A half-finished WebLogic or SOA Suite migration is the single most expensive state your estate can be in, because Oracle counts the old platform and the new one at the same time. This page shows exactly where the double-count comes from, what it costs at list, and the containment moves that reduce the claim before the audit script ever runs.
A half-finished WebLogic or SOA Suite migration is the single most expensive state your estate can be in, because Oracle counts the old platform and the new one at the same time. This page shows exactly where the double-count comes from, what it costs at list, and the containment moves that reduce the claim before the audit script ever runs.
Oracle's audit timing is not a coincidence, and after twenty five years on the buyer side of these negotiations I have never seen it behave randomly. Premier Support for Fusion Middleware 12c ends in December 2026, with Extended Support running out in December 2027. That published date is a calendar of customer distress, and Oracle's install base teams read it the same way you do: every account still on 12c must stand up a target platform before the deadline, which means a predictable twelve to eighteen month window in which old and new stacks run side by side. The mechanic that turns that window into money is one sentence from the price list licensing rules: Processor means all processors where the Oracle programs are installed and/or running, with the count derived by multiplying total cores by the applicable core factor. Read it literally, because Oracle's auditors do. A staged target cluster with WebLogic binaries laid down, domains configured, and zero production traffic routed to it is fully countable. So is a decommission candidate that still has the software on disk because nobody has run the uninstaller yet. Neither state has a "migration" exemption anywhere in the standard OMA or OLSA, and no support renewal quote grants one. The practical consequence is that the migration project, not the eventual decommission, is the exposure event. Your architects measure progress in workloads cut over. Oracle measures it in installed footprint, which peaks precisely when your project plan says you are two thirds done. If you are planning the exit path itself, treat the licensing sequence as a design input, not an afterthought: our guide to migrating off Oracle WebLogic and the licensing exit sets out the order of operations that keeps the countable footprint from doubling.
A staged target cluster with WebLogic binaries deployed and no traffic routed to it is fully countable, because the price list counts installation, not usage.
Put numbers on it. Take a common production estate: four application nodes, two sockets each, twelve cores per socket. That is 96 cores. Apply the 0.5 x86 core factor and you have 48 Processor licences. At the WebLogic Suite list price of $45,000 per Processor, that single cluster is $2,160,000 at list with $475,200 in annual support at the standard 22 percent rate. Now stand up a target cluster of comparable size for a six to twelve month cutover, as almost every 12c migration does, and the installed footprint becomes 96 Processors. At Suite list that is $4,320,000 with $950,400 per year in support. Oracle will not usually invoice the full doubled figure, but that number is the anchor the auditor opens with, and every discount you negotiate is measured against it rather than against the 48 Processors actually serving traffic. The single largest re-scoping lever available is the edition question. The Suite-over-Enterprise Edition gap is $20,000 per Processor. On 48 Processors that is $960,000 at signature and $211,200 every year thereafter. If your teams are not using Coherence data grid, ADF advanced features, or the Suite-only management tooling, you are paying Suite prices for Enterprise Edition function, and mid-migration is the moment to prove it with deployment evidence rather than accept the SKU you inherited. Work the same test on the target platform before binaries land, because re-scoping an unbuilt cluster costs nothing and re-scoping a built one costs a negotiation. Our breakdown of Oracle middleware audit risk and the true-up math shows how these edition assumptions compound once SOA and repository databases enter the count.
| Line item (per Processor) | List price | Annual support at 22% | 48 Processors (single cluster) | 96 Processors (parallel run) |
|---|---|---|---|---|
| WebLogic Server Standard Edition | $10,000 | $2,200 | $480,000 list / $105,600 support | $960,000 list / $211,200 support |
| WebLogic Server Enterprise Edition | $25,000 | $5,500 | $1,200,000 list / $264,000 support | $2,400,000 list / $528,000 support |
| WebLogic Suite | $45,000 | $9,900 | $2,160,000 list / $475,200 support | $4,320,000 list / $950,400 support |
| Suite over Enterprise Edition premium | $20,000 | $4,400 | $960,000 list / $211,200 support | $1,920,000 list / $422,400 support |
Two actions follow. First, cap the parallel window contractually or architecturally: agree a written decommission date with Oracle in the same document that covers the migration, or keep the target platform below the licensed Processor count until the source nodes are uninstalled and the removal is evidenced with timestamps. Second, settle the edition question before the target build, not during the audit. A downgrade decided in a design review is a $960,000 saving on paper. The same downgrade argued from an audit finding letter becomes a concession Oracle will price against your renewal.
SOA Suite never travels alone, and that single fact explains why SOA findings routinely land at three times what the customer expected. At $57,500 per Processor, SOA Suite is already the most expensive line in the middleware stack, but it runs on WebLogic and it requires an Oracle Database repository to hold process state, BPEL execution data, rules and BAM telemetry. Price a modest four-processor Intel SOA deployment honestly and you get $230,000 of SOA Suite plus $180,000 of WebLogic Suite at $45,000 per Processor, roughly $410,000 before anyone opens the database question. Then add the repository: Database Enterprise Edition lists at $47,500 per Processor, and Standard Edition is not a viable production platform for SOA, so the cheap answer is not available. Support runs at 22 percent on all three layers, so the same four processors carry roughly $100,000 of annual support even before adapters. Auditors know this arithmetic better than most architects do, which is why a SOA finding is opened first and everything else is scoped around it.
Two traps convert an already expensive baseline into a claim you cannot argue down. First, there is no discounted standalone Oracle Service Bus SKU. If OSB is running anywhere, including a single message-routing node someone stood up during a proof of concept, the full SOA Suite price applies to every processor where it is installed and running. Teams that assume "we only use the bus" are pricing against a product that does not exist. Second, adapters are a named primary audit finding. SAP, Siebel and industry-specific connectors used without the corresponding pack licence are exactly what LMS looks for, because they are trivially detectable and impossible to dispute after the fact. Mid-migration, this is where discipline collapses: teams leave adapters and partitions enabled on the source platform while they test replacements on the target, and each enabled adapter, partition and managed server adds countable footprint on both sides. Our Fusion Middleware and SOA Suite licensing guide sets out how these layers stack against each other.
If OSB is running anywhere, the full SOA Suite price applies, because Oracle never built a cheaper standalone bus to sell you.
The containment move is unglamorous and effective: before you respond to anything, produce a written inventory of every enabled adapter and partition on the source estate, disable and document what the migration no longer needs, and confirm the repository database edition and processor count separately from the application tier. Do not let a single scoping conversation cover SOA, WebLogic and Database at once, because that is how a three-layer claim gets agreed in one sentence.
Oracle's 2026 partitioning policy still classifies VMware as soft partitioning, and it names the components explicitly: ESXi, vSphere, vCenter, vSAN and NSX. That classification applies to Database, WebLogic, SOA Suite and every other Oracle middleware program. There is no reduction for vCPU limits, CPU affinity settings or DRS host rules, and Oracle has never accepted them as a licensing boundary. Worse for a migration estate, Oracle extends the boundary to every host a virtual machine could reach through live migration, not every host it has actually reached. A vMotion-enabled cluster is therefore one licensable unit in Oracle's reading, regardless of where the workload sat on any given day. Mid-migration this is lethal, because target infrastructure is usually built inside the same cluster as the source for convenience, and the moment both platforms share a vMotion boundary the double-count and the cluster count compound each other.
The quantified pattern is consistent. Across roughly 30 to 40 Oracle virtualisation engagements in 2024 and 2025, soft-partitioned estates faced claims at a median 3.5 times the cores actually running Oracle, with 50 to 70 percent of that exposure avoidable through hard partitioning that was available in the environment but never configured. Read that carefully: the majority of the gap was not a policy dispute, it was an architecture that nobody isolated before the letter arrived. Arguing that the partitioning policy is non-contractual rarely lowers the number on its own, because Oracle prices from what the collection shows and negotiates from there. Containment is the lever, not interpretation, and our Oracle audit defense strategy treats scoping the collection as the first move rather than a technical afterthought.
That last point matters because LMS scripts now read VMware Tools metadata directly. Unless the collection is scoped in writing before it runs, the full cluster map flows into the audit response by default, and once Oracle holds it you are negotiating against your own submission. Three actions in week one: isolate Oracle workloads onto a dedicated, separately licensed cluster with no migration path into non-Oracle hosts, configure hard partitioning where the platform supports it and document the configuration with dates, and scope the script output to named hosts before any data leaves your network. Verbal comfort from an Oracle representative carries no weight; only written terms in the OMA, OLSA or a signed policy amendment change the position.
The clause that ends the most migration programs is the one nobody on the architecture team has read. The Oracle Technology Global Price List (April 16, 2026) states plainly that WebLogic Suite for Oracle Applications is eligible for use with Oracle Agile Applications only, and that any use of a "for Oracle Applications" limited-use program by other Oracle or third-party applications is not permitted. That restricted grant was cheap for a reason: it was never a general-purpose WebLogic entitlement. The single most common way a migration converts a cheap grant into a full-price finding is consolidation. Someone builds a shared target platform, moves the Agile-bundled WebLogic domain onto it alongside two custom Java applications and a third-party product, and in that moment the restricted grant no longer covers anything on that cluster. Oracle will not price the delta as a Suite-for-Applications shortfall; it will price full WebLogic Suite at $45,000 per Processor, with 22 percent support on top, across every processor in the consolidated footprint.
The mechanical problem is that Oracle's collection scripts cannot tell a restricted install from a full one. The binaries are identical. What separates a $0 finding from a seven-figure one is your own entitlement record: the ordering document, the restricted-use SKU number, and evidence of which applications sit in which domain. If your records are thin, the auditor's default assumption is full-use. Build that mapping before you consolidate anything, and keep the restricted domains on physically separate hardware for the duration of the migration. On Named User Plus minimums, treat every published figure as unverified: our own guidance carries both 10 NUP per Processor and 25 NUP per Processor for SOA depending on vintage, which tells you the minimum is contract-specific. Read your ordering documents, not any guide, including ours. The Fusion Middleware and Suite trap breakdown sets out where these restricted SKUs typically hide in older Agile and E-Business estates.
Cloud gets sold as a destination. In an audit, its real value is that it replaces an unbounded argument with a bounded number. Under Oracle's Authorized Cloud Environment policy for AWS, Azure and Google Cloud, the count is two vCPUs per Processor licence where hyper-threading is enabled and one vCPU per Processor where it is not. The trap that catches finance teams and, in our negotiating experience, roughly half of first-draft cloud licence models, is that the Processor Core Factor Table does not apply in Authorized Cloud Environments. Anyone who applies the 0.5 x86 factor on top of the 2:1 vCPU conversion has cut their requirement in half and is under-licensed by exactly that amount.
| Deployment | Counting rule | 64 vCPU / 32 core workload | WebLogic Suite at $45,000/proc |
|---|---|---|---|
| AWS, Azure, GCP, hyper-threading on | 2 vCPU = 1 Processor, no core factor | 32 Processors | $1,440,000 |
| AWS, Azure, GCP, hyper-threading off | 1 vCPU = 1 Processor, no core factor | 64 Processors | $2,880,000 |
| Wrong model: 2:1 plus 0.5 core factor | Not permitted | 16 Processors | $720,000 (half-licensed) |
| On-prem x86, isolated cluster | Cores x 0.5 core factor | 16 Processors | $720,000 |
| Soft-partitioned VMware cluster | All hosts in the vMotion boundary | Median 3.5x running cores | Claim scales with cluster size |
The strategic point is what this does to the conversation. A soft-partitioned VMware estate mid-migration invites an argument about hypervisor policy, DRS rules and cluster boundaries, and our engagement data across roughly 30 to 40 virtualisation cases in 2024 and 2025 shows those estates drawing claims at a median 3.5x the cores actually running Oracle. A dedicated cloud tenancy or a physically isolated on-prem cluster removes the argument entirely. Oracle counts vCPUs or cores you defined, you produce the same number, and the finding closes. Pair that with the processor licence recovery sequence so the vacated on-prem entitlements are formally released rather than left installed and countable.
Acknowledge the letter, name a single point of contact, and confirm nothing about scope. An acknowledgement is not an agreement to a data collection method, a tool, an entity list, or a measurement date, and every one of those four is negotiable before a script runs. Then freeze installs on target nodes immediately: because Processor means all processors where the programs are "installed and/or running," a half-provisioned target cluster is countable the same day it is built. Next, uninstall rather than stop the source-side components you no longer need. Shutting down a managed server, a SOA adapter, or an unused Suite feature changes nothing in the count. Deinstallation, evidenced with dated change tickets and node inventories, is the only action that removes a licensable footprint.
Do not build the defence on the argument that the partitioning policy is non-contractual. In our experience that position rarely reduces a claim on its own, and verbal comfort from an Oracle representative carries no weight at all. Only written clauses in the OMA, the OLSA, or a signed policy contract hold. Negotiate from the exit, not from the deadline: the December 2026 support cliff is your leverage because a customer with a credible, costed WebLogic licensing exit can decline a Suite true-up that Oracle needs more than you do.
Deinstallation is the only action that removes a licensable footprint; stopping a service removes nothing.
Yes. The Processor definition in the Oracle Technology Global Price List covers all processors where the programs are installed and/or running. A staged target node with binaries deployed and no traffic is fully countable, which is why mid-migration estates get double-counted. Uninstall, do not just stop, anything you are not using.
On a 96-core x86 estate (48 Processor licences after the 0.5 core factor), WebLogic Suite is $2,160,000 at list with $475,200 in annual support. Running a comparable target cluster alongside it during cutover doubles that footprint on paper. Downgrading Suite to Enterprise Edition where features allow saves $20,000 per Processor, or $960,000 on 48 processors.
Not under Oracle's stated policy. Oracle classifies VMware as soft partitioning and gives no credit for vCPU limits, affinity rules or DRS configuration, and it extends the boundary to every host a VM could reach through live migration. The only reliable containment is a physically separate, separately licensed cluster with migration paths removed and documented.
SOA Suite runs on WebLogic and requires an Oracle Database repository, so all three layers must be licensed. A four-processor deployment is roughly $410,000 in SOA Suite plus WebLogic Suite before the repository database at $47,500 per Processor for Enterprise Edition. There is no cheaper standalone Service Bus SKU to fall back on.
It can, but the arithmetic changes. In Authorized Cloud Environments two vCPUs equal one Processor licence with hyper-threading enabled, one vCPU per Processor without, and the Core Factor Table does not apply. Applying the 0.5 factor on top of the 2:1 conversion under-licenses you by half, which is a common self-inflicted finding.
Only in a controlled, written form and only after you have established your own count. Disclosing an in-flight migration invites Oracle to price both platforms and to treat the December 2026 Fusion Middleware 12c support deadline as commercial leverage. Verbal assurances from Oracle staff carry no contractual weight, so insist on written terms in the OMA or an amendment.
The strategic approach for Oracle audit defense across LMS, license verification, and contractual response. Beyond the tactical playbook.
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.
