Databricks is acquiring Electric, the company behind PGlite and a real-time sync engine. This gives every AI agent its own lightweight Postgres database in its sandbox, which synchronizes with Lakebase.
Electric builds data primitives specifically designed for agentic applications: a WASM version of Postgres that’s small enough to run within the application or agent itself, plus a sync engine that brings distributed state back to a central database.
Databricks’ reasoning is that agents behave differently from traditional software. Whereas a classic application handles predefined queries and predictable data patterns, an agent determines at runtime which data it needs, sometimes multiple times per second. That inner loop requires data to be available within the same process. And agents often work in groups, in sandboxes where the database is accessible only via a network connection to the cloud.
Building on Neon
The deal isn’t an isolated move. Last year, Databricks acquired Neon for over a billion dollars, laying an important foundation for Lakebase, which was launched in February. There’s also a direct technical connection: PGlite is built on the WASM Postgres work of Stas Kelvich, co-founder of Neon.
Databricks applies this approach within its LTAP architecture, which breaks down the barrier between OLTP and OLAP. Lakebase delivers Postgres at production scale, PGlite runs in the agent’s sandbox, and the sync engine keeps both in sync. This allows developers to build on a single Postgres standard, while control remains centralized on low-cost, sustainable object storage.