① Licensing concept · Exit economics
Most full mainframe exits overrun their budget, slip their timeline, or revert. The honest economics matter for two reasons: deciding whether to leave, and knowing how much a credible exit option is worth at the renewal table.
The exit is rarely the cheap path. The credible option to exit usually is.
The mainframe exit question is whether to migrate workloads off the platform to cut licensing and operating cost, and if so, how. It is a fair question, because mainframe software bills are large and the vendors know switching is hard. But the buyer side answer is rarely the one the migration vendors sell. Published analyses of recent mainframe to cloud programs report typical budget overruns well above one hundred percent, timelines slipping many months past plan, and a majority of programs failing to meet their stated objectives, with a meaningful share abandoned mid project or reverted to the mainframe within a year of cutover.
That does not mean exit is never right. For specific, separable workloads it can be. It means the decision has to be priced honestly, against the often cheaper alternative of optimizing the existing mainframe spend. And it means the most valuable thing the exit question produces, for most estates, is not a migration but a credible, modeled exit option that gives a renewal negotiation something real to compete against. The cost tables below are for making that judgment with eyes open.
The quoted migration figure is usually the direct layer only. The layers below it are where programs overrun. Ranges reflect publicly reported industry patterns and vary widely by estate; treat them as directional, not a quote.
| Cost layer | What it covers | Why it overruns |
|---|---|---|
| Direct migration | Rehost, refactor, or rearchitect the application code | Code conversion is priced per line and rewrites carry low success rates |
| Data conversion | Moving and reconciling large datasets and file structures | The most common cause of mid project abandonment |
| Parallel run | Running both platforms during cutover | Double licensing and operating cost for an extended, often extended again, window |
| Talent | Scarce COBOL and dual platform migration engineers | High day rates and a shrinking workforce inflate and lengthen the program |
| Integration and egress | Reconnecting surrounding systems, ongoing cloud data movement | Recurring costs vendors commonly leave out of the initial quote |
| Revert risk | Returning to the mainframe after performance failure | A reported share of programs revert within a year, paying twice |
Most large estates land on hybrid, keeping core workloads and moving only what is genuinely separable. Use the frame to decide which path each workload belongs in before committing the estate to anything.
| If the workload is | The honest path | The cost lever |
|---|---|---|
| Core, integrated, transaction heavy | Stay and optimize | Metric, capacity, and renewal negotiation on the existing spend |
| Separable, self contained, modest data | Candidate to migrate | A costed exit for that workload, priced against staying |
| Aging and low value | Sunset or retire | Decommission to drop the products it carries |
| The whole estate at once | Rarely the cheapest | Use the modeled option as renewal leverage, not a program |
For most estates the reliable saving is on the mainframe spend itself, not in leaving it. A modeled exit for the separable workloads doubles as the credible alternative that moves a renewal number. The whole estate exit, priced honestly with overrun and revert risk, usually loses to optimization.
Migration quotes commonly price the code path and leave out data conversion, parallel run, integration, and egress. The layers under the quote are where programs overrun by more than the headline figure.
During cutover you run and license both platforms. When the migration slips, and it commonly does, the double cost runs for the extended window too, eroding the saving the exit was meant to deliver.
A vague threat to leave that the vendor can see through is worse than none, because it signals you have not modeled it. Leverage comes from a costed, credible option, not from posture.
A program that reverts after performance failure has paid the migration cost and kept the mainframe. Pricing revert risk into the decision is what separates an honest business case from a vendor pitch.
Price the exit honestly, keep the option, optimize what stays.
Approach the exit question workload by workload, not estate wide. Cost each separable workload's migration honestly, including the layers the migration quote omits, and compare it against the cost of staying after optimization. For the core estate, the reliable lever is the existing spend: read cost per MSU benchmarks and MSU consumption optimization for where that saving lives. Keep the modeled exit for separable workloads as a costed option, because that is the credible alternative the renewal table respects, built well before you need it on the 18 month runway.
Where a genuine exit is right for specific products, the licensing side has its own traps, from reinstatement fees to transfer rules, so the wind down has to be sequenced. When the exit question is on the table, whether as a real program or as renewal leverage, our cost optimization and license negotiation teams cost both sides so the decision is made on real numbers.
Published full migration estimates range from a few million to tens of millions depending on approach, but the headline rarely holds. Industry analyses report typical overruns above one hundred percent and long timeline slips driven by hidden data and integration costs.
Common causes are data conversion problems, performance degradation forcing a revert, scarce and costly talent, and underestimated integration. Several published analyses found a majority of recent programs missed their stated objectives.
For most large estates, yes, because hybrid is the dominant reported strategy and optimizing the existing spend usually beats a wholesale exit once overrun and revert risk are priced in. Specific separable workloads can still be candidates.
A costed, credible exit option is real leverage. A bluff the vendor can see through is not. The value is in the modeled alternative for separable workloads, presented as a genuine option, not in posturing.