A recent survey reveals that almost two-thirds of Java developers leverage the language for artificial intelligence projects, with prominent libraries such as JavaML, Deep Java Library, and OpenCL leading the pack.
Java’s footprint in AI application development is rapidly expanding, as evidenced by Azul’s latest 2026 State of Java Survey and Report. The study shows a significant leap to 62% of respondents utilizing Java for AI, up from 50% in the previous year’s findings.
Unveiled on February 10, this comprehensive report compiles insights from over 2,000 Java users surveyed between September and November 2025. A striking discovery was that 81% of participants have either transitioned, are in the process of transitioning, or intend to transition from Oracle’s Java to an OpenJDK distribution not managed by Oracle, largely driven by 92% of respondents voicing apprehension over Oracle Java’s licensing costs.
The survey underscored a distinct movement towards integrating AI capabilities into the enterprise systems that are fundamentally built on Java. The report highlights the rich ecosystem of AI libraries available to Java developers for implementing AI features, with JavaML emerging as the most frequently used, closely followed by Deep Java Library (DJL) and OpenCL. Notably, nearly a third (31%) of developers indicated that over half of their codebase incorporates AI functionalities.
Regarding AI-driven code generation tools employed for crafting Java application code, OpenAI’s ChatGPT emerged as the frontrunner. It was succeeded in popularity by Google Gemini Code Assist, Microsoft Visual Studio IntelliCode, and GitHub Copilot.
In other key findings from the report:
- 18% have already embraced Java Development Kit (JDK) 25, the latest Long Term Support (LTS) version, released in September 2025.
- 64% reported that Java or a JVM powers over half of their existing workloads or applications, a slight decrease from 68% recorded last year.
- For 43%, Java-based workloads constitute over half of their total cloud computing expenditures.
- A significant 63% acknowledged that the presence of defunct or unutilized code negatively impacts devops productivity, ranging from a moderate to a substantial degree.
