MariaDB and AI: Building the Bridge That Really Matters

Not what AI can do for MariaDB – but what MariaDB can do for AI

Large Language Models are, at their core, vast read-only neural networks. The data that organisations truly depend upon does not reside inside these models; it lives in relational databases. MariaDB’s mission is to make relational data seamlessly accessible to AI.

With MariaDB Server 11.8 GA LTS, this is exactly what we deliver: vector functionality as a standard feature in a familiar, easy-to-manage technology stack. No experimental code, no extra components.

Oracle shares the vision – but misses a crucial point

What makes me especially proud is that the MariaDB Foundation Board wholeheartedly supports this vision. At our September meeting:

Our task is to ensure it becomes just as easy for AI to surface insights from a customer’s relational databases as it is to retrieve information from the Internet.

General purpose beats special-purpose

AI without access to a user’s own data is little more than a mirage. Relational databases are where the living, breathing data resides – and MariaDB’s mission is to bridge that world with AI.

We are not alone in this vision. But with open source at our core, a general-purpose foundation, strong performance, and a unified commitment from our board and community, MariaDB is uniquely positioned to make it real.

Performance is non-negotiable

This is more than a strategy. It is a direction for the industry.

Our vision: Bringing user data to AI

We have seen specialised “vector databases” emerge as pioneers in combining database technology with AI. They have played their role, but history repeats itself: users ultimately do not want to juggle niche tools. They want their everyday database – the one they already trust – to also handle their AI needs.

But one truth has not changed: the world demands open source for infrastructure software, especially databases. This is where MariaDB holds a fundamental advantage.

Strong backing from our Board

MariaDB’s Vector performance has been independently evaluated, and the results are clear: MariaDB performs strongly, not only against PostgreSQL’s PG Vector but also compared with specialised vector databases. Performance remains a non-negotiable truth – and we deliver.

  • Rohit de Souza (MariaDB plc CEO) emphasised how central this direction is for MariaDB’s product strategy, noting both enterprise demand and strong advisory support.
  • Jignesh Shah (Amazon AWS) highlighted the need to make AI features simple and approachable for developers, stressing that MariaDB must be the open source alternative where MySQL is falling short.
  • Barry Abrahamson (Automattic) underlined WordPress’ interest in MariaDB as the natural vector database in their ecosystem, urging deeper collaboration with their open source AI team.

We are not alone in articulating this vision. Oracle has also positioned itself in this direction. Their recently announced partnership with OpenAI and NVIDIA, with headline figures of USD 300 billion in AI infrastructure investment, shows the magnitude of their commitment (OpenAI, WSJ).

Conclusion: A direction for the industry

Here is where it gets truly exciting: MariaDB makes it possible to connect relational data directly to AI. Already today this is happening through RAG applications. The next step is making vibe coding work natively with MariaDB. Beyond that, new approaches to bridge the gap from AI to relational are still being invented – and will have their natural habitat in MariaDB Server.

Other board members echoed the same enthusiasm: this vision is central not only for MariaDB, but for the database industry as a whole.

I have hinted at this before (This Month in MariaDB – August 2025), but it deserves to be said again: our vision for MariaDB’s role in AI is both clear and ambitious. We do not simply want to participate in the AI space – we aim to be the bridge between real-world data and modern AI systems.

Similar Posts