When a divestiture closes, the Oracle licenses the carved-out unit depends on stay with the parent, and the clock on the transition period starts running whether anyone told Oracle or not. This is the sequence that gets both entities licensed without paying twice, plus the two policies (assignment consent and matching service levels) that Oracle will use to price your urgency.
When a divestiture closes, the Oracle licenses the carved-out unit depends on stay with the parent, and the clock on the transition period starts running whether anyone told Oracle or not. This is the sequence that gets both entities licensed without paying twice, plus the two policies (assignment consent and matching service levels) that Oracle will use to price your urgency.
Most carve-out teams open the Oracle Master Agreement, find the assignment clause, and assume the problem is a consent negotiation. It is not. The operative failure is upstream, in the definition of Customer. Oracle's grant runs to the named legal entity and the affiliates it controls as of the effective date. The second NewCo leaves the parent's corporate family at close, it drops out of the license grant. Nothing on the server changes. The databases keep running, the middleware keeps serving, the batch jobs keep firing, and every one of those instances is now unlicensed use accruing daily at an entity with zero entitlement. Oracle does not novate anything automatically. Standard terms typically allow the divested entity to keep running under the seller's agreement for a short transition period, commonly 90 days, after which it must buy its own licenses. And the specific structure you need, moving a defined subset of entitlement rather than the whole company, is generally not permitted absent Oracle's written consent, even in contracts that contain full-company assignment language for a merger or asset sale. Read your assignment clause carefully: full-company language and partial-transfer permission are not the same right, and Oracle will tell you so at the worst possible moment. Meanwhile the parent has its own exposure, because it now holds entitlement it no longer fully consumes and cannot freely resell or shrink. Divestitures are a documented Oracle audit trigger for exactly this reason. Both entities are distracted, the CIO on each side assumes the other one handled it, and Oracle License Management Services knows the calendar as well as your bankers do. Treat the day the SPA signs as day one of the licensing clock, not the day someone gets around to reading the Oracle transfer and assignment restrictions.
The assignment clause is not your problem, the definition of Customer is, and NewCo fails it the moment the deal closes.
There are only four structures that actually work, and choosing among them is a decision about where you want Oracle's veto to sit. Route one is a formal license assignment agreement executed through Oracle Sales, moving a defined slice of entitlement from parent to NewCo. It is clean on paper and slow in practice, because Oracle prices consent and controls the timeline. Route two is NewCo buying a net-new standalone contract at first-time-customer pricing, which sounds expensive until you price the consent premium on route one and the stranded support the parent carries either way. Route three is an extended transition service agreement where the parent retains ownership and provides the service to NewCo as a customer, deferring the licensing event without eliminating it. Route four is migrating NewCo off Oracle entirely inside the TSA window, viable only where the estate is small, non-transactional, or already headed to PostgreSQL. In our negotiating experience, route two paired with a long TSA is the strongest buyer position, because it takes Oracle's consent out of the critical path: NewCo negotiates as a greenfield prospect with a walkaway, rather than as a hostage whose systems go dark in 90 days. Discount bands of 40 to 70 percent off list are achievable on a new standalone deal, and every dollar off net license is worth roughly $2.10 over a five-year hold because support runs at 22 percent of net. Size the NewCo estate for Standard Edition 2 wherever the workload permits: $17,500 per occupied socket against $47,500 per processor on Enterprise Edition, with no options catalog and a materially smaller audit surface.
| Route | Oracle consent risk | Time to close | Parent stranded cost | Buyer leverage |
|---|---|---|---|---|
| 1. Formal assignment of a license slice | High, written consent required for partial transfer | 3 to 6 months, Oracle-controlled | Partial, residual support pool | Low, Oracle prices the consent |
| 2. Net-new standalone contract for NewCo | None, no consent needed | 6 to 12 weeks, buyer-controlled | Full, parent keeps everything | High, 40 to 70 percent discount band in play |
| 3. Extended TSA, parent retains licenses | Low near term, deferred not removed | Immediate at close | Full, but recovered via TSA fees | Medium, resets when TSA expires |
| 4. Migrate off Oracle inside the TSA window | None | 9 to 18 months | Full, plus termination decision | Highest, credible walkaway |
Whichever route you pick, decide it before close and write the mechanics into the TSA, not after. Route three is not a decision, it is a runway, and it expires.
Partial transfers are not permitted unless Oracle agrees in writing, and Oracle knows exactly what that piece of paper is worth to you. The moment you file an assignment request, you have told Oracle three things: a deal is closing, a deadline exists, and you cannot walk away. In my experience across a couple of dozen carve-outs, the request itself is routinely routed through a compliance review before it reaches contracts, which means the first response you get is often a questionnaire about deployment, virtualization, and options usage rather than a redlined assignment agreement. Price the consent before you ask for it. The standard asks are predictable: an OCI or SaaS commitment on the NewCo paper, a first-year support uplift above 22 percent, re-papering the parent's discount level down to something closer to current list, and quietly dropping legacy terms such as an expired ULA's certified quantities, an older audit clause with notice periods you actually like, or a favorable Named User Plus minimum. Treat each as a separate line to be refused, not a package to be traded. The rules governing what Oracle will and will not let you move are the ceiling on your ask, but the discount history is the floor you defend.
Notifying Oracle before your own deal terms are settled converts a licensing formality into a negotiation you have already lost.
Most carve-outs fail in the TSA, not in the Oracle contract, because M&A counsel drafts the TSA around IT service levels, uptime, and cost recovery while treating the underlying software rights as somebody else's problem. Oracle's position is straightforward: the parent is licensed, NewCo is a different legal entity, and the parent running production workloads for NewCo's benefit looks a great deal like unlicensed hosting. That argument is not automatically correct, and the hosting and third-party datacenter rules are worth reading closely before you concede it, but you do not want to be constructing the counterargument for the first time under an audit letter eleven months after close. Either obtain a written Oracle acknowledgement of TSA use, or have counsel document the legal position at signing, with the deployment inventory attached. Then fix the term. A 90-day transition window, which is the default many contracts imply, is not survivable: an Oracle negotiation cycle for a first-time standalone contract, from sizing through discount approval to signature, routinely runs two to three quarters. Draft for 18 to 24 months with pre-priced extension options agreed at close, because an extension negotiated under time pressure while Oracle watches is an extension the parent will price like Oracle does.
Sellers routinely assume that once the divested unit is gone, the parent simply drops the support line attached to those licenses and banks the saving. Oracle's technical support policies say otherwise: all licenses in a given license set must be supported at the same service level, you cannot support a subset of a license set, and if you want to stop paying for part of it, the license set itself must be reduced by terminating the unsupported licenses. There is no reinstatement right on a terminated license, and no partial-support middle ground. Worse, when Oracle reprices the surviving support stream, it recalculates the remaining contract at then-current list against the reduced quantity, which in practice erases the historical discount that was priced across the full volume. In our negotiation experience the repricing usually returns 60 to 90 percent of the support you tried to shed, and in a few cases the parent's post-split support bill exceeded the pre-split bill. That is the trap: the seller pays for capacity it divested, or terminates entitlement it may need back in eighteen months. Fix it structurally before signing, not after: segment CSI numbers and license sets so the divested unit's entitlements sit in their own set, then terminate deliberately. See our guidance on optimizing the Oracle footprint before renewal for the sequencing, and read the Costco support termination case for what a clean, pre-planned reduction actually delivers.
| Scenario, 100 EE Processors at 60 percent discount | Parent support outcome |
|---|---|
| Pre-split: 100 Processors, net $1.90M, support at 22 percent | $418,000 per year |
| Naive drop of 30 Processors' support, no termination | Rejected: cannot support a subset of the license set |
| Terminate 30 Processors, Oracle reprices remaining 70 | Discount recalculated toward list, saving often under 10 percent |
| Segmented CSI and license set agreed pre-close, 30 moved to NewCo | Support falls broadly in line with the 30 percent reduction |
There is no partial-support middle ground: either the license set stays whole and fully paid, or the entitlement is terminated with no route back.
Price the residual honestly. If the modeled stranded support is $300,000 a year for three years, that is a $900,000 line item in the purchase price negotiation, and the buyer, not the seller's IT budget, should carry it. Put the number in the SPA, not in a footnote.
The single most expensive mistake in a carve-out is licensing NewCo as a scale model of the parent. The parent's estate reflects a headcount, a consolidation history, and an options catalog NewCo does not have. Start from NewCo's actual users and actual sockets. Named User Plus is priced at exactly one fiftieth of Processor on the Database lines, $950 against $47,500 on Enterprise Edition under the April 16, 2026 Technology Global Price List, so the crossover is 50 real users per Processor. Below that, NUP wins outright, subject to the floors: 25 NUP per Processor on Enterprise Edition, 10 per server on Standard Edition 2, and you license the higher of the floor and real headcount. Then test the SE2 downshift, which is usually the biggest lever available to a smaller entity: $17,500 per occupied socket, sockets counted rather than cores, a two-socket cap per server, no options catalog, and a materially smaller audit surface because there is nothing to accidentally enable. Leaving the parent's option stack behind is real money: Partitioning at $11,500 per Processor and RAC at $23,000 per Processor are line items NewCo should have to justify from scratch, not inherit. If a feature comes across, confirm whether it is genuinely required or merely default-on, as with Advanced Compression.
| Metric decision for NewCo | List anchor, April 16, 2026 | When it wins |
|---|---|---|
| Database EE, Processor | $47,500 per Processor | Above 50 real users per Processor |
| Database EE, Named User Plus | $950 per NUP, 25 per Processor floor | Under 50 users per Processor, countable population |
| Standard Edition 2, socket | $17,500 per occupied socket, 2-socket cap | Two sockets or fewer per server, no EE options needed |
| Partitioning option | $11,500 per Processor | Only with a documented partitioning workload |
| Real Application Clusters | $23,000 per Processor | Only with a contractual HA requirement |
On price, treat 40 to 70 percent off list as the realistic band for a first standalone agreement, and push toward the top of it by making the deal a new-name contract with a clean-slate audit position rather than a continuation of the parent's paper. First-year support is the negotiable line and can land at 15 to 18 percent of net as part of a package; renewals will revert to 22 percent, so extract the concession where it exists. Then argue in multiplier terms, because that is where the leverage sits: with support at 22 percent of net and 8 percent annual uplifts, one dollar off net license is worth roughly $2.10 across a five-year hold. Negotiate the license discount, cap the uplift, and get the entity definition wide enough that future NewCo subsidiaries do not trigger the same affiliate entitlement problem again. Refuse any commitment sized on the parent's peak footprint.
The order of operations decides the price. Before the SPA signs, run a deployment inventory mapped to legal entity and CSI number, because Oracle's support pool is organized by CSI and you cannot split what you have not first attributed. Then pull the OMA and read four definitions: Customer, Affiliate, assignment, and territory. Those four lines, not the sales rep's tone, determine whether the carved-out unit has any claim at all. Model both entities' post-close support bills next, including the matching service levels exposure that keeps the parent paying 22 percent on licenses it no longer runs, and price the walk-away in the same spreadsheet: SE2 at $17,500 per occupied socket against Enterprise Edition at $47,500 per Processor is usually the credible alternative for a smaller NewCo. Only after those three artifacts exist do you insert the TSA clauses and open the Oracle conversation with a defined ask, a defined date, and a defined exit. Assign owners explicitly: legal owns the definitions, IT asset management owns the inventory, and one named executive owns the Oracle relationship so the vendor cannot work three channels at once. Read the rules on what Oracle licenses you can actually move or reassign before you draft the ask, and the affiliate and subsidiary sharing constraints if either side plans to keep the unit inside the corporate family for a transition period.
No. The licenses were granted to the named parent entity and remain with the parent, and Oracle does not novate them automatically. Standard terms typically allow the divested unit to keep using the programs for a short transition period, often 90 days, under the seller's agreement, after which it needs its own entitlement. Partial transfers of a subset of licenses generally require Oracle's written consent.
Ask for 18 to 24 months with priced extension options, not the 90 days Oracle's standard language implies. A standalone Oracle negotiation plus a possible SE2 or off-Oracle migration realistically takes 9 to 18 months, and a TSA that expires mid-negotiation hands Oracle the deadline. Tie the exit trigger to NewCo contract signature rather than a fixed calendar date.
Not cleanly. Oracle's technical support policies require all licenses in a license set to sit at the same service level, and you cannot support a subset, so Oracle will require termination of the unsupported licenses. Termination is irreversible: there is no reinstatement right, and Oracle's repricing recalculates the remaining contract as if prior discounts never applied. Segment license sets and CSI numbers before close if you want a clean split.
It depends on the discount level baked into the parent's contract versus what NewCo can win as a new deal. Assignment preserves grandfathered net prices and metrics but gives Oracle a consent veto it will price. A net-new NewCo contract at 40 to 70 percent off list, ideally downshifted to Standard Edition 2 at $17,500 per occupied socket, often beats an assigned Enterprise Edition estate at $47,500 per Processor once the options stack is dropped.
It frequently does. Oracle treats mergers, acquisitions, and divestitures as trigger events and knows both sides are distracted and deadline-bound at close. Assume any assignment or consent request will surface a compliance question, so complete your own entity-level deployment inventory before you contact Oracle, and refuse to let a license review be bundled into the assignment transaction.
Check the divestment clause if you have one, since it usually names the responsible party and the notice period before license rights terminate. In practice the parent should lead, because it holds the contract, the CSI numbers, and the commercial relationship. Do not notify until the internal deal terms, TSA length, and your target structure are settled, since an early notification with no position simply starts Oracle's clock.
How Oracle E-Business Suite is licensed in 2026, where the audit traps live, and how to control EBS spend across modules, named users, and the support stack.
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.