MuleSoft vCore pricing: an Abrams tank for a Corolla

MuleSoft, Salesforce's integration platform, prices itself by the vCore — a reserved slice of compute you pay for in full whether it runs at 5% or 95% load. Because hitting your vCore ceiling blocks new deployments, teams buy headroom they never use, then stack premium connectors, a dedicated load balancer, and near-20% renewal hikes on top. It is peak-sized, reserved capacity billed like a column of tanks for jobs a hatchback would finish.

What are you buying when you buy a vCore?

A vCore is MuleSoft's unit of runtime capacity — a fixed allocation of vCPU and memory that hosts your integration applications. You reserve it, and it is yours until renewal whether or not anything flows through it. Third-party pricing guides put list price at roughly $50,000 to $70,000 per vCore per year, discounting to about $8,000 to $14,000 in practice as of December 2025. That is per vCore, and real estates rarely run on one. One documented arrangement came to $210,000 a year for four vCores spread across three environments, before $140,000 of partner implementation to stand it up. The environment split is the quiet multiplier: development, test and production each want their own reserved slice, so you provision the same logical workload more than once.

Why does the sizing always round up?

Because the meter never stops and the ceiling is hard. A reserved vCore is charged at full rate regardless of how much traffic crosses it, and running out of capacity is not a soft limit. Reviewers describe the mechanic plainly:

Reserved vCores incur full charges even during low-traffic periods, and when an organization reaches its vCore limit, new deployments are blocked.

Read those two clauses together and the incentive is unavoidable. If under-sizing means your next release fails to deploy, you size for the worst hour of the worst day and carry that headroom every other hour of the year. There is no cheap burst, so you pre-buy the peak. A Gartner reviewer caught the feeling exactly:

like paying for an Abrams tank when all we really need is a Toyota Corolla.

This is overprovisioning by design, not by accident. The unit is coarse, the billing is reserved, and the penalty for guessing low is a blocked deployment — three properties that all push the same way: buy more than you need, and keep buying it. In our own sizing exercises, the gap between provisioned and used capacity on a reserved-vCore deployment is rarely under a third, and on bursty order-capture traffic it is often far worse, because the reserved figure is set by a peak that arrives only a handful of times a year and idles the rest of it.

What isn't in the vCore number?

The vCore line is where the bill starts, not where it ends. Several things a working integration estate actually needs are metered separately, as of December 2025:

Premium connectors are, predictably, the ones you actually want — the SAP, the mainframe, the higher-value SaaS endpoints — at around $10,000 to $15,000 each per year. A dedicated load balancer, needed for a custom domain or your own SSL handling, adds $36,000 to $60,000. And the total resets upward every year:

close to 20% price increase year on year

— typically surfaced about a month before renewal, when your switching costs are at their highest and your leverage at its lowest. We have written before about paying $48,000 a year to look at your own data; the shape is the same here — a recurring licence indexed to something other than the value you pull out of it.

Who can actually operate it?

Even a perfectly sized MuleSoft estate needs people who can drive DataWeave, its transformation language, and the reviews are consistent that the floor is high. One review puts it simply:

The learning curve is particularly steep.

Another is blunter about who gets shut out:

This isn't a platform where business analysts can self-serve.

That matters to the economics because the licence is only half the story. Standing the platform up is itself a six-figure exercise: the four-vCore arrangement priced at $210,000 also carried $140,000 of partner implementation, because it takes specialists to model your flows. It is the same staffing tax we flagged in a platform that pretty much needs an IT department and in our loyalty platform postmortem: the sticker price of the software is a fraction of the fully loaded cost of running it.

The fix direction is not a cheaper vCore; it is a unit of scale that matches the shape of the work. Integration traffic in a route-to-market estate is bursty and uneven — order capture spikes at depot cut-off, returnables reconciliation runs overnight, loyalty events trickle in all day. Reserving peak-sized blocks to cover that profile is the core mismatch, and an event-driven bus that scales with events rather than reserved cores removes the incentive to pre-buy the peak at all.

None of this makes MuleSoft a bad runtime; it makes it an expensive default. The question for 2026 capacity planning is not whether you can grind the per-vCore rate down at renewal — it is whether your integration layer should be priced by reserved capacity in the first place. As more of the route-to-market stack moves to event streams, the estates that come out ahead will be the ones that stopped sizing for the worst hour and started paying for the work that actually happened.