Java priced for the payroll, running on a tenth of the servers
Oracle prices Java for the whole payroll, not for the people who run it: at list, a 1,000 staff employer pays into six figures a year and a 25,000 staff employer into seven, while the work that genuinely needs Oracle Java support usually sits on a small slice of the estate. That gap is where the cost lives and where the defense lives, and across our engagements the per employee subscription was the most expensive answer in most estates.
Prepared by Redress Compliance · August 8, 2026 · Oracle advisory. Based on 30 to 45 Oracle Java engagements reviewed 2024 to 2025.
Executive summary
The footprint finding is the whole argument: 10 to 30 percent of assumed servers.
Once a sweep separated Oracle builds from the OpenJDK builds already installed, the actual Oracle JDK footprint sat on 10 to 30 percent of the servers the buyer assumed, while the subscription meters the entire workforce: roughly $0.14 million a year at 1,000 staff, $0.63 million at 5,000.
And $2.0 million at 25,000, all at list before discount.
Volume tiering softens the curve without flattening it, because the rate falls slower than the headcount rises, and the subscription bundles support priced against everyone rather than the workloads that use it.
The subscription was the costliest answer in 7 of 10 estates, and the sweep proved it.
The buyer side sequence is an estate sweep first, isolating Oracle Java to the workloads that genuinely need Oracle support, migrating the rest to the free certified builds from Temurin, Corretto, Microsoft, and Azul that cover the large majority of enterprise workloads.
And then negotiating a smaller residual: moving the non critical workloads removed 60 to 85 percent of the proposed subscription cost inside a single renewal cycle, with planned migrations running nine to fourteen months.
The count dispute is worth 18 to 28 percent before any rate conversation.
Oracle's opening employee count ran 18 to 28 percent above the count we could defend, usually by sweeping in contractors, dormant accounts, and entire subsidiaries running no Oracle Java at all, and the defense strips them against the definition before the ladder ever applies.
The levers that moved the realized rate below the opening were structural: a documented count, a multi year prepay, and a visible, costed OpenJDK plan, the strongest single lever on the rate, with Oracle's quarter end windows producing the better numbers.
The Software Investment Advisor letter is the soft front door, and a vague reply opens the hard one.
SIA outreach is how the Java review usually begins, and a vague reply invites a formal audit: the response discipline mirrors the audit playbook, one owner, no volunteered data.
And the internal sweep run before any external answer, because the estate that knows its own footprint answers from knowledge.
The trap on the other side is treating a discount as a win, since a large percentage off a payroll sized number can still exceed the cost of supporting the actual footprint several times over.
The cost at real headcounts
| Headcount band | List rate per employee per month | Annual list cost | What usually runs Oracle Java |
|---|---|---|---|
| 1,000 staff | About $12 | About $0.14 million | 10 to 30 percent of servers |
| 5,000 staff | About $10.50 | About $0.63 million | 10 to 30 percent of servers |
| 25,000 staff | About $6.75 | About $2.0 million | 10 to 30 percent of servers |
| 50,000 plus | Negotiated, unpublished | Multiple millions | 10 to 30 percent of servers |
The right hand column is the whole argument.
Oracle meters the entire workforce while the work that needs Oracle Java support sits on a small slice of the estate, and the subscription's bundled support and updates, presented as the value, are priced against everyone rather than the workloads consuming them.
These are planning ranges at list: the point is the shape, not the cent, and the bill tracks the workforce even though only part of it ever touches Java.
The sweep, isolate, migrate, negotiate sequence
- Sweep the estate first: separate Oracle builds from the OpenJDK builds already installed, the step that found the footprint on 10 to 30 percent of assumed servers.
- Isolate the genuine Oracle needs: the Oracle product certified JVMs and the workloads whose vendors require Oracle support, the residual worth paying for.
- Migrate the rest to certified free builds: Temurin, Corretto, the Microsoft Build, and Azul cover the large majority, removing 60 to 85 percent of the proposed cost.
- Negotiate the smaller residual: with the documented count, the multi year prepay, and the visible OpenJDK plan as the levers, timed to Oracle's quarter.
- Defend the count throughout: stripping the contractors, dormant accounts, and non Java subsidiaries that inflated openings 18 to 28 percent.
The Java SE employee licensing brief
The sweep method, the headcount scenarios, the count defense, and the residual negotiation worked end to end.
Get the white paper →The SIA letter, and answering from knowledge
Software Investment Advisor outreach is the soft front door to a Java review, framed as advisory and functioning as discovery, and the reply calibrates what follows: a vague answer invites the formal audit, while the estate that has already run its sweep answers precisely and from strength.
The response discipline mirrors the wider audit playbook, one owner for all contact, nothing volunteered beyond the question asked, and the internal footprint established before any external statement, because the download logs Oracle already holds are exactly what the sweep maps internally.
The full metric mechanics and the NFTC clocks run in the Java licensing pillar, the order document clauses in the Java pricing guide, the build selection in the alternatives comparison, and the audit sequence in the Java audit defense guide.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
What we saw across Java engagements, 2024 to 2025
Across roughly 30 to 45 Oracle Java engagements we reviewed between 2024 and 2025, the per employee subscription was the most expensive answer in most estates:
Of the proposed subscription cost removed by moving non critical workloads, inside one renewal cycle.
Contractors, dormant accounts, and non Java subsidiaries swept into the opening number.
The discount trap deserves the closing word: list is the opening position and Oracle discounts Java like everything else, widening with headcount, term, and a credible alternative.
But a large percentage off a payroll sized number can still exceed the cost of supporting the actual 10 to 30 percent footprint several times over, which is why the leverage was the alternative and never the negotiation skill.
The report's bands update as Oracle revises its schedule, and the constant across every revision is the gap between the metered workforce and the running footprint, the gap the sweep measures and the migration closes.
Your first five moves
- Run the estate sweep before anything else, separating Oracle builds from the OpenJDK already installed.
- Defend the count against the definition, stripping the 18 to 28 percent of contractors and dormant entities.
- Migrate the non critical majority to certified free builds, the 60 to 85 percent inside one cycle.
- Answer SIA outreach precisely, from the sweep, because a vague reply invites the formal audit.
- Negotiate the residual with the OpenJDK plan visible, the strongest single lever on the rate. The Oracle practice runs the sweep with you.
Frequently asked questions
What does Oracle Java cost at real headcounts?
At list: roughly $0.14 million a year at 1,000 staff, $0.63 million at 5,000, and $2.0 million at 25,000, with 50,000 plus negotiated individually, because the meter is the headcount rather than the Java footprint.
The tiering softens the curve without flattening it, and the bill tracks the workforce even though Oracle JDK usually runs on 10 to 30 percent of the servers buyers assume.
Is the Oracle Java subscription worth it?
In 7 of 10 estates we reviewed it was the costliest answer available: the subscription bundles support priced against every employee while the workloads needing Oracle support sit on a small slice of the estate.
The sweep, isolate, migrate, and negotiate sequence removed 60 to 85 percent of the proposed cost inside a single renewal cycle by paying only for the residual that genuinely needs Oracle.
How accurate are Oracle's Java employee counts?
The opening count ran 18 to 28 percent above what we could defend, usually by sweeping in contractors, dormant accounts, and entire subsidiaries running no Oracle Java at all.
The defense strips the count against the price list definition before any rate conversation, because the definition, not the ladder, is where the money is.
What is Oracle Software Investment Advisor outreach?
The soft front door to a Java review: advisory framed contact that functions as discovery, where a vague reply invites the formal audit.
The discipline is one owner for all contact, nothing volunteered, and the internal sweep run before any external answer, so the estate responds from knowledge of its own footprint rather than hope about it.
How much does moving to OpenJDK save?
Moving non critical workloads to free certified builds, Temurin, Corretto, the Microsoft Build, or Azul, removed 60 to 85 percent of the proposed subscription cost inside a single renewal cycle, with planned migrations running nine to fourteen months.
The free builds cover the large majority of enterprise workloads, and the visible, costed migration plan was also the strongest lever on the residual's rate.
Do discounts fix the Oracle Java cost problem?
Rarely: Oracle discounts Java like everything else, widening with headcount, term, and a credible alternative, but a large percentage off a payroll sized number can still exceed the cost of supporting the actual footprint several times over.
The trap is treating the discount as a win, when the structural answer, the sweep and the migration, changes the number the discount applies to.
