When we think about building the future of AI, our attention tends to gravitate toward LLMs, GPUs, and the latest vector databases. But according to Ed Huang, CTO of TiDB, a cloud-native distributed SQL database, one of the most overlooked but crucial pieces of the puzzle is… SQL.

Yes, that SQL.

“SQL is still the best bridge between the Large Language Models (LLM), the data source, and the human being.”

Rather than replacing SQL, Huang believes AI makes SQL more essential than ever, especially in a world where AI agents are tasked with acting on real-time data, securely and transparently. And with protocols like Model Context Protocol (MCP) emerging, this bridge is only getting stronger.

Why SQL Is Still the Trust Anchor

Large Language Models(LLM) can generate powerful outcomes, but when it comes to interacting with actual databases, structure and accountability matter.

SQL offers both structure and accountability. As suggested by Rahaman (2025), SQL’s structured and interpretable nature makes it well-suited for human-AI collaboration. That’s exactly what makes SQL a perfect anchor for trust in AI-driven systems, it doesn’t just work, it can be reviewed, verified, and validated by humans.

Through MCP, an LLM can craft a query, send it to a SQL-based system like TiDB, and let humans validate the intent before it touches the data source. This is the kind of safety net today’s AI systems need.

Why TiDB Bets on MySQL

While Snowflake and Databricks are doubling down on Postgres, TiDB is taking a different route, with a strong belief that MySQL is the best engine to scale AI-native applications.

TiDB adds several AI-ready enhancements on top of MySQL compatibility:

Vector indexing for retrieval-augmented generation
Full-text search for unstructured data
Python SDK for easy integration with AI workflows
Serverless cloud infrastructure for flexible deployment

Researchers at Microsoft Research emphasize that unified databases improve LLM reliability by managing multimodal data in a single system.

Scaling Intelligence, Not Just Storage

Old database challenges focused on volume. Now, it's about:

Elasticity: scale when needed, shrink to save cost
Metadata scalability: support thousands of schemas for SaaS-style apps
Multi-tenancy: handling millions of tables and connections cleanly

TiDB has spent nearly a decade refining its architecture to handle precisely these needs, making it a fit not just for startups, but for large-scale platforms like Atlassian, HubSpot, and even Databricks, which uses TiDB to manage its Spark metadata system.

Centralizing Data for Smarter AI Agents

Huang warns against scattering your data across five different systems, JSON here, vectors there, documents elsewhere.

“If your data is stored piece by piece in different sources, it becomes really hard for the LLM to access it in a useful way.”

Instead of building silos, TiDB allows everything, structured, vector, and text, to live in one SQL-based platform. That unified access layer makes it easier for AI to reason, correlate, and return accurate responses, fast.

Also read: MySQL vs PostgreSQL vs SQL Server Which One Will Power Your Success

Final Thought

As AI agents grow in capability, the foundation they’re built on becomes more critical.

SQL might not be flashy, but it offers what no black box system can: clarity, control, and collaboration between humans and machines.

Platforms like TiDB are not trying to reinvent the wheel, they’re upgrading it for the road ahead.

At Padepokan Tujuh Sembilan, we believe the future of AI isn’t about replacing everything you know. It’s about building on what already works, with the right foundation, scalability, and ecosystem to bring your vision to life.

Thinking about building your own AI-native platform? Let’s talk.

Reference: analyticsindiamag