The development of AI is producing multiple winners and losers among programming languages. The best-known languages in datasets regularly produce better, more consistent AI-generated code. This is one of several explanations for the continued growth of Java. This is despite problems surrounding Oracle’s revenue model for the programming language, which seems to be driving the popularity of Java-compatible alternatives.
Python was ahead of other programming languages: as the most popular language before the AI hype, it has remained at a lofty height. Java specialist Azul acknowledges Python’s central role in the AI domain, but sees that Java has claimed its place just as well. While AI research and prototyping takes place on a Python basis, according to Azul, Java is responsible for the AI-driven services in production. The researchers see this choice as stemming from requirements around performance, security, and reliability.
In its 2026 State of Java Survey, figures reflect what Azul means by this. Sixty-two percent of the more than 2,000 Java professionals surveyed use Java for AI development, significantly more than the 50 percent who indicated this last year. The exact variant does differ, however. It is more accurate to say that the Java Virtual Machine dominates production AI, not the Java language itself. Scala (47 percent), Groovy (43 percent), Kotlin (39 percent), and Clojure (28 percent) are all used, while Java itself (19 percent) has become a niche.
Oracle exodus
Java, first launched in 1995, was originally a Sun Microsystems initiative. Java’s “write once, run anywhere” promise made it an immensely popular language. Oracle, owner of former Sun components since 2010, decided in 2023 to price Java licenses per employee per month, rather than per user or processor. It soon became clear that the costs would be substantial. Version 21 of the Java Development Kit (JDK), the most popular form of Java, will lose free support in September of this year. This seems to be a reason for organizations to quickly change their development kit. No less than 81 percent have already migrated, are in the process of migrating, or are planning to do so. The move is then to a non-Oracle-based OpenJDK distribution.
The new license costs are offset by faster LTS timelines and simpler upgrade cycles. Oracle’s decision therefore also has advantages, besides the fact that the changed license is not necessarily more expensive. Nevertheless, 92 percent are concerned about Oracle’s pricing model. As a result, the variety in the Java landscape is flourishing, which means that in surveys of popular programming languages, we must take into account that the JVM may be used many times more often than Java adoption itself suggests.
It is striking that 84 percent of organizations that have already made the switch indicate that the migration went more smoothly than expected or according to plan. 72 percent completed the switch within a year. This shows that Oracle’s grip on the Java ecosystem is loosening, despite the fact that the technology behind the language is only gaining in popularity.
Read also: “Oracle considering drastic steps to finance AI spending”