Databases

3/1297 → 8/810; +0.76 percentage points of formation share.

What counts as a "database" startup has quietly split into two camps that barely resemble each other. On one side, companies like ParadeDB and Lantern are not building new databases at all but layering search, vector retrieval, and AI application support directly onto Postgres, betting that the old workhorse database can absorb every new workload rather than be replaced by it. On the other side, Epsilla and Deasy Labs treat the database as a substrate for something else entirely: AI agent infrastructure and metadata orchestration, where Deasy's July 2025 acquisition by Collibra signals that even governance incumbents now see metadata management as a database-adjacent problem worth buying rather than building. PeerDB and Ubicloud represent a third, more traditional strand, moving and hosting data across systems the way infrastructure companies always have. This formation's 4.27x growth outpaces every statistically related category, including Big Data at 3.60x and Cloud Computing at 3.10x, which suggests the surge is not generic cloud demand but something specific to databases absorbing AI workloads. That timing tracks closely with the arrival of practical retrieval-augmented generation techniques and open-source vector extensions for Postgres around 2023, which made it suddenly cheap for a decades-old database to compete with purpose-built vector stores rather than lose ground to them.

No comparable outside activity is connected yet.