Your Enterprise AI Strategy Must Start With Java, Not Python
Source: VMware Blog
2026 Enterprise AI Predictions
Thinking AI means you need Python? Using your existing Java and Spring infrastructure is the most efficient and least risky path to AI success.
It’s prediction season. Nowadays that means reading a lot of fanfic about how AI will replace your entire org chart by Q3 and, even more, why you’ll fail because you haven’t completely upended your organization. No AI ROI? That’s your fault for not changing the soul of your enterprise culture.
Annual predictions tell you more about what the author hopes will happen than what actually will. There’s nothing wrong with that; it’s just better to spray off the prediction fiction.
So, here’s what I hope happens in 2026 with enterprise AI: executives mandate that new AI development starts in their organizations’ dominant programming language, which is likely Java (Java at 30: the genius behind the code that changed tech).
Go Where Your Developers Are
If you want to start using AI to run your organization better, you don’t need to hire thousands of Python devs.
You need to give the thousands of Java devs you already have the tools they need.
Why Java Still Matters
- Enterprise‑grade language – Java powers banking, logistics, health‑care, and countless other mission‑critical systems.
- Consistently top‑ranked – Java has been in the top three programming languages for years (see the RedMonk programming index and other analyst surveys such as IDC).
- Dominant framework – The Spring Framework is the de‑facto standard for Java enterprise development. Ask any large organization what’s in their stack and you’ll hear “Java, JavaScript, and Spring.”
The Misconception
Many believe that “doing AI” requires a legion of Python (or TypeScript) developers to rewrite everything. While the AI community currently leans heavily on those languages, enterprises run on Java. Throwing away existing Java code, domain models, and operational knowledge would be absurd and irresponsible.
“We have invested a lot in domain models, some of which are even very good… to be able to leverage that as we move to the new world is really, really important.”
— Rod Johnson, founder of Spring Framework and CEO of Embabel (talk on using Java for AI)
What You Lose by Switching Languages
| Area | What You Lose by Switching |
|---|---|
| Domain knowledge | Years of business‑specific models and logic |
| Codebase | Hard‑won, well‑paid‑for Java code |
| Operational expertise | Production‑run practices, troubleshooting skills, and reliability patterns |
| Security & compliance | Established controls and audit trails |
| Skill investment | Training, hiring, and up‑skilling costs for new languages |
Re‑building all of this from scratch would also require re‑creating the Day 2 operational processes that have been refined over decades.
The Right Path Forward
- Leverage existing Java assets – Use the code, models, and data architectures you already own.
- Adopt AI‑enabled Java tooling – Platforms like VMware Tanzu accelerate AI‑driven development while preserving your Java investment.
- Extend Spring for AI – Spring’s ecosystem now includes AI extensions (e.g., Spring AI) that let you embed generative AI without leaving the Java world.
- Modernize incrementally – Introduce AI capabilities on top of your current stack rather than rewriting everything.
Bottom Line
If your organization runs Java applications, your AI strategy should start with Java. Trying to pivot to a different language adds unnecessary risk, cost, and time. By building on the Java and Spring foundation you already have, you can:
- Accelerate AI adoption
- Preserve existing domain knowledge and operational expertise
- Reduce the total cost of ownership
Your developers are already where the AI future is headed—on Java.
How to Reach the Developers
If your goal is to work with the Java applications your Java developers have built, you need a clear path to get there.
1. Treat the AI stack as an integrated platform
An AI platform isn’t just:
- a “blinking cursor” on Kubernetes (an empty shell with no services or guardrails), or
- a raw endpoint for hosting models.
It should resemble a pre‑engineered, AI‑ready Platform‑as‑a‑Service (PaaS) – e.g., the offering described in the VMware Explore 2025 Tanzu announcement.
For Java & Spring, this means tight integration between:
- Spring Framework and the platform, and
- Secure, self‑service developer services such as:
- Databases
- Message brokers
- Large Language Models (LLMs)
- Model‑Context‑Protocol (MCP) servers
- AI inference services
2. Balance Developer Experience (DX) with Operator Experience (OX)
Running, securing, and cost‑controlling applications is an operator responsibility. A good platform:
- Improves DX (easy onboarding, self‑service, consistent APIs)
- Improves OX (automation, observability, compliance, cost‑control)
Think in terms of operator‑to‑developer ratios: the higher the ratio, the more automation and reliability the platform must provide. Ideally, a handful of platform engineers can support thousands of developers and applications – see the talk on this topic here.
3. Keep the Java stack fresh
Many organizations hesitate to upgrade Java libraries and runtimes because they fear breaking “successful” applications. This leads to the legacy trap:
- Out‑of‑date dependencies
- High maintenance cost
- Slower delivery of new features
If you’re stuck in this trap, read the classic description on Cote’s “Legacy Trap” article.
Why staying current matters
- Executives want new features, not endless maintenance.
- AI‑driven products require continuous experimentation (e.g., trying Spring AI).
- New frameworks become available quickly; you need to adopt them continuously, not just once.
4. Adopt emerging standards early
Model‑Context‑Protocol (MCP) is now the de‑facto standard for adding AI capabilities to applications.
- Released: November 2024 (≈ 1 year ago) – still “yesterday” for enterprises.
- Spring integration: The Spring team shipped an official Java SDK for MCP just months after the protocol’s release – see the announcement on the Spring blog.
To use MCP (or any new AI feature), you must run a recent version of the Spring Framework.
5. General lesson for all AI programming
The same dynamics apply regardless of language:
- New languages feel easier because there’s no legacy baggage.
- Legacy languages (like Java) can still stay agile—provided you regularly upgrade the runtime, libraries, and frameworks.
TL;DR Checklist
| ✅ | Action |
|---|---|
| 1 | Choose an AI‑ready PaaS that integrates tightly with Spring. |
| 2 | Ensure the platform supports both DX and OX (high operator‑to‑developer ratio). |
| 3 | Implement a regular upgrade cadence for Java, Spring, and related libraries. |
| 4 | Adopt emerging standards (e.g., MCP) as soon as they become stable. |
| 5 | Keep a “continuous experimentation” mindset – treat new AI frameworks as first‑class citizens. |
By following these steps, you’ll move from “hand‑coddling legacy Java apps” to a modern, AI‑enabled development environment that serves both developers and operators efficiently.
The Easiest Path With the Best Chance of Success
The choice of where and how to start your AI journey is a business decision, not a technology one. By grounding your AI efforts in the operational muscle you’ve already built with Java and Spring, you minimize friction and risk. Everyone just keeps doing what they’re doing—only now with a new tool.
That hard‑won operational and development experience—paid for in decades of maintenance and gut‑wrenching production failures—gives you a platform that can continuously absorb new capabilities like Spring AI and MCP.
Get the developer experience and the operator experience right, and you unlock the organizational velocity needed to turn tentative AI experiments into a mission‑critical competitive advantage.
Join us at DevNexus 2026 (March 4–6, Atlanta) – the largest Java‑ecosystem conference in the U.S., with VMware Tanzu as the featured (Unobtanium) sponsor and a strong slate of can’t‑miss sessions.
Learn more: