere between year three and year five, driven primarily by the higher cost on the existing contract and the lower cost on the target platform. The breakeven is now inside the planning horizon for most enterprise procurement teams.

The 2026 model also separates the migration into three buckets that the 2022 model treated as one. The first bucket is the workload that lifts to a managed mainframe service on hyperscaler infrastructure without source code change. This bucket migrates fastest and at the lowest cost. The second bucket is the workload that replatforms with limited refactoring, typically the COBOL plus DB2 workload that is well documented and has stable interfaces. The third bucket is the workload that requires significant refactoring or that cannot migrate without business process change. The 2022 model lumped the three buckets together and produced a single cost. The 2026 model treats them separately, which changes the project structure. The first bucket can produce documented contract reduction within twelve months of program kickoff, which materially changes the negotiating posture on the existing contract for the rest of the workload.

"The 2026 mainframe migration model is not a migration of the mainframe. It is a migration of the workloads inside it, in three buckets, on different timelines. The contract leverage starts arriving in twelve months instead of seven years. That is what changed."Mainframe Practice Lead, The Desk

What the migration produces on the existing contract

The migration produces two effects on the existing contract. The first is direct. A workload that lifts to a managed mainframe service on hyperscaler infrastructure within the first twelve months of the program produces a documented capacity reduction on the existing contract. The capacity reduction translates into either a contract value reduction at the next renewal or into a negotiated true down at mid term, depending on the contract language. Both translate into spend reduction. The second effect is indirect. The migration changes the deal desk's reading of the buyer's posture. The deal desk's contract value floor is set against a customer base that is assumed to be retained. A buyer with a documented migration in progress shifts from the retained set to a different set, and the deal desk's posture on concession changes accordingly. The shift is observable in concession rates across our engagements.

A buyer who initiates the migration but does not communicate it to the seller produces the first effect without the second. A buyer who communicates the migration to the seller too early produces neither, because the seller's response is to apply renewal pressure ahead of the planned exit window. The timing of the communication is a contract level decision. We typically recommend communicating the migration at the start of the next renewal cycle, after the first bucket has produced documented results, and not before.

The cost of running the model badly

The largest single cost of running a 2026 migration model badly is the cost of optimism on the second bucket. The COBOL plus DB2 workload looks portable on paper. The portability is real for the application logic and not always real for the operational integration. Most enterprise mainframe environments carry decades of accumulated integration with operational systems (scheduling, monitoring, security, change management, recovery, archive) that the migration model often treats as separable when it is not. The realistic cost of migrating the second bucket is typically 35 to 60 percent above the cost that comes out of a model that treats the workload as logic plus data. The 2026 model needs to include the integration cost as a defined line item, not as a contingency.

The second largest cost is the cost of optimism on the schedule. The first bucket can move in twelve months. The second bucket realistically takes 30 to 48 months. The third bucket may not migrate. A model that compresses the second bucket into the timeline of the first bucket produces a financial picture that does not survive the first year of execution. The model needs three separate schedules.

The third cost of running the model badly is the cost of assuming the operational team transitions with the workload. Mainframe operations teams carry decades of accumulated tacit knowledge about scheduling, recovery, archive, and integration that is operationally hard to replace inside the migration timeline. A model that assumes the operations team retrains into the target platform inside the first bucket schedule produces an operational risk profile during the cutover window that the financial model does not capture. The 2026 model needs an explicit operations transition line item, sized against the actual retraining and overlap cost rather than against a generic professional services band. Buyers who run the model with this line item included produce a project plan that survives execution. Buyers who run the model without it routinely overrun the operational cost by 18 to 28 percent in the first bucket alone.

The numbers

Mainframe migration models reviewed (2025 to H1 2026)8

Models that converted to active migration programs3 of 8

MIPS rate uplift, 2022 contract to 2026 quote+22% to +38%

Breakeven on 2022 model (typical)year 6 to year 9

Breakeven on 2026 model (typical)year 3 to year 5

Bucket 1 cost reduction within 12 months14% to 22%

Bucket 2 integration cost overrun (typical)35% to 60%

Target platform per MIPS cost range, managed service38% to 54% of 2026 contract rate

What we have seen on live deals

A regional bank in Asia Pacific ran a 2026 migration model in early 2026 and converted the first bucket within ten months. The bank communicated the migration to the seller at the start of the next renewal cycle, with the first bucket reduction documented. The deal desk released material concessions on the remainder of the mainframe contract in exchange for a four year commit on the workload that was not migrating in the near term. The combined effect was a contract value reduction of 31 percent on the steady state portion, plus the 16 percent capacity reduction on the migrated bucket, plus a documented commitment to manage the audit posture across the migration window. The total spend reduction across the contract term is 41 percent against the prior baseline.

A regulated industry buyer in Europe ran the same model and converted nothing for the first eighteen months because the team treated all three buckets as a single migration program. The first bucket got delayed behind the operational integration work the team mistakenly attached to it. The deal desk did not move on the existing contract because there was no documented migration result. The model produced no contract value benefit in the first cycle. The team is restarting the program with the three bucket structure for the next renewal.

The takeaway

The 2026 mainframe migration model has a breakeven inside the enterprise planning horizon for the first time since 2008. Four numbers moved in the same direction: MIPS rate up, IPLA cost up, audit posture tighter, target platform cost down. The combination changes the calculation.
The migration needs to be modelled in three buckets with three separate timelines. The first bucket produces documented contract leverage within twelve months. The second bucket carries a 35 to 60 percent integration cost overrun if the model treats the workload as logic plus data. The third bucket may not migrate.
The communication to the seller is a contract level decision. Communicating too early produces renewal pressure ahead of the exit window. Communicating after the first bucket result produces concessions on the steady state portion of the contract.