IBM virtualization rules by hypervisor type.

Sub-capacity licensing lets you license only the virtual cores running IBM software rather than every physical core in the host. That single principle saves enormous money, but it is conditional. The condition that trips most buyers is not whether they deployed ILMT, it is whether the hypervisor underneath the workload is recognized and counted the way they assumed. IBM publishes eligibility and counting rules per virtualization technology, and the differences are large enough to swing a finding by an order of magnitude.

The headline example shows the stakes. A WebSphere instance using four of thirty two host cores under an eligible, correctly reported hypervisor is licensed at roughly 480 PVU. The same instance defaulted to full capacity is 3,840 PVU. Eight times the exposure, decided entirely by whether the virtualization layer and its reporting hold up.

Power and LPAR: allocation, not utilization

On IBM Power systems, capacity is counted by the cores allocated to a partition, not by how busy those cores are. An LPAR assigned eight cores is licensed for eight cores even if the workload idles most of the day. Capping with the right shared processor pool settings is what controls the number. Buyers who assume light utilization protects them are reading the wrong meter, because allocation is what IBM counts.

VMware: the cluster is the boundary

On VMware, the counting question is what the virtual machine can reach, not only where it sits today. Where vMotion and Distributed Resource Scheduler can move a guest across hosts, IBM expects the capacity calculation to reflect every host in that movement boundary unless cluster mobility is constrained and that constraint is evidenced in ILMT. A loosely configured cluster can pull every host into scope. This is the single most common virtualization dispute in our work, and it is covered in depth in the VMware article below.

Hyper-V and other x86 hypervisors

Microsoft Hyper-V, KVM, and similar x86 hypervisors are eligible for sub-capacity, but only on supported versions and only when ILMT, or an approved tool such as Flexera One ITAM or HCL BigFix Inventory, sees the environment correctly. Running IBM software on an unsupported or end of life host operating system can place the deployment on ineligible technology, which voids the sub-capacity claim and reverts that environment to full-capacity charging.

Cloud: eligible regions and instance types only

Public cloud is eligible for sub-capacity only on approved providers and instance types, with capacity counted against the virtual cores of the instance. Land the same workload on a non-eligible configuration and the sub-capacity position does not apply. Cloud also tends to interact with the VPC metric rather than PVU, which changes the math again.

• Power and LPAR count allocated cores regardless of utilization
• VMware counts the cluster mobility boundary unless movement is constrained and evidenced
• Hyper-V and KVM are eligible only on supported host versions seen correctly by ILMT
• Ineligible or end of life host operating systems void the sub-capacity claim
• Cloud eligibility is restricted to approved providers and instance types

The practical defense is to map every IBM workload to its hypervisor, confirm eligibility for that exact platform and version, and prove the counting boundary with continuous ILMT data. A finding that assumes full capacity collapses the moment you produce a correctly scoped, hypervisor-specific calculation backed by evidence.