Volume is worth 5 to 12 points on a VCF deal, term is worth 18 to 38, so buying more cores stops buying more discount above roughly 10,000

Broadcom's discount authority steps at approximately 2,000, 10,000 and 50,000 cores, but the scale layer contributes at most 12 points against a $350 to $400 per-core list while term commitment contributes up to 38 and a credible alternative another 8 to 15. That ranking decides your next move: if you are sitting at 5,000 cores in the dead zone between thresholds, adding cores to chase a tier is the most expensive discount you will ever buy.

Prepared by Redress Compliance · August 31, 2026 · Broadcom VMware advisory. 35 to 50 renewals and migrations advised 2024 to 2026, cross-checked against a 380-plus deal benchmark database.

Broadcom does not publish a core-count discount grid, and that is deliberate: the absence of a published table means every buyer negotiates against a rep's claim about where the next threshold sits.

Reconstructed from advisory engagement files, discount authority steps at roughly 2,000, 10,000 and 50,000 cores, applied against a list band of $350 to $400 per core per year on a one-year term. Read the headline tiers against those steps and the geometry gets uncomfortable.

At 1,000 cores you sit below the first step entirely and have no scale argument at all. At 5,000 cores you are in the dead zone: past the 2,000-core step, 5,000 cores short of the next one, with nothing left to earn from volume.

At 20,000 cores you have banked the second step and face 30,000 cores of empty runway before the third. The scale layer contributes 5 to 12 points in total. Term contributes 18 to 28 points at three years and 28 to 38 at five. A credible alternative adds another 8 to 15.

If you are ranking your levers by yield, volume is third and distant, which is why benchmarking by spend tier rather than by core count gives you a more honest target.

Ask for the minimum-order clause by name and demand the paper. If Broadcom cannot produce a current policy document, the number does not exist for your deal.

The 34% ceiling reported in one 380-plus negotiation benchmark database and the 55% ceiling reported elsewhere are not contradictory: they are the same market split by whether the buyer brought term and an alternative or arrived with core count alone.

Every tier conversation is conducted in billable cores, not physical cores, and Broadcom knows the difference is in its favor. The floor is 16 cores per CPU socket. A two-socket host with 8-core CPUs licenses as 32 cores, double its physical count.

Across engagement files, the minimum added 10 to 25 percent to billable cores in estates that had optimized for many small hosts under the old per-CPU model. Those are the estates that did the right thing under the previous rules and got taxed for it.

Here is where it turns into a negotiation trap. An estate with 1,700 physical cores spread across small hosts can bill at 2,000 or more, which floats it over the first authority step. The rep will present that as good news: you qualified for the scale tier.

Run the arithmetic before you accept the framing. You just paid full negotiated rate on 300 cores you do not run, in exchange for a slice of a scale layer that is worth 5 to 12 points across the whole estate. At a $250 realized per-core, 300 phantom cores cost $75,000 a year.

The scale points you bought with them were available for free by moving from a three-year to a five-year term, which is worth 10 points more on its own. Crossing a threshold on phantom cores is a net loss every time.

So the sequence matters. Rationalize hosts first, establish billable cores second, and only then let anyone talk about tiers. A core count you have independently verified before the quote arrives is the difference between negotiating your estate and negotiating Broadcom's model of your estate.

Expect the rep to resist consolidation modeling, because fewer, denser hosts shrink the billable base and the commitment that rides on it.

Read the two discount layers side by side and the strategy becomes obvious. Term commitment moves the number 18 to 28 points at three years and 28 to 38 points at five. Scale, on estates already above 10,000 cores, moves it 5 to 12.

A vendor that wanted volume would have built the opposite ratio, because a genuine volume curve is a share-gaining instrument: you pay buyers to consolidate their estate on your platform and you accept a thinner margin per unit in exchange for footprint. Broadcom is not buying footprint.

It already has it, and it decided in 2024 that the footprint it wanted to keep was the top decile of accounts. So what got built was a commitment curve with a volume garnish.

The garnish exists because sales needs something to say to a 20,000-core account that a 2,000-core account cannot get, and because unpublished thresholds at roughly 2,000, 10,000 and 50,000 cores give a rep a story to tell about why this quarter's number is the best available.

The economics sit in the term, the bundle, and the escalator, not in the core count.

The clearest proof is in the spread of realized outcomes. One benchmark database of 380-plus post-Broadcom negotiations records an average VCF discount of 22 percent off list, with the best outcomes reaching 34.

Advisory files from the same market describe negotiated clearing between 30 and 55 percent once a core audit and a credible alternative are in play. Those two ranges are not describing different vendors or different account sizes. They are describing prepared and unprepared buyers.

If size drove the outcome, the 22 percent average would be a function of the mix of small accounts in that database and the 55 percent ceiling would belong exclusively to the largest estates. It does not work that way.

The observable variance, roughly 20 to 30 points, dwarfs the 5 to 12 points that volume tiering is capable of contributing at its maximum. Preparation is the bigger variable, and preparation is available to a 3,000-core buyer on identical terms to a 30,000-core one.

That is unambiguously good news if you are mid-sized. The advantage the very large account used to hold, that its raw purchasing scale would be converted into structurally better pricing, has largely been engineered out.

What remains available to everybody is the term lever, the exit lever, and the accuracy of your own core count.

A 4,000-core buyer who has audited billable cores against the 16-core-per-CPU floor, priced a partial migration, and is prepared to sign five years can realistically clear better per-core economics than a 25,000-core buyer who walks in with a renewal spreadsheet and a deadline.

Our engagement experience is that this inversion happens often enough to be a pattern rather than an anomaly, and Broadcom's reps know it, which is why the tier conversation gets offered so readily. It is the cheapest concession in the room.

For the largest estates the same flattening reads as a problem. Above 10,000 cores you have already collected everything the scale layer has to give.

The next published step sits at approximately 50,000 cores, which for all but a few dozen accounts globally is a theoretical construct: it exists in the discount authority matrix, it is not a threshold your estate will ever cross.

And any rep who dangles it is selling you a destination you cannot reach.

So the 20,000-core buyer has exhausted volume, has probably already signed multi-year on a prior deal, and is left with exactly one lever that still moves price: credible partial exit.

Not a threat to leave, which nobody believes from an estate that size, but a costed, dated plan to move a defined workload tier off VCF, with the target platform named and the migration budget approved.

That is the mechanism described in the costed exit that moved the quote further than the discount did, and at the top of the market it is not one lever among several, it is the lever.

The structural reason all of this holds is that Broadcom's post-acquisition model monetizes the installed base through term lock and bundle escalation rather than through unit-price competition.

Moving a customer from VVF to VCF, or adding an add-on SKU inside the same subscription, generates more incremental revenue than any volume discount concedes.

Every point of scale discount given at 10,000 cores is recovered several times over across a five-year term through package uplift and renewal escalators.

That is why the curve is allowed to flatten: the flat part costs Broadcom almost nothing to concede, and the money it protects is somewhere else entirely.

Push on tier and you will get four responses in a predictable order. First, the unpublished-thresholds deflection: the rep confirms tiers exist, declines to name them, and asks what number you need, which converts a structural question into a haggle.

Counter by requiring the offer be expressed as an approved percentage off a named price list, with the list version and effective date stated in writing, rather than a blended per-core figure that hides which SKUs carry which discount.

Second, the growth path: add cores, or upgrade a VVF population to VCF, and you reach the next tier. Refuse any tier-driven addition unless the incremental cores price below your existing blended rate, and model the five-year cost of the added cores against the discount points they buy.

In every file we have seen at 5,000 cores, they do not. Third, the bundled uplift, where the scale points arrive alongside add-on SKUs or a package move that recovers the concession and more. Price the bundle separately or refuse it.

Fourth, the quarter-end tier promise conditional on signature inside ten days, which is a timing tactic, not an authority event. Ask for the approval to be documented and dated, and let it expire once. If it reappears next quarter, it was never scarce.

Throughout, keep the discount conversation anchored to the specific price list your quote is built from, because a percentage off an unnamed list is not a concession.

The numbers in this article come from three overlapping bodies of evidence, and they do not agree with each other in the places that matter most.

The first is our own advisory file: 35 to 50 Broadcom renewals and migrations handled between 2024 and 2026, from 400-core departmental estates to 60,000-core global platforms.

The second is a published benchmark database of 380-plus post-acquisition negotiations, which records an average VCF discount from list of 22 percent and a best observed outcome of 34 percent.

The third is the set of list anchors circulating in the channel: $350 to $400 per core per year for VCF and $135 to $155 for VVF, with vSphere Standard near $55. Those three sources describe the same market from different distances, which is why the ranges are wide rather than false.

Two conflicts are live and you should resolve both before you quote a number in a meeting. The first is threshold placement: one body of evidence puts discount authority steps at roughly 2,000, 10,000 and 50,000 cores, while our own file shows a distinguishable scale layer only above 10,000.

The practical reading is that the 2,000-core step is real but small, and the 10,000-core step is the one worth planning around. The second is the 72-core minimum order announced in 2025: one source records it as still applying since April 2025, another records it as withdrawn after customer protest.

It only bites below about 1,000 cores, so ask your reseller in writing which version applies to your quote rather than arguing it in the abstract.

The recurring pattern across all three evidence sets is the same: the spread between a 22 percent outcome and a 47 percent outcome is explained almost entirely by term, costed alternative and core hygiene, not by how many cores you bought.

Use the per-core benchmark bands by spend tier to sanity check where your quote sits.

Related advisory: the full topic guide.

Does Broadcom publish a VCF volume discount table by core count?

No. Broadcom does not publish a universal price list or a core-count discount grid, and account teams will not confirm thresholds in writing. Every tier table in circulation, including ours, is reconstructed from advisory engagement files, and the reconstructions disagree on where the steps sit.

Treat any tier claim as a negotiating hypothesis to test, not a policy you can cite back to the vendor.

What discount should 1,000 VCF cores realistically get?

At 1,000 cores you sit below the first reconstructed authority step at roughly 2,000 cores, so expect essentially zero scale contribution.

Your entire discount comes from term and competitive pressure: a three-year commit is worth 18 to 28 points and a five-year commit 28 to 38, with a credible alternative adding 8 to 15. A realized figure in the $250 to $300 per-core range against a $350 list is a defensible outcome at this size.

Is it worth adding cores to reach the next discount tier?

Almost never. Scale is worth 5 to 12 points and only above 10,000 cores, so buying incremental cores to cross a threshold means paying full freight on capacity you do not need to unlock a single-digit improvement on the rest.

Run the arithmetic: if the added cores cost more than the points they unlock save across the whole estate, and they usually do, the answer is no. Buy the term band instead.

Why do two customers at 20,000 cores pay different per-core prices?

Because realized enterprise-scale VCF spans $185 to $275 per core, a 26-point spread inside one volume band. The difference is term length, whether a costed alternative exists, and whether the billable core base was audited before the quote.

Advisor-assisted deals clear at 30 to 55 percent off list while the general population averages 22 percent, and that 8 to 21 point gap tracks preparation, not size.

How does the 16-core-per-CPU minimum change my tier position?

It inflates the base your tier is calculated on. A two-socket host with eight-core CPUs licenses as 32 cores, double its physical count, and across estates built on many small hosts the minimum added 10 to 25 percent to billable cores.

That can float you over a threshold on cores you do not use, which is the worst way to buy a discount: you pay for phantom capacity to earn points you could have bought with a term extension.

Is the 72-core minimum order still in force?

Sources conflict. One holds that a 72-core minimum order has applied since April 2025, another that it was withdrawn after customer protests as of mid-2026.

Get it resolved in writing before you model a small deployment, because it only materially affects buyers at or below the 1,000-core tier and remote sites. Above a few thousand cores it is noise.

What is the fastest way to move a VCF quote if my core count is fixed?

Cost a partial exit. Competitive pressure is worth 8 to 15 points, roughly the same as or more than the entire scale layer, and sub-$200 per core generally requires both a multi-year commitment and a credible alternative.

Price migrating 20 to 30 percent of workloads to a named alternative with real dates and real numbers, then bring that to the table alongside a five-year term option.