Java in 2025: LTS Release, AI on JVM, Framework Modernization
Transcript:
2025 was not a quiet year for Java. The Java platform and the ecosystem around it continued to evolve rapidly, so let’s look at the major updates of the year.
Java turned 30 this year, which in tech terms usually means becoming a legacy fossil. But that did not happen. Java keeps adapting, modernizing, and still runs a large portion of the world’s critical infrastructure.
Java 25 is the new long-term support release. It brings numerous updates and 18 JEPs with new, improved, or revised functionality. LTS releases are the ones teams adopt and run in production for years, so this is a release worth checking out if you haven’t already. Among the important features are continuous improvements to Java Flight Recorder with better visibility and lower overhead, scoped values that move Java toward safer and more reliable concurrency with fewer thread-local risks, and the ahead-of-time cache from Project Leyden that improves startup and warm-up times. This is especially important for cloud-native and serverless environments. Compact object headers also reduce the memory footprint of Java applications. If you want a deep dive into every single JEP, check out my dedicated video on Java 25 features.
AI on the JVM kept gaining momentum in 2025. This year it crossed the line from an interesting experiment to something actually usable in production. The first production-ready version of Spring AI reached general availability in 2025. It provides a familiar programming model for integrating LLMs, vector stores, and AI workflows into Spring applications. LangChain4j was also very active this year and reached its first general availability release, bringing structured AI workflows to Java with a focus on strong type safety and JVM-friendly abstractions. Embeddable is a new project from Rod Johnson, the founder of Spring, aimed at building advanced AI agents on the JVM. It demonstrates that serious AI orchestration on the JVM is absolutely possible.
Kotlin 2.2 continued refining the language, tooling, and compiler performance. It also improved multiplatform support across web, server, and mobile. Kotlin’s popularity keeps growing, and it now stands shoulder-to-shoulder with Java in backend development. Kotlin also owns the multiplatform story across platforms. According to general feedback, KotlinConf 2025 felt very backend-focused, with a lot of attention on performance tooling and real-world use cases.
The Spring Framework team was also very busy this year. There were numerous releases across the Spring ecosystem, with Spring Boot 4 and Spring Framework 7 standing out. These releases modernized the ecosystem with cleaner APIs, better performance, and closer alignment with newer Java versions.
Quarkus released a new LTS version, 3.27, reinforcing its position as a serious choice for cloud-native Java applications with fast startup times, low memory usage, and aggressive optimizations. Apache Flink 2 marked a major shift in real-time data processing, introducing new asynchronous models and improved APIs that make streaming systems more powerful and expressive.
Vaadin merged Hilla into Flow, simplifying its ecosystem and reducing fragmentation. Vaadin 25 focused on being slimmer, faster, and more modern. The Micronaut framework continued refining its ecosystem around HTTP, messaging, security, and observability, with better OpenTelemetry integration and support for modern Java and Kotlin versions. Ktor kept gaining traction as a lightweight, Kotlin-first framework, continuing to improve its HTTP engine, coroutine-based concurrency model, and plugin system.
IntelliJ IDEA added support for Java 25 and Spring Framework 7, keeping pace with the ecosystem and making upgrades easier. JetBrains also introduced a new unified IntelliJ release strategy, simplifying versioning and making it easier to stay up to date. The Spring Debugger saw significant improvements in 2025, providing better insights into Spring applications at runtime without additional libraries or agents. JetBrains Research introduced LitmusKT, a testing tool for Kotlin multiplatform concurrency that helps uncover subtle concurrency issues that are otherwise very hard to reproduce.
Gradle 9 landed with a strong focus on performance and reliability. Builds are faster, more predictable, and better suited for large, complex projects. The configuration cache is now the preferred path in Gradle, delivering major speedups that are especially important for CI pipelines. OpenTelemetry continued evolving its stabilization and release practices, increasing confidence that today’s instrumentation will not break tomorrow.
BellSoft introduced hardened container images that are a strong fit for enterprise Java applications. These minimized images are based on Liberica JDK and Alpaquita Linux. They provide an immutable component set, up to 95% fewer vulnerabilities, and a single SLA from one vendor that covers the OS, runtime, and vulnerability management.
These were the major updates in and around the Java ecosystem in 2025. Of course, this is only the tip of the iceberg. There were many more releases across tooling, frameworks, and platforms. It was a busy year for Java, and we are looking forward to another year full of progress and innovation. If you enjoyed this video, don’t forget to like it, subscribe to the channel, and until next time.
