This article is the first in "The Future of Java" series.

Enterprise Java has been defined by one principle since its inception: standards matter. The success of Java in the enterprise wasn’t an accident of timing, it was the deliberate result of building a vendor-neutral foundation that developers could rely on for decades. Servlets, EJBs, JPA, JMS, and later Jakarta EE all codified proven practices into specifications that every vendor had to implement consistently. Enterprises could invest with confidence because they knew that they were not locking themselves into one provider’s interpretation of a fad.

The landscape has changed. Artificial Intelligence is the new frontier, and the rate of innovation is orders of magnitude faster than what we have seen in previous technology waves. Every month brings new frameworks, new protocols, and new expectations from business stakeholders. The dilemma for software architects and IT decision makers is clear: how do you harness the potential of AI without betting your future on technologies that might be obsolete next quarter?

That tension between the stability of open standards and the speed of emerging innovation is exactly where Quarkus comes in.

The standards dilemma in the AI Era

Jakarta EE remains the stabilizing force of the Java ecosystem. Its value is precisely that it does not chase every shiny new tool. By standardizing only what has been proven in practice, Jakarta EE offers interoperability, security, and long-term reliability. If you are running core banking, ERP, or regulated workloads, you know the value of having a runtime and APIs that are stable across years, not months.

But AI development does not follow that cycle. Frameworks evolve weekly. New protocols such as MCP (Model Context Protocol) and A2A (Agent-to-Agent) are still in early community discussions. Retrieval augmented generation (RAG) patterns shift as vector databases evolve. Agentic workflows are actively being reinvented across multiple languages and platforms. No committee could standardize this landscape today without freezing innovation.

This is why the ecosystem needs a dual approach. Enterprises need a secure place where rapid innovation can be tried, tested, and matured. They also need the discipline of Jakarta EE to take the best ideas and codify them into long-term contracts.

Quarkus as the innovation engine

Quarkus is the cloud-native Java stack that thrives on being ahead of the curve. Its build-time optimization model, small footprint, and instant startup makes Quarkus the natural candidate for cloud-native workloads. But its role has expanded: Quarkus has become the proving ground for new ideas that will eventually shape Jakarta EE’s AI future.

Today, Quarkus already provides concrete support for AI. The seamless integration of the LangChain4j extension makes it possible to build AI-powered applications directly in Quarkus, supporting vector stores, model connectors, and agentic workflows. RAG is available: developers can combine enterprise data with Granite models or other LLMs and deliver grounded, context-aware responses. Early frameworks like LangGraph4j allow experimentation with multi-agent orchestration while keeping a clear Java API layer. Even community protocols such as MCP and A2A are being tested in Quarkus environments, aligning AI interactions with the same messaging and security patterns Java developers already understand.

Quarkus is not just adopting these technologies, it is shaping how they become accessible to Java developers. Its extension model provides a consistent developer experience whether you are binding to PostgreSQL, Kafka, or watsonx.ai. Most importantly, Quarkus acts as the sandbox where enterprises can innovate safely. Architects can design AI-driven services using Quarkus and LangChain4j today, deploy them in production, and know that as the patterns mature, there is a path to a supported solution, based on established Jakarta EE components and approaches.

LangChain4j and the role of Context and Dependency Injection (CDI)

But it’s not only Quarkus paving the way. LangChain4j gets even more attention. While AI frameworks for Python dominate the conversation, LangChain4j brings those capabilities natively into Java. It abstracts LLM connectors, vector stores, embeddings, agentic approaches, and Retrieval-Augmented Generation patterns so architects do not need to reinvent low-level APIs.

The langchain4j-cdi project, for example, takes this further by aligning LangChain4j with Jakarta CDI. This matters because CDI is the dependency injection standard that underpins both Jakarta EE and MicroProfile. With langchain4j-cdi, any Jakarta EE runtime can provide consistent, standard-aware integration with LangChain4j features. This is not a side project, it is the seed of future standardization. What Quarkus and LangChain4j make possible today is exactly the type of innovation that Jakarta EE can potentially formalize tomorrow.

IBM’s dual investment strategy

This is where IBM’s position in the ecosystem becomes critical. Unlike many vendors who pick either speed or stability, IBM is investing in both. IBM is helping drive innovation in frameworks that allow Java developers to build AI-infused applications today through Quarkus and LangChain4j. IBM remains a core contributor to open standards such as Jakarta EE and MicroProfile, ensuring that the enterprise Java platform continues to evolve in a stable and interoperable way.

On the AI side, IBM is releasing the Granite models as open, transparent LLMs that align with the same open philosophy that made Java standards successful. At the same time, watsonx and Red Hat OpenShift AI provide the enterprise-grade environment where models can be governed, monitored, and scaled responsibly. This dual investment means architects do not need to choose between experimenting with AI now and maintaining long-term stability. IBM and Red Hat are ensuring that the Java ecosystem can move fast without breaking the trust enterprises have built in the platform.

Standards as the safety net

The lesson from Java’s history is clear: enterprises that adopt innovation too early risk fragmentation, while those that wait too long risk irrelevance. Standards exist to balance those pressures.

Jakarta EE will not be the place where MCP or A2A protocols are defined next month. But when Quarkus and LangChain4j prove the value of those patterns in real applications, Jakarta EE can codify them into APIs that every runtime must support. This is the path from innovation to stability.

For architects, this is not just a technical point. It is a risk-management strategy. By aligning with Quarkus for near-term innovation and Jakarta EE for long-term stability, enterprises can adopt AI confidently without creating a tangle of one-off integrations that will need to be rewritten in two years.

Standards as a vehicle for scaling knowledge

Interoperability, security, and long-term reliability are the traditional benefits of standards. But in the age of AI, they also serve another critical purpose: scaling knowledge across developer communities.

AI frameworks and patterns evolve at a pace that makes it almost impossible for individual developers to keep up. One month the conversation is about retrieval, the next about multi-agent orchestration, the next about fine-tuning or prompting strategies. Without a stable foundation, developer skills risk becoming disposable as the ecosystem shifts. And finding the experts for the next project is becoming a hiring adventure.

Standards change that dynamic. By codifying proven programming models, like dependency injection, messaging, persistence, observability, Jakarta EE, and MicroProfile create continuity. Developers can apply their skills to new challenges without constantly relearning the basics. The mental model of how to build, secure, and operate enterprise-grade applications remains consistent, even as the underlying AI technologies change.

For enterprises, this means that their teams can stay current without being overwhelmed by hype cycles. Developers can learn how to integrate AI through Quarkus and LangChain4j today, knowing that the patterns that prove valuable will mature into Jakarta EE or MicroProfile specifications tomorrow. The same skills will still be relevant when the next generation of AI protocols arrives.

In other words, standards do not just stabilize runtimes, they stabilize the workforce. They provide a knowledge baseline that scales across languages, platforms, and evolving implementation approaches. In an era when talent is as scarce as compute, that continuity might be the most valuable outcome of all.

Practical scenarios

It is worth making this concrete. Imagine a financial services company building an AI-assisted compliance checker. With Quarkus, the development team uses LangChain4j with Granite models to analyze incoming trade data. A vector store provides contextual retrieval of relevant regulations, and the system flags anomalies in real time. As Retrieval-Augmented Generation and MCP mature, the same application can be migrated to use standardized APIs for vector access and model invocation. That means if the company changes its runtime or platform, the application logic stays intact. Meanwhile, the models run in watsonx, ensuring governance and auditability, and the whole workload is deployed on Red Hat OpenShift AI, giving the operations team consistent lifecycle management.

This is the future path for most enterprises: experiment now, standardize later, and stay aligned with an ecosystem that will not disappear.

Future-proofing enterprise AI investments

Architects are the ones who will be asked to explain whether adopting AI in enterprise Java is a safe bet. The business will ask if building on Quarkus today locks them into a path that becomes obsolete tomorrow. The answer must be that innovation and standards are not competing forces, they are complementary. Quarkus is where enterprises can adopt AI safely today. Jakarta EE and MicroProfile will ensure that the same innovations are still usable and interoperable five years from now. IBM and Red Hat are investing across the spectrum of frameworks, standards, models, and platforms, to ensure that enterprises are not caught off guard by the rapid evolution of AI.

Innovation and standards: Both sides of the Java future

The AI wave is not the first time enterprise Java has faced rapid change, and it will not be the last. What made Java successful was the balance between innovation and standardization. That balance still applies today.

Quarkus provides the speed enterprises need to experiment and deploy AI-driven applications immediately. Jakarta EE and MicroProfile provide the long-term stability and interoperability that enterprises require. IBM and Red Hat are ensuring that both sides of the equation remain strong, by investing in next-generation frameworks like Quarkus and LangChain4j, in long-term standards like Jakarta EE and MicroProfile with the Red Hat Enterprise Application Platform, WebSphere Liberty, and in open AI foundations like Granite, Red Hat OpenShift AI, and the watsonx platform.

For software architects and IT decision makers, the message is simple: the future of enterprise Java and AI is secure. You can move fast with Quarkus without sacrificing the stability and trust that Jakarta EE has always delivered.

For an overview on what's happening at the standards level with Jakarta EE and AI, see the article What's Up With Open Standard Enterprise Java and AI.

From Standards to Speed: Why Quarkus Leads the AI-Ready Jakarta EE Future

From Standards to Speed: Why Quarkus Leads the AI-Ready Jakarta EE Future