Oracle prices Java against your total workforce, not your Java usage, and the definition reaches into your suppliers' payrolls. The populations that count, the exclusions that survive challenge, and how to build a number Oracle cannot rewrite.
Oracle's opening employee number almost never comes from you. It comes from a filing, a careers page, or a figure somebody mentioned on a call, and in our engagements it runs 18 to 28 percent above the number the buyer can actually defend.
It is a license quantity equal to your entire workforce, including specified supplier staff, regardless of how many people use Java. On 23 January 2023 Oracle withdrew Named User Plus and Processor licensing for Java SE and replaced both with this single metric.
The mechanic is deliberately simple. Count the people, drop the number on Oracle's published band ladder, multiply by twelve months, and that is the annual list price.
Under the retired model a company with 50 Java users and a handful of server processors paid for 50 users and those processors. Under the employee metric the same company with 5,000 staff pays for 5,000 people.
The bill is now welded to who you employ rather than to what you deploy. That is not an accident of drafting, and no amount of technical evidence about your deployment will change the arithmetic.
Our companion analysis, Only 50 developers use Java, so why are you licensing 10,000 employees, walks the gap between usage and headcount. This page decodes the count itself.
It says the count is all of your own full time, part time and temporary employees, plus the full time, part time and temporary employees of your agents, contractors, outsourcers and consultants that support your internal business operations. It adds that the quantity is set by that number, not by how many people use the programs.
Read the second half twice. The definition reaches past your payroll and into the payrolls of your suppliers, which is why two companies with identical Java estates and identical staff numbers can face very different bills.
The wording is published by Oracle in the Java SE Universal Subscription price list on the Oracle price list index and summarized on the Java SE Universal Subscription product page. Read the version attached to your own order, not a summary, including this one.
Oracle's default negotiating position treats every supplier person with system access as countable. The contract language is narrower than that position.
The words "that support your internal business operations" limit the second half of the definition. They do not limit the first half, and they do not appear anywhere near the words "use the programs".
The practical test we apply is whether the supplier is delivering an outcome or supplying people into your operating model. A firm engaged to deliver a fixed scope project is arguably not supporting internal operations in the sense the clause describes. A body shop supplying twenty developers into your platform team plainly is.
Contractor counting is the single most common live dispute we see, and it is treated at length in Do contractors and consultants count toward your Java employee total?.
Oracle amended the model language twice in early 2023 and the second amendment is easy to miss. The January wording said the subscription was not eligible for application specific full use, embedded, ISV licensing or redistribution.
The March wording replaced that list of exclusions with a positive statement that the subscription is available only for internal business operations under an applicable enterprise license. The shift tightens the scope and reinforces the same internal operations language that governs the contractor count.
It is not a footnote. It is the drafting Oracle will point at when it argues which supplier staff count. Check the effective wording printed on your specific order.
Oracle publishes a scenario in which a company has 28,000 employees, made up of 23,000 own staff plus 5,000 agents, contractors and consultants, and lands in the 20,000 to 29,999 band at $6.75 per employee per month. That is $2,268,000 a year.
The example exists to establish one thing in writing. Oracle folds supplier staff into the count in its own published material, so arguing that contractors are out of scope entirely is not a viable opening position.
The viable position is narrower and stronger. Argue about which supplier staff meet the qualifier, and make Oracle name individuals rather than assert supplier headcounts.
Every person on your payroll, plus the named supplier staff who work inside your operating model, and nobody else. The table below is the population map we build at the start of every count engagement.
Print it, walk it line by line with human resources, procurement and the outsourcing owners, and write down a decision and a reason for each row. That document is the count.
| Population | Counts? | Why | Buyer action |
|---|---|---|---|
| Full time employees on payroll | Yes | First half of the definition, uncontestable | Take the number from the HRIS on the order date, dated and exported |
| Part time employees | Yes, as whole people | The definition names them and contains no proration | Do not convert to full time equivalents. It will not match your own evidence |
| Temporary and seasonal staff | Yes, if engaged on the count date | Explicitly named in the definition | Time the order away from your seasonal peak. This is legitimate and material |
| Interns, apprentices, graduate trainees | Yes, if on payroll | They are temporary employees | Include them and say so, so the number reconciles |
| Staff on parental, medical or sabbatical leave | Yes | Still employed on the count date | Include. Excluding them creates an inconsistency Oracle will find |
| Board members and non executive directors | Usually no | Officers rather than employees, and not part of internal operations | Exclude with a one line written rationale in the count memo |
| Retirees and leavers with live accounts | No | Not employed | Close the accounts before anybody counts them. Stale identities inflate every number |
| Independent contractors engaged directly | Yes, where they support internal operations | Second half of the definition | Extract from the vendor master and accounts payable, not from HR |
| Consultancy staff on a time and materials basis | Yes, for the named individuals assigned to you | The definition reaches the supplier's employees, not the supplier | Count assigned individuals. Never accept a supplier total headcount |
| Fixed price, deliverable based supplier staff | Arguable, often excludable | They deliver an outcome rather than supporting internal operations | The most winnable exclusion. Document the contracting model in writing |
| Shared service and BPO agents serving many clients | Arguable and material | One agent supporting forty clients is not a whole employee of each | Argue for the dedicated, named population only. Get it from the supplier in writing |
| Dedicated offshore development center staff | Yes | Dedicated and inside your operating model | Include. They are already visible in your access and badge systems |
| Managed service provider staff running a function for you | Yes, for dedicated staff | Outsourcers are named in the definition | Ask the provider for a dedicated headcount attestation at the order date |
| Franchisee owners and their staff | Usually no | Independent businesses, neither your employees nor your agents' employees | Strong exclusion. Expect Oracle to push. Hold the line with the franchise agreement |
| Joint venture staff where the venture is a separate entity | Depends on the agreement | The obligation follows the contracting party and its defined affiliates | Read the affiliate definition before conceding a single person |
| Employees of an acquisition that has not closed | No | They are not yours on the count date | Plan the order date around the closing calendar. Do not volunteer the pipeline |
| Employees of a business you have divested | No, from completion | Same principle in reverse | Make the reduction visible and evidenced at the next count |
| Your customers and end users | No | They do not support your internal business operations | Never volunteer a user number. It invites confusion with the employee number |
Not every row on that table is worth an argument. Three of them carry enough value to justify the effort and enough contractual support to survive it.
Everything else on the table is either clearly in or worth conceding quickly to buy credibility on the three that matter.
Source: Redress Compliance advisory engagement file, 2024 to 2025.
Every argument above operates inside a boundary that most buyers never examine. The count applies to the contracting party, and what the contracting party includes is set by the agreement, not by the price list.
If your Oracle Master Agreement defines the customer to include affiliates and subsidiaries, the count is group wide and every argument about contractors is a rounding error. If the order is placed by one legal entity with no affiliate extension, the count is that entity's.
Check three things before you concede anybody. Which legal entity is named on the ordering document, how the governing agreement defines affiliate, and whether any prior amendment extended the definition across the group.
Two levers, two orders of magnitude
| Lever | Typical movement in the count | Where the evidence comes from |
|---|---|---|
| Contractor and supplier scope | 10 to 25 percent | Statements of work, vendor master, supplier attestations |
| Entity and affiliate scope | Nothing, or the majority of the count | The ordering document and the governing agreement |
| Count date timing | 2 to 15 percent in seasonal businesses | Twelve months of HRIS headcount history |
| Stale identity cleanup | 1 to 6 percent | Directory and joiners, movers, leavers reconciliation |
Buyers spend most of their negotiating energy on the first row. The second row is where the money is.
The quantity is fixed as of the order date and stays fixed for the term unless the order says otherwise. That single sentence has two consequences that are worth real money.
The first is that the date you place the order is a variable you control. The second is that growth during the term is a contractual question, not an automatic one.
None of this works retrospectively. It works when the count date is treated as a planned event with a named owner, six to nine months before the renewal.
Oracle's standard expectation is that material growth is trued up, and the mechanism will be whatever your order says. Silence is not protection. It is an invitation to a conversation you will have from a weak position.
Ask for three things in the order. A stated growth threshold below which no true up applies, a price hold on the rate applied to any true up, and a symmetrical reduction right if the workforce shrinks.
The third is the one Oracle resists and the one that matters most in a downturn. The negotiation levers are set out in Oracle Java employee count negotiation levers.
With a dated, reproducible count built from systems of record and owned by finance or human resources, not by IT. The test is simple: two people following your written method a month apart should reach the same number.
Almost nobody passes that test the first time, which is why Oracle's number so often becomes the number.
Artifact six is the one that wins arguments. Artifact four is the one that takes longest, so start it first.
Oracle's opening number almost always comes from a public source. An annual report, a filing, or a professional network page, and each of those measures something different from the contractual definition.
Annual report figures are frequently full time equivalents averaged over a year, which understates whole person headcount while excluding contractors entirely. Professional network counts include former staff and people who never worked for you.
Do not argue that the public number is wrong. Build the bridge from it to yours, line by line, and hand over the bridge rather than the underlying data.
The common advice is to fight the contractor clause. We disagree, and the disagreement is worth real money. Contractor scope is arguable and in our engagements it moves the count by ten to twenty five percent. Entity scope moves it by everything or nothing. If the agreement binds only the entity named on the order, a global group may be licensing one division rather than the whole enterprise, and no contractor argument comes close to that. Buyers open with contractors because it feels like a compliance discussion. Open with the entity boundary instead, because it is a contract discussion, and those are the ones that change the quantity on the order.
There is a second reason to lead with entity scope. Arguing about contractors requires you to disclose your supplier landscape, which hands Oracle a map of your operating model. Arguing about entity scope requires nothing but the documents Oracle already has.
Your number lands in one of eight published bands, and the band sets the rate applied to every person you counted. The ladder below is Oracle's published list pricing, meaning before any discount.
The full arithmetic, including the band edge effect where ordering more people than you employ is cheaper, is worked in Oracle Java employee tier pricing with worked examples.
Published Java SE Universal Subscription bands
| Employee band | List per employee per month | Annual list at the top of the band |
|---|---|---|
| 1 to 999 | $15.00 | $179,820 |
| 1,000 to 2,999 | $12.00 | $431,856 |
| 3,000 to 9,999 | $10.50 | $1,259,874 |
| 10,000 to 19,999 | $8.25 | $1,979,901 |
| 20,000 to 29,999 | $6.75 | $2,429,919 |
| 30,000 to 39,999 | $5.70 | $2,735,932 |
| 40,000 to 49,999 | $5.25 | $3,149,937 |
| 50,000 and above | Quoted by Oracle | Negotiated, and it can fall below $5.25 |
List prices change. Confirm the current ladder on Oracle's published price list before you model anything.
Java on as many machines as you like, up to a stated ceiling of 50,000 processors, with desktops and laptops excluded from that ceiling. Above it you need additional licenses from Oracle.
For most buyers the cap never binds, because desktops and laptops sit outside it and few estates run Java on that many server processors.
Model it anyway if you operate a large virtualized estate or a substantial cloud footprint. Oracle's processor counting rules in virtual environments are unfriendly to buyers, and a breach converts an unlimited deployment right into a metered one at exactly the point where you have least leverage.
Two mechanics do most of the damage: the annual escalation and the minimum annual floor. Both are easy to miss and both are close to impossible to remove once signed.
Oracle's renewal uplift on Java subscriptions has been running at up to 8 percent a year, compounding, and it applies automatically unless you negotiate a hold. Confirm the figure on your own ordering document rather than on any published summary.
On a subscription of $1.62 million, an 8 percent compounding uplift turns a three year commitment into roughly $5.26 million against the $4.86 million a flat price would imply. Over five years the gap widens sharply.
If you sign nothing to cap it, you have agreed to it. The forward modelling sits in the Java bill increase forecast.
In roughly one in three order documents we have reviewed, Oracle sets a minimum annual floor between $50,000 and $100,000. The floor prevents the bill from falling even if your count does.
That matters most for smaller buyers and for any business expecting a workforce contraction. Negotiate it down or out before signature, because afterwards it is a ratchet.
Oracle's published position is that customers of the legacy Java SE Subscription and Java SE Desktop Subscription may renew under existing terms, to the extent their existing order permits, subject to confirmation that usage still reflects the licensed counts.
That is the escape hatch for anyone who licensed a subset of their estate under the retired per user or per processor model. Keep the legacy metric and you keep the cheaper arithmetic.
Two warnings. A published FAQ position is not contractually binding, and any scope change, count increase, co term or added product can be treated as the trigger to convert you.
The disciplined renewal path is mapped in how to renew the Oracle Java legacy metric, and the underlying validity question is answered in are your pre 2023 perpetual and NUP Java licenses still valid.
Far more than the two to five times figure that circulated when the change landed, unless you were already licensed broadly under the old model. The multiple depends entirely on how much Java you were actually licensing.
The retired Java SE Subscription list prices were $2.50 per named user per month for desktop use and $25.00 per processor per month for server use, both with volume tiers that reduced them further.
Work a real case. A company with 50 Java users and ten x86 cores, counted as five Oracle processors after the core factor, paid roughly $3,000 a year at list. The same company with 5,000 employees now pays $630,000 a year at list.
That is not two to five times. For a usage light, headcount heavy organization it is two orders of magnitude, and it is why the metric change, not the price rise, is the story.
A retailer with one hundred Java servers and forty thousand store associates now pays as if every associate were a Java developer. Nothing in the estate changed. The sentence that defines the quantity did.
The two to five times range describes organizations that were already licensing thousands of named users. For everyone else it understates the impact by a wide margin, which is why so many 2023 budget forecasts were wrong.
Three, and they recur in almost every engagement. None of them is about Java technology.
Oracle asserts a supplier population, usually derived from your public presence or from a number somebody mentioned on a call. You respond with the contracting model for each supplier and a dedicated headcount attestation.
The resolution is almost always a negotiated population rather than a principled one. Go in with the attestations already collected, because the party holding evidence sets the anchor.
Oracle prefers the highest recent number. You are entitled to the number as at the order date, evidenced. If the request arrives inside a formal audit, sequence your reply using the GLAS formal notice response guide before you send any count at all.
Disputes here are won with twelve months of dated HRIS history showing the shape of the workforce, not with a single export that could have been produced on any day.
Oracle will treat almost any change to a legacy order as a conversion trigger. The defense is procedural rather than technical.
Renew like for like, change nothing, add nothing, co term nothing, and put the renewal instruction in writing well before the anniversary.
It means the decision is commercial rather than technical, because nothing you do to the estate reduces the bill while you remain on the metric. Only removing Oracle binaries removes the charge.
That reframes the business case. The comparison is not Oracle support against community support. It is a headcount linked annual fee that grows with the company against a one off engineering program.
No, it counts your entire workforce regardless of Java use. Oracle's definition states expressly that the quantity is determined by the employee number rather than by the number of people who use the programs. A company with 5,000 staff and 40 Java servers licenses 5,000 employees.
Yes, where they support your internal business operations, and Oracle's own published example folds 5,000 contractors into a 28,000 person count. The argument worth having is which supplier staff meet that qualifier, not whether supplier staff can ever count. Outcome based suppliers and shared service agents are the defensible exclusions.
Yes, the definition names full time, part time and temporary employees with no proration anywhere in the wording. Converting to full time equivalents produces a number you cannot reconcile to your own payroll evidence. Count whole people and manage the exposure through the count date instead.
Usually no, because a franchisee is an independent business that pays you rather than an agent or contractor supporting your internal operations. Expect Oracle to assert otherwise in retail, hospitality and automotive estates. Hold the position with the franchise agreement itself and be prepared to explain the commercial relationship.
They do not count until the transaction completes, because the quantity is fixed as of the order date. That makes the sequencing of an order against a closing calendar a legitimate planning question. Do not volunteer your acquisition pipeline in a licensing conversation.
As of the order date, and it holds for the term unless your order says otherwise. Growth during the term is handled by whatever true up language you agreed, so silence favours Oracle. Ask for a growth threshold, a price hold on any true up, and a symmetrical reduction right.
Yes, in principle, where your existing order permits renewal on existing terms and your usage still matches the licensed counts. The risk is procedural rather than legal, because almost any change to the order can be treated as a conversion trigger. Renew like for like and put the instruction in writing early.
By producing a better number first, with a written method and dated evidence behind it. Public figures are often averaged full time equivalents and exclude contractors, so they measure something different from the contractual definition. Build a reconciliation from the public number to yours and share the bridge rather than the source data.
Oracle prices Fusion ERP Cloud per employee, not per user, which inflates true cost. The buyer side guide to module economics and the modernization discount.
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.
