VMware

What VMware migration economics actually look like in 2026.

A clear eyed read on what it costs to leave VMware, what it costs to stay, and how the gap between the two has narrowed in some places and widened in others over the last twelve months. Numbers, not narrative.

By The Negotiation Desk · VMware · Archetype: The Exit

Published 29 Jan 2026 · Updated 28 Feb 2026

Migration economics are the most asked about and the least carefully modeled subject in every Broadcom renewal conversation we have had this year. The asking is understandable. The renewal quote is high, the bundle has changed, and the buyer wants to know whether the alternative is genuinely available. The careless modelling is also understandable, because the available comparisons are noisy, the inputs are contested, and the seller side framing of the question has been, predictably, pessimistic about the alternative. This piece presents the Desk's working model as of the publication date. It is the model we apply when a buyer asks us to read the exit option, and it is the model we are willing to defend in writing.

The point of the piece is not that migration is the right answer. Migration is the right answer for some buyers and the wrong answer for others, and the wrong answer for most buyers in most years. The point is that the buyer side analysis should be done early enough and rigorously enough that the renewal conversation can reference it, because the documented alternative pathway is one of the two largest concession band shifters we measure in a Broadcom renewal, the other being the entitlement correction.

The four cost categories

The migration model has four cost categories. The first is platform replacement cost. This is the licensing or subscription cost of the destination platform, modeled across the same term as the comparable VMware renewal. The second is migration execution cost. This is the labour, downtime, parallel run, and tooling cost of moving workloads. The third is operating cost differential. This is the steady state cost difference once migration is complete, including infrastructure, support, training and retention. The fourth is risk adjusted exit cost. This is the cost of remaining VMware exposure that does not migrate, including any termination payments, any residual support, and any continued audit exposure.

Each category requires its own inputs and produces its own range. The categories are not independent. The platform replacement cost depends on the destination platform's licensing model. The migration execution cost depends on the workload mix and the destination's compatibility. The operating cost differential depends on the comparison of steady states. The residual exposure depends on the migration scope. Modelling the categories without modelling their interaction produces a number that is misleading in the buyer's favour. The Desk's working model handles the interaction with a sensitivity overlay on each category.

What the platform replacement cost actually shows

Across the engagements we have modeled in the last 12 months, the platform replacement cost on a like for like basis sits in a wide range. For estates with simple infrastructure profiles, the destination platform cost runs at 35 to 60 percent of the comparable VCF renewal. For estates with heavy reliance on advanced features specific to VMware, the destination platform cost runs at 70 to 105 percent of the comparable VCF renewal, with some destinations costing more than VMware once feature parity is required.

The single largest driver of the range is the workload mix. Estates dominated by general purpose virtualisation can migrate cheaply on a license basis. Estates dependent on specific networking, storage policy, automation or container orchestration features tied to VCF migrate at much higher license cost, because the destination has to provide those features either natively or through additional products.

What migration execution actually costs

The migration execution cost is the part of the model the seller side framing has historically understated and the buyer side framing has historically overstated. The Desk's working numbers, drawn from completed and partially completed migrations in our book, sit at 1.4 to 3.2 times the destination platform's first year cost. This is the all in execution cost, including labour, downtime exposure, parallel run period, retraining, and the tooling required to support the migration.

The range varies primarily by estate size and migration approach. Lift and shift migrations sit at the lower end. Replatform migrations, where workloads are modified to match the destination's idioms, sit higher. Refactor migrations, where workloads are rebuilt for the destination's native capabilities, sit higher still. The right approach for a given workload depends on the workload's value, the team's capacity, and the timeline available.

"The migration cost the buyer should fear is not the license cost of the destination. It is the labour cost of moving, and the labour cost has not gone down. The license gap has narrowed. The labour gap has not."Exit Planning Lead, The Desk

What the operating cost differential actually shows

The third category is the one most buyers underweight. After migration completes, the destination platform produces a steady state operating cost that differs from the steady state operating cost of remaining on VMware. The differential can be favourable or unfavourable to the destination, depending on the operational maturity, support contract structure, and infrastructure profile.

Across our modeled engagements, the steady state operating cost on the destination platforms we have modeled runs at 80 to 130 percent of the steady state operating cost on VMware, with median around 95 percent. The median is below VMware's steady state, but only by a small margin. The buyers who model this category carelessly assume the destination is cheaper at steady state by a large margin. That assumption is rarely defensible.

What the residual exposure actually shows

The fourth category is the residual exposure. Few migrations move every workload. The workloads that remain on VMware after a partial migration still incur licensing, support and audit exposure. The economics of the migration model have to include the cost of the unmigrated tail. We see buyers either include it well, in which case the model produces a defensible number, or include it badly, in which case the model shows an exit saving that does not exist in practice.

The Desk's working assumption is that the residual exposure should be modeled at full retail VMware pricing rather than at the buyer's negotiated bundle pricing, because the bundle pricing depends on the bundle, and the bundle assumptions change when the buyer is no longer purchasing the bundle at the same scale. This conservatism in the model tends to make exits look less attractive on paper than they look in seller side comparisons. We are comfortable with the conservatism.

What the model tells the renewal conversation

The most useful output of the model is not the answer to "should we migrate". The most useful output is the price point at which the renewal becomes worse than the documented alternative. This price point is the buyer's effective ceiling. Below the ceiling, the renewal is the right answer on the numbers. Above the ceiling, the renewal is worse than the documented alternative, and the buyer has a credible position to walk.

The ceiling is rarely close to the seller's opening quote. The ceiling sits somewhere between 30 and 70 percent below opening quote for most estates we have modeled. The opening quote is therefore almost always negotiable down to the ceiling, because the seller knows the ceiling exists and the buyer knowing it exists changes the conversation. Buyers who model the ceiling rigorously, document it, and reference it in the renewal conversation, close at the ceiling or close. Buyers who do not, close at opening quote less whatever modest concession the seller offers without challenge.

Destination platform cost vs comparable VCF renewal, range35% to 105%

Migration execution cost, multiple of destination first year1.4x to 3.2x

Steady state operating cost on destination vs VMware, median95%

Buyer ceiling typical position below opening quote30% to 70%

The three pitfalls in buyer side modelling

The first pitfall is using the seller's discounted bundle price as the comparison baseline. The bundle price is a moving number. The right baseline is the rack rate for the components the buyer actually consumes, applied across the term, with the bundle discount itemised separately as a variable. This is more work. It is also the only way to produce a comparison that survives the seller's pushback.

The second pitfall is ignoring the timing curve. The migration cost is concentrated in the first 12 to 24 months. The migration savings, where they exist, accumulate over the full term. A net present value model that does not handle the timing curve produces a number that flatters migration in the early years and understates the residual VMware exposure in the later years. The right model puts both flows on the same discount curve and reports the present value comparison.

The third pitfall is single point estimation. The four cost categories each carry meaningful uncertainty. The right model carries a low, central and high case for each category, propagates the cases through the term, and reports the comparison as a distribution rather than a single number. The single number gives the buyer false confidence. The distribution shows where the comparison is robust and where it is sensitive. Buyers who present a distribution to the seller during renewal produce stronger concession movement than buyers who present a point estimate, because the distribution is harder for the seller to dismiss.

What we have seen on live deals

An energy major in late 2025 modeled a full migration of its VMware estate and presented the documented pathway to the seller during renewal. The model showed a ceiling 52 percent below opening quote. The seller's commercial team, on receiving the documented pathway, moved the offered price to 41 percent below opening quote inside two working sessions. The buyer did not migrate. The buyer signed at the negotiated price. The migration model paid for itself many times over without a workload moving. This is the more common outcome in our book. The model is used. The migration is not. The buyer pays for the model in the renewal cycle, and recovers the investment in the closing position.

A North American transport operator took the opposite outcome. The model showed a ceiling 38 percent below opening quote. The seller's commercial team moved 22 percent and held. The buyer began a phased migration on the non critical tier of the estate during the renewal term, structured to be reversible if the seller offered better terms at the next checkpoint. The seller has since offered better terms. The migration continues in parallel because the buyer's operations team has built capability that has value beyond the renewal cycle. Both outcomes are reasonable. Both started from the same kind of model.

The takeaway

Migration economics have four cost categories and they interact. Modelling the categories without modelling their interaction produces a misleading number. The interaction sensitivities are the work.
The license cost gap between VMware and destinations has narrowed in some places. The labour cost gap has not. Buyers who plan around license savings without planning around execution cost end up paying twice.
The most useful output of the model is the buyer's ceiling, not the answer to whether to migrate. The ceiling is referenced in the renewal conversation, and the renewal conversation is where the model usually pays back, with no workload moving.