PVU in public cloud: AWS, Azure and the BYOL question.
Moving IBM software to AWS or Azure does not leave the PVU metric behind. Bring your own license and the entitlement, the counting rules, and the tooling obligations all travel with you. Where buyers assume the cloud simplifies licensing, it usually just relocates the audit risk.
Most IBM middleware in public cloud runs under a bring your own license arrangement, often shortened to BYOL. You own the IBM entitlement and you deploy it on cloud compute you rent. The cloud provider supplies the infrastructure, but the IBM licensing obligation is entirely yours, and it is still measured in PVU. That is the point buyers miss: the cloud does not relicense the software, it just changes the hardware underneath it.
How cloud cores count
PVU in cloud follows the same formula, PVU per core times the cores the IBM software runs on. On AWS and Azure the unit you are billed for is typically the virtual CPU, and IBM publishes how cloud virtual cores map into the PVU calculation. The practical effect is that the size of the instance you choose, and how many of them you run, drive the entitlement you need. Scaling out an auto scaling group or a fleet of instances scales the PVU requirement with it, and that is exactly where cloud deployments drift past their entitlement without anyone deciding to over deploy.
Sub-capacity still applies, and so does ILMT
Cloud is not an exemption from sub-capacity rules. To license only the virtual cores the software uses rather than something larger, you still need the IBM License Metric Tool or an approved alternative deployed within 90 days of first eligible deployment, running continuously, with quarterly reports retained for two years. Running ILMT across ephemeral cloud instances is harder than on a static estate, and that difficulty is precisely why cloud sub-capacity claims fail: agents do not follow instances that spin up and down, and the coverage gaps that result read as undefended capacity.
Bringing your own license is not bringing your own rules. The PVU metric, the sub-capacity conditions, and the ILMT obligation all move to the cloud with the software. A clean on premises position can quietly break the moment the same workload is rehosted without the tooling following it.
Where cloud PVU becomes a finding
- Elastic scaling outruns entitlement. Instances added under load lift the cores running IBM software, and the peak, not the average, is what an auditor measures.
- ILMT does not cover ephemeral instances. Short lived instances appear and vanish between scans, leaving capacity that was never measured and defaults to full counting.
- Mixed estates lose their reporting thread. Workloads split across on premises, AWS, and Azure fragment the ILMT record and create gaps at the seams.
- Provider billing mistaken for license compliance. Paying the cloud provider for compute says nothing about whether the IBM entitlement covers it.
BYOL keeps the IBM metric, the sub-capacity conditions, and the ILMT obligation fully in force in the cloud. Confirm that tooling actually covers your AWS and Azure instances, including the elastic and short lived ones, and that peak core counts are entitled, before IBM measures the gap for you.