A 500-person company licensing Oracle Java under the 2023 employee metric paid $90,000 a year at list. Under the price list Oracle published on October 8, 2026, the same company pays $10,000. A 100-person company drops from $18,000 to $1,000. Those figures come from House of Brick’s line-by-line comparison of the old and new lists, and they are not a typo.
The cut applies only to organizations with fewer than 3,000 employees. Oracle framed it as a response to “the increasing use of Java among small, medium, and emerging organizations, as AI adoption brings new capabilities,” according to the announcement reproduced by SD Times. Oracle’s own blog post on the new SMB pricing adds online purchasing through the Oracle Store for organizations of 500 or fewer.
That raises a real budgeting question for any mid-size IT shop that started an OpenJDK migration to get away from Oracle Java pricing: does a $10,000 subscription make the migration project a waste of money? For some organizations, yes. For most, the answer depends on headcount, server footprint and three contract terms that got stricter while the price got lower.
What the October 8 list actually charges
The new list replaces per-employee monthly pricing with flat annual bands below 500 employees and a per-head add-on above that:
| Employees | Old annual list (2023) | New annual list (Oct 2026) | Change |
|---|---|---|---|
| 100 | $18,000 | $1,000 | -94% |
| 500 | $90,000 | $10,000 | -89% |
| 1,000 | $144,000 | $83,600 | -42% |
| 2,000 | $288,000 | $230,800 | -20% |
| 2,500 | $360,000 | $304,400 | -15% |
| 3,000 | $378,000 | $378,000 | 0% |
Figures from House of Brick. The bands are $100 for up to 10 employees, $1,000 for up to 100 and $10,000 for up to 500. From 501 to 2,999, the price is $10,000 plus $147.20 per year for each employee above 500. Enterprise pricing at 3,000 and above did not move: it still starts at $10.50 per employee per month.
Look at the shape of that table rather than the headline percentage. Below 500 employees, Java SE went from a meaningful line item to a rounding error. Between 1,000 and 3,000 employees, House of Brick calculates the savings “level off at roughly $55,000 to $60,000 a year.” The $147.20 add-on works out to $12.27 per employee per month, slightly more than the old $12.00 rate for the 1,000 to 2,999 band. Mid-size firms got a fixed discount on the first 500 heads, not a new rate.
Three terms that got stricter
The price went down. The compliance surface went up. Licenseware’s vendor price tracker and House of Brick both flag the same changes in the new list.
The employee count is now continuous. Under the 2023 terms, your licensed count was fixed as of the order date. The new list requires an additional license if your employee count exceeds the licensed quantity “at any time during the subscription term.” The definition of employee did not change: full-time, part-time and temporary staff, plus the agents, contractors, outsourcers and consultants who support internal operations, whether or not they touch Java. A company that licenses at 480 employees and onboards a 40-person outsourced service desk in month seven is out of compliance in month seven.
Bands cannot be stacked, so the cliffs are steep. A 101-person company cannot buy two “up to 100” entitlements; it buys the 500 band. Going from 100 to 101 employees moves the bill from $1,000 to $10,000. That is a planning issue for any fast-growing firm, and it belongs in the hiring forecast the same way a SaaS tier threshold does.
Small bands now carry server caps. The 2023 list had a single 50,000-processor ceiling for everyone. The new list ties processor limits to the band: 1,000 processors for up to 10 employees, 10,000 for up to 100, 20,000 for up to 500, and 50,000 above that. Processors are counted with Oracle’s core factor table, so the 1,000 cap is about 2,000 x86 cores. Licenseware’s warning is the one to take seriously: “The processor limits can push an organization with few employees and many servers into a higher band.” A lean SaaS company with a large Java server estate may not get the cheap band at all.
Two smaller points round this out. SMB pricing is sold direct by Oracle only, not through resellers, and the list now spells out a one-year minimum term. Existing subscribers should expect the new rates at renewal rather than mid-term, which makes your renewal calendar the place to act.
Who should keep migrating, and who can stop
The cut lands while Oracle’s enforcement is escalating, not easing. Pamela Fulmer of Tactical Law Group wrote in April that “the soft outreach is giving way to formal audit notices” issued by Oracle’s renamed license function, GLAS, with letters addressed to a named CIO, CFO or general counsel and a response window “commonly forty-five days” (Legal 500 reprint of the Tactical Law Group article). Azul’s 2026 State of Java survey of more than 2,000 Java professionals found 21% had already been through an Oracle Java audit, and 81% had migrated, were migrating or planned to migrate at least part of their estate to a non-Oracle OpenJDK distribution (IT Brief coverage of the Azul survey). Cost was the most cited reason at 37%. Azul sells an OpenJDK distribution, so read its numbers as directionally right rather than neutral.
A cheaper list price shifts the economics of that migration, and the shift is not uniform. A practical way to sort it:
- Under 500 employees, modest server footprint. The subscription is now $10,000 or less. A migration that needs weeks of engineering time, regression testing and a vendor support contract for the replacement JDK will struggle to beat that over a three-year horizon. Buying the subscription and redirecting the migration budget is defensible, provided your processor count fits the band.
- 500 to 1,000 employees. The cut is real (65% at 750 employees, by House of Brick’s math), but you are now paying roughly $147 per head above 500. Rerun the migration business case with the new number. Many will still favor migration; few will find it urgent.
- 1,000 to 3,000 employees. The saving levels off around $55,000 to $60,000 a year against a bill in the low to mid six figures. The structural problem with the employee metric (paying for every head regardless of Java use) is untouched. The migration case barely moves.
- 3,000 and above. Nothing changed. Neither should your plan.
Two cautions apply at every size. First, a migration that leaves any Oracle JDK running in production keeps the whole employee count in scope, so a half-finished project gets the cost of both options. Second, a lower forward price does not settle historical use. If Oracle’s audit team believes you ran Oracle JDK unlicensed in 2024 and 2025, the October list price is a starting point for that negotiation, not a release from it. As House of Brick puts it, “A low price is not much of a bargain if it comes with an audit finding attached.”
The negotiation lever is the same one that works with any vendor that prices on a metric you don’t control: a credible exit. Oracle cut SMB pricing in the same year it moved to formal audit letters and while four in five Java shops told Azul they were leaving. That is a vendor trying to keep a customer base, which means your migration plan has value in the room even if you never finish it. Treat Java like any other enterprise license agreement: count what you actually run, check it against the license inventory before Oracle does, and bring the alternative’s real cost to the renewal. The same discipline applies to the run of vendor repricing this fall, from Cisco’s November 7 software increase to the metered add-ons now standard in SaaS renewals.