Rapid Java Migration to Corretto + Temurin — $3.1M Removed in 16 Weeks

The Situation: A $3.6M Java Renewal That Did Not Need to Be Paid

A mid-cap retail and logistics group with roughly 19,000 employees had been on Oracle Java SE Universal Subscription for two contract cycles. The third-cycle renewal landed at $3.6M for three years on the Employee Metric — a 41% step-up against the prior subscription, driven by Oracle's continued price corrections on the Employee count band. The CIO's question to us was unambiguous: how fast can we exit, and what does that cost. The answer was a 16-week rapid Java migration to Amazon Corretto and Eclipse Temurin, with a controlled bridge subscription on a defensible Employee Metric count covering the migration window. The end-state run-rate Oracle Java spend: zero.

The Java footprint, when we ran the forensic inventory, was concentrated in seven distinct workload classes. Customer-facing e-commerce on Java 17. Inventory and warehouse management on Java 11 LTS. Back-office reporting on Java 8 LTS (the legacy reason for staying on Oracle in the first place — Java 8 commercial support). Two internal mobile apps on Java 11. A data integration platform on Java 17. The customer telephony stack on Java 8. And a small handful of internal-only developer tools on whatever JDK the engineering team had installed locally. Seven workloads, three Java versions, two acceptable OpenJDK migration targets. The technical migration was tractable; the contract negotiation against Oracle's predictable retention playbook was the harder part.

Why "Rapid" Java Migration Is the Right Strategy for Most Enterprises

Oracle's Java SE retention playbook depends on the customer believing that migration is a six-to-eighteen-month exercise. For most enterprises it is not. Amazon Corretto and Eclipse Temurin are production-grade OpenJDK distributions with security-update cadences matched to (and in Corretto's case, integrated with) the major LTS releases. A regression-tested migration of a typical enterprise Java workload runs at 2–4 weeks per workload class with parallel CI runs and a controlled cut-over. The reason most enterprises do not do it is internal — engineering teams do not own the licence spend and procurement teams do not own the build pipeline. Bridging that gap is most of what a rapid Java migration engagement actually delivers.

Our Approach: 16 Weeks, Seven Workload Classes, Two OpenJDK Targets

1. Forensic Java Distribution and Version Inventory

   Every Java runtime across the group's production estate was identified by version, vendor, deployment topology, and security-patch cadence. The audit found two non-production workloads still on unsupported Java versions — a separate compliance gap that would have surfaced in the next Oracle audit cycle. We closed those before the migration started.

2. Target Distribution Selection

   Customer-facing e-commerce, the data integration platform, and the inventory system migrated to Amazon Corretto for AWS-native security update alignment and Corretto's quarterly LTS cadence. Back-office reporting, the mobile apps, the telephony stack and developer tools migrated to Eclipse Temurin for broad ecosystem compatibility and the Adoptium Working Group's vendor-neutral release process. Two workload classes per pilot wave; full regression in CI; controlled blue-green cut-over.

3. Bridge Subscription on a Defensible Employee Metric

   The Oracle Java SE Universal Subscription was bridged at the right-sized Employee Metric count for a single 12-month term to cover residual Oracle JDK production deployments during the migration window. Oracle's opening Employee count was 19,200. The defendable count, once contractors, board members, retail-store cashiers without any Java endpoint and warehouse pickers were excluded, was 4,860. The bridge subscription cost $480K — versus the $1.2M annual run-rate Oracle had quoted on the full Employee Metric.

4. Regression Testing and Cut-Over

   Each workload class was migrated in a controlled wave: build with the new JDK in CI, run the full regression suite against the new build, deploy to a blue-green slice in production, monitor for 7 days, then cut traffic. Two workloads — the telephony stack and the legacy reporting platform — required minor code changes to address Java module-system issues; both were resolved inside the wave timeline. No production incidents attributable to the migration. No customer-impact events.

5. Audit Moratorium and Exit Documentation

   The bridge subscription terms included a 24-month audit moratorium and an explicit Oracle acknowledgement that the migration to Corretto and Temurin was a permitted distribution change under the OTN licence terms. The exit documentation — what was on Oracle Java, what moved when, what is now on Corretto or Temurin — was produced as a board-grade artefact and is the defence record if Oracle's Java audit team ever revisits the historical window.

The Results

By month four the group's entire Java footprint was on Amazon Corretto and Eclipse Temurin. The bridge subscription closed at the end of its 12-month term and was not renewed. Oracle's account team made the usual attempt to revive the conversation in fiscal Q4; the audit-moratorium clause and the forensic exit documentation closed it. The retained run-rate Oracle Java spend at the group: zero.

Key Takeaways for Java SE Universal Subscription Holders