A drive costs a fraction inside a server of what the same NAND costs inside an array. Storage licensing reverses that math, and the array comes back.
Hyperconverged infrastructure made a hardware argument and won it. Put the data on drives inside standard servers, buy those drives through the channel, and skip the array vendor’s price list. That argument still holds today. Walk a data center anyway and the array is often back, sitting beside the HCI cluster and holding the data that grew. Storage licensing explains the reversal, and the teams who bought that array were reading their spreadsheets correctly.
Key Takeaways
- A server drive and an array drive often hold the same NAND. The server drive costs a fraction of the array drive, so HCI wins the hardware comparison outright.
- Storage licensing on several converged platforms runs two meters. One counts cores, with a floor of 16 cores per CPU. The other counts raw terabytes.
- Each licensed core includes raw capacity. VCF includes 1 TiB per core and VVF includes 0.25 TiB per core. Anything past that bills as an add-on license.
- The meter reads raw physical capacity, so deduplication and thin provisioning never lower the licensed number.
- Capacity-priced arrays win the arithmetic that follows, which sends the data back onto the more expensive hardware.
The Hardware Math Favors HCI
Price the same terabyte two ways. Bought as a server drive it moves through the channel, quoted by several suppliers against each other and installed in a chassis you already own. Bought inside a dedicated array it arrives at the array vendor’s price, qualified against that vendor’s own compatibility list, and carries a controller and a support contract around it. The NAND underneath is often identical, and the server drive still costs a fraction of the array drive.
That gap matters more this year than last. DRAM contract prices rose 90 to 95 percent quarter over quarter in the first quarter of 2026, with another 58 to 63 percent following in the second, and NAND climbed 70 to 75 percent alongside it. Every terabyte now carries a rising acquisition cost, so the freedom to shop a drive across suppliers turned into real money.
What Storage Licensing Meters
Broadcom runs two meters rather than one. The first counts cores, and the second counts raw terabytes.
Key Terms
Raw capacity
The physical capacity of the disks vSAN claims, counted before failure tolerance and before data reduction. This is the number licensing measures.
TiB entitlement
The raw capacity included with each licensed core. VCF includes 1 TiB per core. VVF includes 0.25 TiB per core, rounded up.
Add-on capacity license
A separate license sold in TiB increments, required once raw capacity in the cluster passes the entitlement the licensed cores earned.
Entitlement recalculation
The re-measure of entitlement against consumption. It fires on a license assignment, a host joining, a host leaving, and any change in total raw TiB.
Controller VM
The virtual machine Legacy HCI runs on every node to deliver storage services. It reserves cores and memory on each host before a workload starts.
- Cores. vSAN licenses every core on every host in the vSAN cluster, with a floor of 16 cores per CPU.
- Capacity. Each licensed core carries an entitlement of raw capacity. VCF includes 1 TiB per core. VVF includes 0.25 TiB per core.
- Overage. Raw capacity past the entitlement bills through an add-on license sold in TiB increments.
Broadcom’s own example puts three hosts with two 16-core CPUs each at 96 licensed cores. That earns 96 TiB. Fill those hosts with 100 TiB of raw NVMe and you buy an add-on license for the 4 TiB overage.
The Meter Reads Raw, Not Usable
This detail costs more than the rate does. Storage licensing here measures the raw capacity of the disks vSAN claims, ignoring what you get back after failure tolerance and data reduction. Dedupe and thin provisioning cut what you store. The licensed number stays where the drives put it.
Modern flash passes the ratio on the first tray. Look at one dual-socket host with 32 cores.
The VCF entitlement covers under a third of that tray. Under VVF, one drive of that size passes the whole host’s included capacity.
Density lowers your cost per terabyte and raises your license at the same time.
Both Meters Read the Same Servers
The two meters run on the same hardware at once, which closes the obvious escapes. Filling the bays with denser drives trips the capacity meter. Consolidating onto fewer, higher-core-count servers holds the core count roughly flat and trips the other one. The capacity those cores earn was paid for through the core count all along, so the included terabytes arrive as a ration rather than a gift.
Shrink the Cluster and the Bill Grows
Broadcom recalculates entitlement against consumption at four moments. The list covers a new license assignment, a host joining the cluster, a host leaving it, and any change in total raw TiB.
That third one should worry anyone consolidating. Retire a host and its cores leave with it, taking the entitlement they carried. Your data stayed exactly where it was, and the cluster now sits over its included capacity. An add-on license covers the gap, so shrinking the footprint writes you a storage bill.
The Controller VM Charges Its Own Toll
Nutanix reaches the same place by a different road. Its storage runs inside a controller VM on every node, documented at roughly 16 vCPU and 64 GB per host. A twelve-node cluster gives up about 192 vCPU and 768 GB before an app starts. That reservation holds whether the cluster serves 50 TB or 500. Both vendors certify their own hardware, so a compatibility list decides which servers you can buy.
The Array Comes Back to More Expensive Hardware
External arrays price by usable capacity, which matches how storage teams think about data. Put a capacity-priced array next to a core-and-raw-terabyte meter and the arithmetic picks the array. Read that result closely. The team just moved its capacity onto the more expensive hardware to escape the software bill, and the move penciled out.
Storage licensing beat the physics.
The tier the converged purchase retired comes back with its own controller, its own support contract, and its own refresh calendar. The network repeats the pattern. When the software-defined network costs too much or asks too much of your team, the switches and the network vendor stay. The three-tier data center rebuilds itself on the far side of a converged purchase, an outcome familiar to anyone who has weighed turnkey against software-based HCI.
Three Tiers Means Three Masters
Three tiers is a form of lost control rather than an older diagram. Split compute, storage, and network and you get three vendors, three support contracts, three refresh calendars, three compatibility lists, and three licensing models, each one working to grow its share of your budget. The bill comes due during an outage, when three vendors read one another’s logs and you own an integration nobody else supports.
Two Questions That Separate the Platforms
Ask any converged vendor to price two scenarios. First, double the raw capacity and add no compute. Second, pull two hosts out and hold the data flat. A vendor who charges more in either case has tied your storage bill to something other than your data.
Then ask for the ratio in writing. How many raw terabytes does each licensed core carry? What happens at the line? Does the meter read raw or usable? A raw meter means every efficiency feature you paid for stops short of the invoice.
Scaling on the Workload’s Terms
Storage and compute scale on their own when one operating system runs both. VergeOS works that way. Storage services live in the OS instead of a controller VM on every node, so the cores and memory Legacy HCI reserves go back to your workloads. Capacity grows by adding drives or storage-focused nodes on commodity x86, priced at what the channel charges for the drive. Inline global dedupe and thin provisioning cut the flash you buy in the first place. Hardware choice still matters, and the choice returns to you.
That restores an ordinary expectation. Data growth costs money in proportion to the data. Compute growth costs money in proportion to the compute. You make both calls, the hardware advantage holds, and the array stays retired.
Go Deeper
Watch the demos, then read the migration
VergeIO ran a session on how Legacy HCI took control of the data center, and it was mostly live demo. Three of them, on screen. A running VMware virtual machine imported into VergeOS in seconds, with the original retained so a bad import restores and tries again. A live VM moved between nodes from the command line with no interruption, which is the refresh clock handed back. And a batch of new VMs built from a plain-language request, using AI models running locally on the box rather than a cloud service.
Watch the session: Legacy HCI Took Control of Your Data Center (registration includes the on-demand recording)
Case study: Formica standardizes its global infrastructure on VergeOS, replacing VMware and StarWind
Storage Licensing FAQ
Is storage less expensive in HCI than in a dedicated array?
On hardware, yes. A server drive is sourced through the channel and quoted across several suppliers, and the same NAND sold inside a dedicated array carries the array vendor’s price, qualification, controller, and support contract. Storage licensing is what reverses the comparison, since a license metered against raw capacity adds a software charge that array pricing handles per usable terabyte.
How does vSAN licensing work?
vSAN licenses per core across every host in the vSAN cluster, with a 16-core-per-CPU floor. Each licensed core also carries an entitlement of raw capacity. VCF includes 1 TiB per core and VVF includes 0.25 TiB per core. Raw capacity above that entitlement needs an add-on license.
Does storage licensing count usable or raw capacity?
vSAN counts raw. The meter measures the physical disks the cluster claims, before failure tolerance and before data reduction. Deduplication and compression lower what you store and leave your licensed capacity untouched.
Can removing a host raise my vSAN license cost?
Yes. Entitlement is recalculated when a host leaves the cluster. Those cores took their included terabytes with them. The same data on the same drives can now exceed your entitlement and trigger an add-on license.
Why does dense NVMe cost more under storage licensing metered per core?
The ratio is fixed per core and modern drives are large. A 32-core host earns 32 TiB under VCF and 8 TiB under VVF. Eight 15.36 TB drives put roughly 112 TiB of raw capacity in that host, so the overage bills separately.
Do hosts outside the vSAN cluster need a vSAN license?
No. The scope is the vSAN cluster. Hosts in other clusters carry no vSAN licensing requirement, though every core on every host inside the vSAN cluster does.

Leave a comment