Contents
Key takeawaysWhat changed in VCF licensingThe core minimumCore count against discountWhat we see in renewalsWhat Broadcom will sayContract terms to ask forWhat to do nextFAQVCF is sold per core with a 16 core minimum per processor, so small hosts are billed for cores they do not have. Correcting that count lowers every year of the subscription and the next renewal, which a discount cannot do.
- Three changes stacked. Bundled scope, subscription pricing and the per processor core minimum raised annual cost 2 to 5 times for customers who had licensed only the hypervisor.
- Small hosts pay for phantom cores. Every CPU is billed at 16 cores or more, so horizontal builds on low core count processors carry capacity that does not physically exist.
- A lower count lasts longer. A lower core count carries into the next renewal baseline, while a discount is renegotiated from zero each cycle.
- Consolidation pays first. Moving onto fewer, denser hosts and switching off idle ones cut the bill in our renewals before any discount was discussed, and the lower count holds for every year of the term.
- Sequence the steps. Fix the count, then table a costed alternative, then negotiate rate, term and price protection.
- The discount was often a distraction. In 22 of the 35 renewals we reviewed, the discount conversation displaced the billable core count as the thing that decided the invoice.
What changed in VCF licensing, and why did costs rise so sharply?
Broadcom replaced perpetual VMware licenses with a per core subscription that carries a minimum per processor. For customers who had only licensed the hypervisor, annual cost rose 2 to 5 times. The increase came from three changes that landed together, and each one needs its own answer.
- Scope. The portfolio collapsed into bundles, with VMware Cloud Foundation (VCF) as the flagship. A customer who wanted compute now pays for storage, networking and management in the same subscription.
- Ownership. Subscription replaced a one time purchase plus annual support. Capital spend became a recurring cost, and there is no license left at the end of the term.
- Metric. Every processor is billed at a core minimum, so hosts with low core count processors are billed for more cores than they have.
Broadcom announced the end of availability of perpetual licenses on December 11, 2023. Existing perpetual licenses keep working, and support continues until the current support contract runs out. The change bites at renewal, when support is no longer offered and the subscription quote arrives instead.
| Dimension | Before the acquisition | VCF today | What it means for you |
|---|---|---|---|
| License type | Perpetual plus support | Subscription only | Recurring cost, no ownership |
| Unit | Per processor or per core | Per core with a minimum | Higher billable core count |
| Scope | Pick the products you need | Bundled platform | You pay for capability you may not use |
| Renewal | A support renewal | A full subscription renewal | Larger annual exposure |
Why did customers with the narrowest VMware footprint see the largest increases?
The steepest increases landed on customers who had licensed the fewest VMware products. That is the opposite of how a volume based price rise usually spreads, where the largest buyers absorb the most.
A customer that ran vSphere alone had the most bundle scope to absorb and the smallest support bill to compare it with. If that customer also ran many small hosts, the core minimum added a third layer on top.
The VMware Estate After the Repackaging
How does the VCF core minimum bill you for cores you do not have?
VCF is licensed per physical core, with a minimum of 16 cores for every CPU in each ESXi host. A processor with fewer than 16 cores is billed as if it had 16. Broadcom's own counting guide gives the example of a host with two 8 core CPUs, which is counted as 32 cores.
In the renewals we benchmarked, the minimum added 10 to 30 percent of phantom cores to the billable count, on hosts whose processors do not physically carry that many cores. The penalty is structural. It follows from the hardware you already own, so there is no error in the quote to point at.
Why are horizontal builds on small hosts hit hardest?
Under a per processor perpetual model, building out horizontally on many modest hosts was sound engineering. It spread failure domains, kept the blast radius of any one host small, and cost nothing extra in licensing because the unit was the socket.
The core minimum reverses that. The same layout now bills for cores the processors do not have, so an architecture chosen for resilience is charged as if it had been chosen for capacity. The customers paying most are the ones who followed the old advice most closely.
Under the core minimum, your data center layout became a licensing decision after the hardware was already bought.
No one made a bad decision here. The metric changed underneath a decision that was already made, so the lasting remedy runs on your hardware refresh cycle rather than through negotiation. The VVF against VCF guide covers the bundle side of the same question.
How do you count your own billable cores?
Build a host level count yourself before Broadcom sends a quote, then reconcile its number against yours line by line. These sources give you what you need:
- RVTools, vHost tab. Lists CPU sockets and cores per socket for every host in vCenter. Add a column that bills each socket at the greater of its real cores or 16.
- Broadcom's PowerCLI counting scripts. KB 313548 publishes FoundationCoreAndTiBUsage.psm1, plus a multi vCenter version, so customers can run the same count Broadcom uses. Run them yourself and keep the output with a date.
- Disabled cores. Broadcom warns the scripts can report inaccurate results where cores are disabled in the BIOS. Do not rely on disabled cores to lower the count unless Broadcom agrees in writing.
- Idle and retiring hosts. Mark every host that runs nothing, or that leaves service during the term. A host with no workloads still bills at the floor.
Our explainer on VMware core licensing covers the counting rules in more detail.
Broadcom VMware negotiation brief
The per core arithmetic, the core minimum rule and the renewal terms that hold against Broadcom's opening quote.
Get the white paper →Why does cutting billable cores beat negotiating a discount?
A core reduction changes the quantity for good, while a discount changes the rate for one term. Cutting billable cores by 20 percent beats a 20 percent discount because the lower count carries into the next renewal as the starting number, and the discount has to be won again from zero.
A discount also decays as list prices move underneath it. Even if you win the same percentage at the next renewal, it comes off a higher list, while the core reduction keeps working in every year of the subscription.
A worked example with 40 small hosts
Say you run 40 hosts, each with two 12 core processors. That is 960 physical cores, but the minimum bills each processor at 16, so the quote counts 1,280. Eight of those hosts run nothing. The table uses an illustrative rate of $300 per core per year, not a Broadcom list price.
| Path | Billed cores | Rate per core | Annual cost |
|---|---|---|---|
| Quote as counted | 1,280 | $300 | $384,000 |
| 20 percent discount, count unchanged | 1,280 | $240 | $307,200 |
| 8 idle hosts removed, a 20 percent cut in cores, no discount | 1,024 | $300 | $307,200 |
| Cores cut and discount won | 1,024 | $240 | $245,760 |
| Remaining 32 hosts consolidated onto 16 hosts with two 24 core processors, plus the discount | 768 | $240 | $184,320 |
The second and third rows produce the same first invoice. They separate at renewal. Suppose list rises 10 percent to $330. The discounted path starts the next negotiation at 1,280 cores, or $422,400 at list. The reduced path starts at 1,024 cores, or $337,920. That $84,480 gap exists before anyone discusses price.
The last row also shows why consolidation matters beyond the idle hosts. The same 768 physical cores on denser processors carry no phantom cores at all, so every point of discount lands on capacity you actually run.
Why we disagree with "accept the bundle and negotiate the discount"
The usual reseller advice is to accept the bundle and negotiate a discount off the new list. In 22 of the 35 renewals we reviewed, that advice pointed the customer at the smaller of the two available savings.
A discount on an inflated count partly pays for cores that do not exist, and it resets at the next renewal. Correct the count first, then ask for the discount.
Why does the order of the steps change the total?
A standard subscription order commits the core count for the whole term. If you sign for 1,280 cores and consolidate in year two, you keep paying for 1,280 until renewal unless the contract gives you a reduction right. So the consolidation has to be finished, or written into the order as a ramp, before you sign.
The same applies to an exit quote. In our migration work, a costed alternative quote cut the renewal by an average of 22 percent, whether or not the customer went on to migrate. That percentage works on whatever base is on the table, so produce the quote against the corrected count.
- Exit costs. The migration economics analysis sets out what a move to Nutanix costs in licenses and effort.
- Other platforms. The alternatives comparison covers the main candidates beyond Nutanix.
What have we seen in Broadcom VMware renewals in 2024 and 2025?
Across roughly 30 to 40 Broadcom VMware renewals we benchmarked between 2024 and 2025, the shift to subscription raised annual cost for most customers. The sharpest jumps were where the customer had licensed only the hypervisor. Three patterns came up again and again.
- Scope drove the biggest multiples. The customers who moved from a support bill on vSphere licenses they owned to a subscription for the full bundle saw the largest multiples, as described at the top of this page.
- Phantom cores followed the hardware. Wherever hosts ran processors below 16 cores, the minimum inflated the count as described above, and no discount conversation touched that part of the bill.
- Consolidation paid before price did. Moving onto denser hosts recovered 15 to 25 percent of the increase before any discount was discussed, and that reduction held for every year of the subscription.
The renewals that went best redesigned the host layout first, so the minimum stopped charging for cores that do not exist. Then they priced a credible alternative against the corrected count, and only then negotiated rate, term and price protection.
How does this differ for a small and a large environment?
The sequence is the same at any size. What changes is where the phantom cores sit and how much of the bill they represent.
- A few dozen hosts. Idle and undersized hosts are often a large share of the count. Consolidation and decommissioning usually outweigh the last few points of discount, and one hardware refresh can remove most phantom cores.
- Thousands of hosts. Most processors may already carry 16 cores or more, so phantom cores are a smaller share, but they still add up. At 2,000 two socket hosts, 3 percent of a 64,000 core count is 1,920 cores, about $576,000 a year at the illustrative $300 rate. Per cluster scope and a renewal cap deserve as much attention at that size.
Current per core ranges are in our VCF pricing guide.
What will the Broadcom account team say, and how should you reply?
Expect the conversation to steer toward rate and deadline and away from the count. These are typical lines on a VCF renewal, with replies that keep the count on the table.
| What you will hear | What to say back |
|---|---|
| "The 16 core minimum applies to every customer." | We accept the rule. We are disputing which hosts are in the count. Here is our host list, with idle and retiring hosts marked. |
| "This discount is only valid until quarter end." | Our signing date follows the consolidation plan. We will sign once the count reflects the hosts we will run. |
| "Full VCF gives you the best value per core." | Quote VCF and vSphere Foundation per cluster. We will decide on the clusters that use NSX and vSAN. |
| "Migrating would cost more than renewing." | Here is our costed alternative. We are happy to compare it line by line with your offer. |
Quarter end pressure is real on the Broadcom side, and it can work for you once your count is ready. Our note on fiscal quarter end timing covers when that pressure peaks.
What contract terms should you ask for in a VCF subscription?
Ask for terms that protect the corrected count for the life of the subscription. Broadcom does not grant these by default, so each one is a request you make in writing before signature.
- A host schedule attached to the order. The count is tied to named hosts and processors, so a later count dispute starts from an agreed list.
- A reduction right at hardware refresh. When you consolidate or decommission during the term, the billable count falls at the next anniversary.
- A price hold on added cores. Growth during the term is priced at the same per core rate, not at the list price of the day.
- A renewal cap. A ceiling on the per core rate at the next renewal limits how far the rate can reset. The price cap guide covers the wording.
- A ramp. If consolidation cannot finish before signature, a ramp steps the count down on a documented schedule.
When should each step happen before the renewal?
Consolidation takes hardware lead time, so the work starts long before the quote. Use this timeline as the default and bring it forward if a hardware refresh is already planned.
| Months before renewal | What to do |
|---|---|
| 12 | Build the host level count and flag phantom cores, idle hosts and clusters that need only compute |
| 6 | Finish consolidation or agree a ramp, and price a costed alternative against the corrected count |
| 3 | Reconcile Broadcom's count with yours and table the alternative |
| 1 | Negotiate rate, term and price protection, and confirm the host schedule in the order |
What to do next
- Count physical cores per host. Find every processor where the minimum bills for cores that are not there, and total the phantom cores.
- Consolidate and decommission. Move workloads onto fewer, denser hosts and switch off idle ones before the quote is built, so the lower count holds for every subscription year.
- Challenge the bundle scope on its own. Price vSphere Foundation for clusters that do not use NSX or vSAN, and keep that argument apart from the count.
- Cost an exit against the corrected base. The alternative quote works on whatever number is on the table, so produce it after the count is fixed.
- Negotiate rate, term and price protection last. Ask for the host schedule, reduction rights and a renewal cap in the same order.
- Get a second check on the count. The Broadcom practice runs the core audit with you and reviews the quote against current deals.
Holding a Broadcom VMware quote? Our VMware renewal negotiation team works only for buyers, for a fixed fee or 25 percent of what we save you.
Frequently asked questions
What changed with VMware Cloud Foundation?
Broadcom folded the VMware catalog into a small number of bundles, with VCF at the top, sold per core on subscription. New perpetual licenses are no longer sold. A customer who needed only vSphere now buys a platform that also includes vSAN, NSX and the management tools, which accounts for most of the increase.
What is the minimum core count for VCF licensing?
Sixteen physical cores per CPU. A two socket host is never billed below 32 cores, whatever its processors carry. Only physical cores count, so hyperthreads add nothing, and a processor with more than 16 cores is billed at its real count. Modern high core parts are unaffected by the minimum.
How do core minimums catch buyers out?
They inflate the count on hardware you bought for other reasons. In our file the minimum added 10 to 30 percent of phantom cores. Low core processors are typical in edge sites, management clusters and older hosts, so look there first. Because the quote counts hosts rather than workloads, a cluster built for failover capacity is billed in full.
Why did costs rise so much for hypervisor only customers?
All three changes hit them at once, on the smallest previous bill. Their old cost was support on licenses they already owned. When you report the increase internally, split it by cause: added bundle scope, the switch from support to subscription, and phantom cores. Each has a different fix, and only the last one yields to consolidation.
Is negotiating a discount the right first step?
No. Fix the count first. A discount won on an inflated count lasts one term and spends part of its value on phantom cores. Once the count is right, push for the discount, a renewal cap and a price hold on added cores together, so the rate cannot simply reset later.
How much does consolidation recover?
In our renewals, 15 to 25 percent of the increase, before price was discussed. The saving depends on how many hosts run processors below 16 cores and how many sit idle. Removing idle hosts carries no engineering risk and is the fastest part to deliver.
When is an exit worth modeling?
Once the core count is corrected, and early enough that the numbers are credible. A costed alternative cut renewals by an average of 22 percent in our migration work even when the customer stayed. Scope it to a real workload group, such as one cluster or site, with migration effort included.
Is hardware layout really a licensing decision now?
Yes, and it became one after the hardware was bought. Processor choice and host count now set the subscription bill, so infrastructure and procurement teams should review refresh plans together. A refresh that looks marginal on performance can pay for itself through the cores it removes.