Why Oracle Java Still Prices Your Headcount, Not Installs
Ninety-two percent of more than 2,000 Java professionals are concerned about Oracle Java pricing. Only 7 percent say they are not concerned at all (nearly half the “not concerned” share a year earlier). As of January 23, 2023 Oracle moved Java SE to a per-employee metric. Eighty-one percent of those respondents have migrated, are migrating, or plan to migrate some or all of their Oracle Java to a non-Oracle OpenJDK distribution. Twenty-one percent say their organization has already been through an Oracle Java audit.
That’s the problem.
The parent pack is the Oracle Java SE subscription cost page: published bands, Employee definition, and stay versus OpenJDK.
For the audit angle, see Oracle Java audits count headcount, not installs.
The figures come from Azul’s 2026 State of Java Survey, fielded by Dimensional Research among 2,039 qualified respondents. Treat them as directional. Azul sells Java support. This is not a market census. What the survey still tells you is that three years after the metric changed, concern is still climbing, and a lot of organizations are still hearing the commercial model for the first time from a salesperson.
Oracle Java does not ask how many copies you run
It asks how many people you employ.
Oracle’s own price list is the primary source, not a survey. “Employee for Java SE Universal Subscription” is defined as all of your full-time, part-time, and temporary employees, plus the employees of agents, contractors, outsourcers, and consultants that support your internal business operations. The quantity required “is determined by the number of Employees and not just the actual number of employees that use the Programs.” Pricing starts at $15 per employee per month on the published card, with volume tiers down to $5.25.
A thin Oracle Java footprint and a large employee base is a large position. Your inventory export cannot tell you which you are until someone recasts it against the metric Oracle actually sells.
If your reports still treat Java as an install count, you are reading a reporting layer. The vendor will not walk in with your install count. The vendor will walk in with your headcount.
This is not a criticism of the teams building those reports. Traditional SAM tools collect comprehensively and answer slowly. They were built to prove what is deployed, not to recast that deployment against a metric that ignores volume. The gap is interpretation work. An analysis that arrives after the salesperson has framed the meeting is a document. One that arrives during the meeting is a decision input.
Most people still hear it from a salesperson
Azul’s own commentary on the 2026 numbers is operational, even if you discount the commercial interest. Concern keeps rising because awareness is still filtering through the user community. For most organizations, the way they find out is a conversation with an Oracle salesperson. That conversation starts somewhere in the organization. It takes a while to reach the people who own the financial implications and the budget.
Picture the room. That conversation has already started (just not with the people who own the budget). Your CIO wants to know whether you are exposed, whether a migration is already underway, and what it costs if you stay. If that answer takes three weeks of specialist reading, you have already lost the framing of the meeting.
Azul’s 2026 survey puts current audit experience at 21 percent of respondents, with 26 percent naming audit risk as a reason to migrate. A prior Azul-commissioned wave put three-year audit incidence much higher. You do not need either figure to be a census. You need them as a reminder that Oracle License Management Services documents an Audit Service that compares usage to entitlements, and that the metric on the price list is Employees, not the one in your inventory export.
If that pattern feels familiar, the problem is not your team’s maturity. A per-employee conversation is being answered with a deployment report.
A migration plan is not a license position
The same 2026 survey says 81 percent have migrated, are migrating, or plan to migrate some or all Oracle Java to an OpenJDK alternative. Sixty-three percent intend to migrate the entire estate. Cost is the top driver at 37 percent, then open source, licensing uncertainty, and audit risk.
That is a useful directional signal. It is not a license position. Stay-versus-move on that remainder is a renewal-or-replacement decision, not a migration slide.
“Some or all” is doing a lot of work. A plan to migrate the estate does not take the last Oracle Java install off the employee metric. A completed migration of the applications you know about does not cover the JDK that landed on a laptop, a build agent, or a vendor appliance. Oracle’s FAQ is blunt on architecture: “The underlying application architecture and deployment model does not affect pricing.” If you cannot say, this week, where Oracle Java still runs, what employee number the metric would use, and what you will do about the remainder, you are walking in with a project plan. Oracle is walking in with a subscription.
Do not let the vendor frame the meeting as “are you still on Oracle Java.” The question is narrower: is there still one copy that puts the employee base on the hook, and what number do you accept if there is.
The number to walk into the next Java conversation with
Do this in the next thirty days, as a decision pack rather than a project.
| What you produce | Why it changes the meeting |
|---|---|
| Current Oracle Java footprint, this week | Distinguishes “we have a migration” from “one copy remains.” Oracle’s metric does not care how few people touch Java. |
| Employee count Oracle would use | Full-time, part-time, temporary, plus supporting contractors and outsourcers. Not named users. Not processors. |
| A stay-versus-move cost on that headcount | Published list starts at $15 per employee per month. Price the employee metric against your estate, not last year’s processor or NUP number. |
| A remainder plan, not a roadmap slide | 81 percent in the survey are somewhere on a migration. “Plan to migrate” is not an answer if one Oracle Java instance is still enough. |
Can a non-specialist brief the CIO in fifteen minutes from the pack you have today? If the answer is no, you do not have a Java position. You have a library.
Treat the survey percentages as a prompt, not as your estate. Azul commissioned the work. Your number is the one you can defend against your own inventory, your own HR file, and the contract in front of you.
The decision layer, not another inventory
You already have discovery data and an employee count. The gap is not another inventory. The gap is turning that estate into a decision: where Oracle Java still runs, what employee number the metric would use, and what should happen before the next conversation. LICENSEWARE sits on the inventory and ITSM tools you already run. It is not a rip-and-replace SAM suite. An Oracle/Java ELP is the artifact: a current, defensible position you can walk in with, not another report that still needs a specialist to translate. Java versions and editions have to be recast against the employee metric Oracle actually sells, not left as an install list.
If you are heading into an Oracle Java conversation or an audit and your current tools still need three weeks to turn an install list into a headcount position, book an Audit Readiness Review. You can also start on the free plan and run the analysis on your own data.
The question to walk in with
Do not let Oracle frame this as “are you still using our Java.” The right question is: is there still one copy that puts the employee base on the hook, what is that employee number, and what does stay-versus-move cost against this estate. That is a data question, not a sales question.
The 2026 survey is a temperature check from an interested party. Use it that way. The commercial model has not changed since 2023. The vendor will walk in with a number. The only question is whether you have yours first: current, defensible, and tied to the contract in front of you.
FAQ
Does one Oracle Java install put the whole company on the hook? Oracle’s price list says the quantity is determined by Employees, not by the people who use the Programs. Defend your contract and your use, not the survey.
Is the 92 percent figure a market census? No. Dimensional Research surveyed 2,039 Java professionals in work commissioned by Azul. Treat the percentages as directional.
Does a planned OpenJDK migration close the exposure? Not by itself. The survey’s 81 percent includes people who have migrated, are migrating, or only plan to. A partial move leaves the employee metric in play if one Oracle Java instance remains.
What number should SAM take into the next Oracle Java meeting? Where Oracle Java still runs, the employee count the metric would use, and a stay-versus-move cost on that headcount. Not the install export.
