Oracle unveils Java 25
- Oracle releases Java 25, and it eases learning and offers eight years of support.
- A language for both learners and enterprises.
Oracle has released Java 25, the latest long-term support (LTS) version of the world’s most widely used programming language. The update underscores Java’s dual role as both a platform for enterprise systems and a language for new developers, while also extending its reach into fast-changing areas like AI and post-quantum security.
Java 25 introduces changes designed to make the language easier for newcomers while expanding its capabilities for advanced users. JEP 507, for instance, extends pattern matching to primitive types, making code more expressive and useful in AI inferencing. JEP 511 simplifies imports by allowing developers to bring in all packages exported by a module, a feature that lowers friction for those still learning how modular libraries work.
Beginners also benefit from JEP 512, which trims boilerplate code so small programs can be written without advanced structures, and JEP 513, which improves constructor flexibility by enabling input validation before a constructor is fully executed.
Libraries also see upgrades—structured concurrency, refined under JEP 505, helps manage multithreaded applications by grouping related tasks into a single unit of work. JEP 506 introduces scoped values for sharing immutable data across threads, while JEP 502 adds stable values that behave like constants but with more control over initialisation. The Vector API, updated under JEP 508, boosts performance by compiling vector computations into optimised CPU instructions at runtime.
Security and performance
Security enhancements reflect growing concerns about encryption and resilience. JEP 470 adds support for the PEM format, allowing cryptographic keys and certificates to be encoded and decoded more easily. JEP 510 introduces a key derivation function API, laying the groundwork for hybrid public key encryption and post-quantum standards.
On the performance side, JEP 519 reduces object header sizes to save memory and improve efficiency. JEPs 514 and 515 improve startup times and code generation, shifting profiling into training runs so applications can deploy optimised code from the start.
Oracle has also extended the JDK Flight Recorder with profiling and tracing features. JEP 509 improves CPU-time profiling on Linux, JEP 518 enhances thread sampling reliability, and JEP 520 adds method timing and tracing, helping developers pinpoint bottlenecks and debug issues more effectively.
Making Java easier to learn
“The idea is to smooth the path for beginners without creating a ‘Java Lite’ that would later require unlearning,” Arimura said. Features like compact source files reflect this principle by giving students a way to start coding quickly while still learning the full language.
He pointed to Learn.java, a new resource for students who may not yet call themselves developers. It complements dev.java, which is targeted at professionals looking to advance their skills. “Once you’re a professional developer, we have great material on dev.java to help you improve and master your craft. But if you’re a student… you need a different path. That’s what Learn.java is for,” Arimura explained.
Why long-term support matters
“We always start with eight years, but we fully expect it will be extended,” Smith said. Java 8, for example, was extended to 2030, while Java 11 was extended to 2032 to meet customer needs for decade-long projects. “It comes down to which versions have strong adoption and are tied to long-term implementations,” he added.
Oracle prepares Java for AI and post-quantum cryptography
With AI and security at the core of Java 25, Oracle executives emphasised the importance of balancing stability with innovation. Smith noted that lessons from the adoption of TLS 1.3 in Java 8 are shaping the approach to post-quantum cryptography. “The last thing you want is to provide an update with advanced cryptography but force customers to choose between stability and security,” he said.
Bernard Traversat, vice president of software development at Oracle, highlighted the performance challenges. “TLS performance is critical for most enterprise and other Java applications, so as we transition to post-quantum, we need to ensure customers don’t see major performance drops,” he said.
Traversat also pointed to ongoing work to prepare Java for AI workloads. Projects such as Valhalla, which reduces memory pressure, and Loom, which brings virtual threads for handling large numbers of interactions, are being developed with agent-based systems and AI inference in mind. Project Leyden, which accelerates startup times, is expected to be especially important for AI agents and LLM-driven interactions.
Looking ahead: Oracle’s Java roadmap
He added that Oracle sees growing opportunities in generative AI and is ensuring Java’s ecosystem can take advantage of new tooling while remaining energy-efficient at scale. Projects like Babylon aim to expand Java’s role in heterogeneous computing, supporting both enterprise efficiency and next-generation applications.
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology eventsclick here for more information.
AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
TNG – Latest News & Reviews

