Oracle classifies Nutanix AHV as soft partitioning, which means every physical core in every node can be pulled into scope, not just the cores running your database. This page breaks down the AHV-specific core-count mechanics, how live migration widens exposure, and whether affinity policies survive an audit.
How to Negotiate Your Oracle SaaS Renewal: The Five Moves at the Table
Scope before price: strip the 18 to 32 percent of inactive bundle modules first. Kill the escalator with a 0 to 3 percent cap that survives the term, trade term for protections, refuse the easiest-path module bundling, and close on Oracle's May 31 clock.
Oracle classifies Nutanix AHV as soft partitioning, which means every physical core in every node can be pulled into scope, not just the cores running your database. This page breaks down the AHV-specific core-count mechanics, how live migration widens exposure, and whether affinity policies survive an audit.
On Nutanix AHV, the license unit is the physical node, not the virtual machine. You cannot license at the vCPU level, at the VM level, or at any sub-node granularity. That single fact drives everything else on this page. When Oracle audits a Nutanix estate, the opening question is not "how many cores is the database using" but "how many physical cores exist in every node the database could touch." The gap between those two numbers is where seven-figure claims come from.
Oracle's partitioning policy is binary. A hypervisor either qualifies as hard partitioning (only the physical cores running Oracle need licensing) or it is treated as soft partitioning (every physical core across the entire cluster must be licensed). AHV sits firmly in the soft partitioning category, alongside VMware ESXi, Microsoft Hyper-V, and generic KVM. Oracle's published server partitioning policy does not name AHV at all, and that silence is not neutral: it means AHV gets treated exactly like the technologies Oracle already contests. This is the same trap documented on our Oracle on VMware and Oracle on Hyper-V pages, applied to Nutanix's KVM-derived hypervisor.
The smallest licensable unit on Nutanix is a full physical node. There is no vCPU-level license, so the cluster boundary is the cost boundary.
Oracle's Processor metric requires that every physical processor core on every server where the software is installed and/or running be counted as a fraction of a processor license, with the fraction set by the Core Factor Table. For the current Intel Xeon portfolio the factor is 0.5, meaning each physical core counts as half an Oracle processor license. The January 2026 Core Factor Table update pulled the newer Xeon 69xxP, 67xxE, 67xxP, 65xxP, and 63xxP variants plus the X55xx and E54xx series all into the 0.5 category, so the entire current Intel Xeon lineup now sits at 0.5. AMD EPYC also carries 0.5. In practice, on modern hardware, you divide total physical cores by two to get processor licenses required.
The severity comes from cluster geometry. A four-node cluster of 32-core servers is 128 physical cores. If isolation evidence fails, that is 128 cores of Oracle exposure, which at 0.5 becomes 64 Enterprise Edition processor licenses. At the $47,500 list price per Database Enterprise Edition processor, that is roughly $3.04M in license before any support. The 22 percent annual support (about $10,450 per processor) then attaches on top, indefinitely. Bigger hardware makes the number worse fast: a two-socket AMD EPYC 9754 box carries 256 cores, which at 0.5 is 128 processor licenses, or $6,080,000 at list before discounts and before support.
| Cluster | Physical cores | Core factor | Processor licenses | EE list license |
|---|---|---|---|---|
| 4 nodes x 32 cores | 128 | 0.5 | 64 | $3,040,000 |
| 6 nodes x 32 cores | 192 | 0.5 | 96 | $4,560,000 |
| 2 sockets x EPYC 9754 (256c) | 256 | 0.5 | 128 | $6,080,000 |
| 8 nodes x 48 cores | 384 | 0.5 | 192 | $9,120,000 |
Every figure above is list price only. Enterprise Edition discounts ran 35 to 70 percent across 2024 and 2025 Oracle negotiations (an observed band from our engagement files, not a published Oracle rate), so real exposure varies. But note carefully what the discount does and does not fix: a discount reduces the price per license, it does not reduce the count of cores Oracle claims. The count is the fight. Win the core-count argument and a 60 percent discount on 64 processors beats a 60 percent discount on 128 processors every single time.
AHV includes live migration, its VMware-equivalent feature for moving running VMs between nodes. This is the specific mechanism that converts a single-node database into a whole-cluster liability. Even if your Oracle VMs are pinned today, Oracle examines historical placement data. If a VM crossed a node boundary at any point, Oracle argues all nodes were in scope at the time of migration. You are not defending a snapshot, you are defending a history.
The audit triggers on Nutanix are predictable: cluster expansions, node replacements, and any live migration event that crosses the isolation boundary. Each of those is a moment where the scope can widen and where the audit team looks for drift. House of Brick has flagged the same mechanic, describing live migration "drift" as the core cause of Nutanix audit risk. The precedent everyone cites is Oracle vs Mars, where Oracle's LMS team took the position that because vMotion allowed VMs to migrate between hosts, all cluster processors required licenses. That reasoning transfers directly to AHV live migration.
Oracle prices the question by what your VMs could do. You settle it by proving what they actually did.
The practical implication is that your defense is evidence-driven, not architecture-driven. The buyer who keeps a machine-generated migration audit trail can prove where VMs actually ran and cap the claim to reality. The buyer who cannot produce that record is arguing against Oracle's worst-case reading with nothing but assertions. Our soft partitioning audit-defense guide and the VMware cluster-scope page both walk the same evidentiary logic in detail.
This is the central question and the sources genuinely conflict, so we will not pretend the answer is clean. There are three positions in the market, and you should understand all three before you bet an estate on any of them.
One camp holds that AHV affinity rules are an operational configuration, not a licensing control mechanism. They can be overridden manually or modified by an administrator, so Oracle does not treat them as a boundary. In this reading, VMware vSphere, Hyper-V, and AHV never qualify regardless of pinning, affinity rules, or host groups. Everything soft-partitioned licenses the whole machine.
House of Brick has stated that Oracle recognizes Nutanix affinity rules for license containment "when configured properly," arguing that affinity rules pin Oracle VMs to specific hosts and satisfy the contractual requirement to license only where Oracle is installed or running. The same source qualifies this: soft partitioning on AHV is "technically, yes, but risky," because configuration changes can easily break the setup, and they still recommend licensing all cores on the physical host as the safe route.
The reconciling position, and the one we work from in practice, is that AHV containment is defensible only when sub-cluster isolation is documented and enforced. Without documented isolation, Oracle counts every core in every node. Affinity by itself is a preference, not a wall. The distinction that matters is between a soft affinity rule (a scheduler hint that can be violated under load) and hard isolation with enforcement plus an audit trail proving it was never violated. The first loses at audit. The second gives you a real seat at the negotiation table.
| Control | What it is | Audit standing |
|---|---|---|
| Soft affinity rule | Scheduler preference, can be overridden | Weak, treat as no containment |
| Hard affinity / pinning | VM bound to named hosts | Contested, needs evidence |
| Documented sub-cluster isolation + audit trail | Enforced boundary with machine-generated proof | Strongest defensible position |
| No policy at all | VMs float freely | Whole-cluster claim |
Our companion analysis on whether CPU pinning reduces the license count lays out the exact evidence test, and the KVM and OLVM core-count page covers the same argument on the underlying KVM technology that AHV is built on.
There is a clean win available in AHV, and it is worth taking. AHV lets you designate certain physical hosts as storage-only nodes. These serve only I/O to the cluster and are incapable of running virtual machines. Because Oracle cannot run on them, storage-only hosts do not require Oracle licenses. This is not a gray area, it is a structural exclusion.
Nutanix strengthens this with a "Never Schedulable" node policy plus AHV Auditing. Never Schedulable cannot be accidentally toggled off, which removes the risk that an errant click turns a storage node into a compute node and silently expands your licensed footprint. Combined with the machine-generated audit trail of migration changes, this is exactly the kind of enforced-plus-evidenced control that survives scrutiny. If you are running Oracle on AHV, the target architecture is a compute pool small enough to license fully, with storage-only nodes carrying the capacity, and Never Schedulable locking the boundary.
Storage-only nodes require zero Oracle licenses. Never Schedulable makes that boundary tamper-resistant, which is what turns it from a claim into evidence.
Nutanix environments routinely generate the largest compliance gaps in Oracle audits. A single Oracle Database VM on a four-node Nutanix cluster can create $1.9M or more in license obligations, and findings of $2M to $14M are common in the field. Across roughly 30 to 40 Oracle virtualization engagements between 2024 and 2025, soft-partitioned estates faced license claims a median 3.5 times the cores actually running Oracle. That multiple is the whole game: Oracle is billing you for the 128 cores in the cluster when maybe 32 are doing the work.
The counter-evidence is equally clear. Defended estates that produced real placement evidence settled the VMware element at 10 to 25 percent of opening claims. In one Nutanix case, a technical inventory demonstrating VM pinning challenged the LMS calculation and produced a settlement covering actual allocated cores plus a 20 percent buffer, cutting the claim to $1.4M. The lesson is consistent: the opening number is designed to anchor high, and evidence is what moves it. Buyers who show up with a documented, enforced, and audit-logged boundary do not pay the cluster-wide number.
Here is the leverage most buyers leave on the table. Oracle's partitioning policy is not part of your contract. It is a published position paper. That cuts both ways: it is not binding on you, but Oracle's audit teams apply it without exception as if it were. Your signed agreement licenses processors where the programs are installed and/or running. Oracle's published licensing positions sit outside that contract. The soft-partitioning-widens-scope argument is a policy position, and your "installed and/or running" contract language is the counter-position.
That does not mean you can wave the contract and walk away, because Oracle applies the policy in every audit regardless. It means your negotiation runs on two rails at once: the contractual argument that scope is limited to where the software actually ran, and the evidentiary argument proving where that was. The two reinforce each other. When you understand both the partitioning policy mechanics and the path to Oracle-approved hard partitioning, you can hold the line at settlements measured against cores actually running Oracle rather than cores that theoretically exist in the cluster.
Soft partitioning. Oracle's published partitioning policy does not recognize AHV as an approved hard partitioning technology, so it falls in the same bracket as VMware ESXi, Hyper-V, and generic KVM. In practice that means all physical cores across the entire cluster can be pulled into scope, not just the cores running Oracle. AHV is Nutanix's flavor of KVM, and Oracle does not name it in the policy, which defaults it to soft.
No. The smallest licensable unit on Nutanix is a full physical node. There is no vCPU-level or VM-level license, so you count every physical core on every node where Oracle is installed and/or running, then apply the 0.5 core factor. This node-level minimum is why cluster geometry drives the whole cost.
The sources genuinely conflict. One camp treats affinity as an operational hint Oracle ignores, another says it can hold when configured properly, and the reconciling view is that only documented, enforced sub-cluster isolation with an audit trail is defensible. Soft affinity by itself is treated as no containment. If you rely on affinity, you must be able to prove enforcement and that the boundary was never violated.
Live migration is the mechanism that widens scope from one node to the whole cluster. Oracle examines historical VM placement, and if a VM crossed a node boundary at any point, Oracle argues all nodes were in scope at that time. Cluster expansions, node replacements, and boundary-crossing migrations are all documented audit triggers. Your defense is a machine-generated migration audit trail proving where VMs actually ran.
No. Storage-only nodes serve only I/O and cannot run virtual machines, so Oracle cannot run on them and they require zero Oracle licenses. Combined with the Never Schedulable node policy, which cannot be accidentally disabled, this is the one uncontested license reduction available on AHV. Use it to shrink the compute pool you have to license fully.
A single Oracle Database VM on a four-node cluster can create $1.9M or more in license obligations, and audit findings of $2M to $14M are common. Soft-partitioned estates have faced claims a median 3.5 times the cores actually running Oracle. Defended estates with real placement evidence, however, settled at 10 to 25 percent of the opening claim, so evidence is what moves the number.
Oracle Exadata can lock you into full core licensing across X9M, X10M, and Cloud at Customer. The buyer side strategy to size the platform and cut the bill.
Gated with a work email on the download page. No sales follow up you did not ask for.
Get the White Paper →
500+ enterprise clients. 11 vendor practices. Industry recognized. One conversation can change what you pay for the next three years.
One buyer side briefing a week. Renewal signals, audit moves, and the levers that work. No vendor spin.
