Ardent clones a 1.6TB Postgres database in six seconds

Ardent clones a 1.6-terabyte production Postgres database in under six seconds, per a report on the startup's launch. Founder Vikram Chennai knows why that matters. His last company, an AI agent called AI Data Engineer, reached $200,000 in annual recurring revenue, then stalled. It had no safe way to verify the SQL it generated.

Each clone is isolated at both compute and storage, so testing against it can't touch the original database. It also matches production exactly, not a sampled copy, which matters because SQL that passes against fake data can still break on the real thing.

Storage costs nothing until data changes, and compute for an idle clone scales to zero. That keeps a full test copy nearly free to spin up and tear down.

Crane Venture Partners led a $2.15 million pre-seed round for the two-person company, which came out of Y Combinator's Spring 2026 batch.

PlanetScale, Neon and Supabase have all shipped MCP servers that let AI agents query and migrate databases directly, so Ardent isn't the only way in. Branching speed, not raw access, is what will separate them, and Ardent's six-second number so far has no independent benchmark behind it.