At AWS re: Invent this month, I spoke to one and only Ed Huang, Co-Founder and CTO of TiDB, powered by PingCAP, to talk about something many teams are quietly struggling with.

The last few years in data were all about unbundling.

- One database for transactions.

- One for analytics.

- One for vectors.

That worked when workloads were separate.

But AI changed the rules.

As Ed put it, today the same application, or even a single AI agent, jumps across transactions, analytics, and vector search in milliseconds. No developer wants to manage five databases. Agents cannot do that at all.

That is why PingCAP is betting on unification with TiDB X.

One system that can:

- Scale like an analytical warehouse

- Behave like a transactional database

- Support vectors and search natively

- Run on object storage with elastic compute

Keep a single SQL interface with strong guarantees

This does not mean specialized databases disappear. Many of them will continue to win in narrow use cases.

But the center of gravity is clearly shifting.

As AI agents become first-class users of data systems, platforms that remove boundaries matter more than platforms that optimize for one workload in isolation.

This was a sharp, honest conversation on where databases are heading in the AI era.

#data #ai #awsreinvent #qlik #agenticai #sovereigncloud #aws #theravitshow