Oracle
Oracle ULA exit, $42M kept off the table.
Oracle quoted $48M to renew. The retailer certified 41,000 processors instead, kept every license, and walked. The certification play, step by step.
Oracle quoted $48M to renew. The retailer certified 41,000 processors instead, kept every license, and walked. The certification play, step by step.
Oracle priced the renewal at $48M for growth that had already stopped. Certification turned the same estate into 41,000 perpetual processor licenses at no license cost, and kept roughly $42M off the table.
The retailer exited because the renewal quote was $48M for three years while its Oracle deployment had stopped growing 18 months earlier. A ULA renewal prices Oracle's forecast of your growth. Certification prices what you actually deployed, and here the two numbers had nothing to do with each other.
Eighteen months of instance level data showed the Oracle footprint flat while the application portfolio kept growing around it. That single chart reframed the renewal conversation internally. The question stopped being whether $48M was a fair price and became what, exactly, it would be buying.
The dataset behind the chart was deliberately simple: instances by environment by quarter, options usage, and where each new application workload had landed. Nothing exotic was required. The discipline was pulling it before Oracle framed the conversation, not after.
The gate was simple: if measured growth could not consume the renewal price, certification wins. It could not, so the team built the exit. Nothing about the choice was ideological; it was arithmetic with a deadline attached.
Setting the rule before engaging Oracle mattered as much as the rule itself. A decision criterion agreed in a calm quarter does not bend in a pressured one.
Priced honestly, the renewal path cost $48M plus a repriced support stream, while the certification path cost the advisory and sweep effort and nothing in license fees. The exit economics of a ULA are asymmetric like this more often than buyers expect.
The comparison has to run over the same window, on the same estate, with the same assumptions about growth. Held to that standard, the renewal needed phantom deployment to justify itself.
One path buys insurance against growth that may not come. The other converts money already spent into a permanent asset.
The $48M covered three more years of unlimited deployment rights the telemetry said would go unused. It also reset the contractual baseline, since support under a new agreement is recalculated on the new deal's value under Oracle's support policies. And it deferred the counting question to a later certification, on whatever terms applied then.
Three cost lines never appear in a renewal quote, and each one landed in the model here.
Certification carried no license fee. The costs were real but bounded: the 14 week sweep, external advisory, and internal time from infrastructure and procurement teams. Netting those against the avoided renewal left roughly $42M kept off the table across the three year comparison window.
The certified estate left the retailer free to keep moving workloads off Oracle without wasting a prepaid unlimited right. That freedom does not appear in either quote, but it compounds. Every workload that left Oracle after the exit widened the gap between the two paths.
The internal case fit on two pages. Page one held the telemetry chart and the two priced paths. Page two held the risks of each path and the mitigation already underway for the exit: the sweep plan, the counting positions, and the review calendar.
Framing matters here. The exit was presented as capturing an asset already paid for, not as a bet against Oracle. That framing survived finance review because every number on the page traced to a document.
Renewal path versus certification path, three year view
| Dimension | Renew the ULA | Certify and exit |
|---|---|---|
| License cost | $48M over three years | $0, existing rights certified |
| License position | Unlimited during term, recount at next exit | 41,000 perpetual processor licenses |
| Support | Recalculated on the new agreement value | Flat, continues on the existing base |
| Audit posture | Deferred to next certification | Fixed entitlement, defensible count |
| Flexibility | Locked to Oracle growth assumptions | Free to move workloads off Oracle |
The team maximized the certified count by sweeping every legally countable deployment before the ULA expired, landing at 41,000 processor licenses. Certification converts unlimited deployment rights into a fixed perpetual grant, so every properly deployed processor on the certification date becomes a permanent asset. After that date, nothing can be added.
The sweep inventoried every Oracle instance across data centers, VMware clusters, and authorized cloud environments against the Oracle pricing and licensing definitions. Standby, disaster recovery, and partially deployed environments were validated line by line against the ULA certification clause.
The point was completeness, not inflation. Capture what genuinely exists, and finish planned rollouts while the unlimited right still covers them.
The certified position came in at 41,000 processor licenses across Database, RAC, and Diagnostics and Tuning. Oracle challenged the VMware cluster counting; the contract language and deployment evidence held, and the certification letter was accepted without adjustment.
How that evidence pack is built, and how the review conversation runs, is the subject of the companion ULA certification case study. The method itself lives in the certification guide.
Every workstream in the exit was scheduled backward from expiry, because value created after the freeze is value lost. The sweep finished with margin, the counting positions were papered, and the decision gate was revisited once more with final numbers before the letter was signed.
Deadlines are leverage in a ULA exit, but only for the side that planned for them.
The renewal conversation stayed open, deliberately, for the entire nine months. Closing the door early hands Oracle certainty for free. Keeping both paths alive keeps the pricing honest on the one you might still take, and costs nothing on the one you intend to take.
The rules the team held to were simple, and they are repeatable in any exit.
Across comparable engagements in 2024 to 2025, renewal quotes moved down 25 to 40 percent once a credible certification track was visible. The pattern held because the alternative was real, not performed. Leverage in a ULA endgame is not a negotiating style; it is the documented ability to walk.
This retailer certified and walked regardless of where the quote might have moved. That is the point of the gate: once the arithmetic says the renewal cannot pay for itself, a better bad deal is still a bad deal.
The exit workstreams and their owners
| Workstream | Owner | Deliverable |
|---|---|---|
| Deployment sweep | Infrastructure, with advisory oversight | Complete countable inventory |
| Counting positions | Licensing advisory | Evidenced VMware, DR, and cloud positions |
| Economic model | Procurement and finance | Both paths priced over three years |
| Oracle communications | Single negotiation owner | Controlled message, both paths open |
| Certification letter | Legal, signed at officer level | Accepted declaration |
The exit avoided the $48M renewal entirely, kept 41,000 perpetual processor licenses, and left annual support running flat on the existing base. Net avoided cost against the renewal path was roughly $42M over three years after advisory and sweep costs. No compliance finding followed, because the count had been built to survive one.
Certification itself triggered no support repricing; the certified licenses simply continued on the support contracts already in force. The discipline that matters afterward is structural. Oracle ties matching service levels to whole license sets, so any later attempt to drop support on part of the estate has to be planned along set boundaries, not assumed line by line.
The retailer left the exit with three things it did not have under the ULA: a fixed, evidenced entitlement, a support bill it could forecast, and no contractual reason to keep new workloads on Oracle. The estate stopped being a growth commitment and became an asset under management.
The certified record also changed the audit calculus. An auditor arriving after certification meets a documented entitlement and a documented count, which is the least profitable audit Oracle can run.
Three numbers define this engagement, and each one carries a lesson that scales beyond it.
Source: Redress Compliance advisory engagement file, ULA exit work 2024 to 2025.
The dollar figure scales with estate size, but the ratio does not change. When deployment has plateaued, certification converts a recurring negotiation into a one time exercise with a permanent asset at the end. The renewal, by contrast, guarantees you will hold this same negotiation again in three years, from a weaker position.
Benchmark your own situation against the nine month runway above. If your expiry is closer than that, the honest conclusion is not that exit is impossible, but that every remaining week costs certified value.
A ULA renewal prices the vendor's forecast of your growth. Certification prices what you actually built. Measure before you believe either number.
The reusable pattern is measure first, decide second. Every ULA decision should start from deployment telemetry, not from the renewal quote, because the quote is a sales document and the telemetry is not.
Four habits from this exit transfer directly to any estate, at any scale.
For the full decision method, including when a renewal genuinely is the better answer, see the ULA decision framework and the broader exit strategy guide.
Balance matters, because certification is not always the answer. A renewal earns its price in three situations.
None of the three applied here. That is what the decision gate exists to establish.
The standard Oracle account team advice is that growing companies should renew the ULA because certification caps your rights at a moment in time. We disagree. In the ULA certifications Fredrik Filipsson ran in 2024 to 2025, most estates had plateaued on Oracle 12 to 24 months before expiry while growth claims in the renewal pitch reflected Oracle's forecast, not the customer's telemetry. The buyer side move is to pull 18 months of deployment data before engaging Oracle at all, then price both paths. A renewal that prices phantom growth is the most expensive insurance a CIO can buy.
Suppose your ULA expires inside the next 18 months. Six recommendations fall straight out of this engagement.
The decision sequence below is the one this retailer ran. It applies to any ULA inside 18 months of expiry.
Run the steps in order. Each one produces the input the next one needs, and skipping ahead to the Oracle conversation is the most common way buyers surrender the sequence's leverage.
White Paper · Oracle
Oracle ULA Exit Strategy Playbook
The buyer side playbook for exiting an Oracle ULA: the certification trap, the support reset, and the timing that protects your renewal leverage. Read it free.
Certifying out means counting deployed processors at the end of the unlimited term and converting them into a fixed perpetual grant at no additional license cost, instead of renewing.
Oracle can challenge the count but cannot refuse a certification run per the contract. This retailer's 41,000 processor count was challenged on VMware grounds and accepted once evidence was presented.
Yes. Properly certified deployments become perpetual licenses you own permanently. The unlimited deployment right ends; the certified entitlement does not.
Support continues on the existing base and certification itself triggers no repricing. The renewal path, by contrast, recalculates support on the new agreement value. Later support reductions must respect license set boundaries.
Deployed Oracle software on VMware clusters is countable when the deployment is real and the contract definitions support it. Counting positions must be evidenced; this is where Oracle pushes back hardest.
Nine to twelve months. The deployment sweep alone took 14 weeks in this engagement, and the count, evidence, and certification letter consumed the rest.
Then a renewal can be rational, but price it against measured growth, not Oracle's forecast. Most estates we measured in 2024 to 2025 had plateaued despite growth claims in the renewal pitch.
No. It is cost avoidance: the difference over three years between paying the $48M renewal and certifying the existing estate at no license cost, net of the exit's advisory and sweep effort.
The ULA decision gates, the certification math, and the audit posture that protects the exit.
Used across more than five hundred enterprise engagements. Independent. Buyer side. Built for procurement leaders running the next renewal cycle.
The most expensive insurance a CIO can buy is a ULA renewal priced on growth that already stopped.
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. Pricing moves, audit signals, and the levers that work. No vendor spin.
