Databricks announced Tuesday that it's acquiring Electric, the startup behind the WebAssembly-based Postgres project PGlite and the Electric sync engine, as agentic applications transform how developers interact with databases. The Electric team will become part of Neon, the serverless Postgres company Databricks purchased for roughly $1 billion last year and the foundation of its Lakebase database service. The companies did not reveal financial terms of the transaction.
PGlite, a complete Postgres database compiled to WebAssembly, has surged from 1 million to 13 million weekly downloads over the past year, according to the companies. The database runs in browsers, Node.js processes, or the sandboxed environments that agents use for code execution, and it supports dynamic extension loading including pgvector, the favored Postgres vector extension. On Lakebase, agents now generate roughly four times more databases than human users do, the report states. The typical project carries about 10 database branches, while some projects exceed 500 branch iterations in depth. For certain application types on Lakebase, the average database compute session now runs for less than 10 seconds.
The Electric sync engine sits at the heart of Databricks' interest in the acquisition. This engine maintains a central Postgres database that synchronizes in near real-time with browser tabs, mobile applications, or agents—what Databricks describes as the multiplayer model of Figma or Google Docs applied to Postgres and the agents that rely on it. The Neon team notes that "complex problems like conflict resolution, partial replication, and reconnection logic make real-time sync difficult to build from scratch," explaining why Databricks likely chose acquisition over internal development. Electric's founders, James Arthur and Valter Balegas, confirmed that "everything we've previously open sourced stays open source," covering the sync engine, PGlite, Durable Streams, and TanStack DB, though Electric Cloud, the company's hosted service, is shutting down.
Traditional non-agentic applications share a single database among many clients, and that database represents the most permanent component of the stack, Databricks explains. Agent workloads flip this model. Agents prefer to branch databases the same way they branch code, a pattern that Neon designed its architecture to accommodate. In practice, a coding agent creates a sandbox, launches PGlite inside it, builds and tests against the database, then either discards everything or synchronizes the outcome with a Lakebase branch. Because Lakebase separates storage from compute and stores data in Postgres page formats on object storage, creating that branch becomes an inexpensive copy-on-write metadata operation. As coding agents push creation costs toward zero, the volume of applications explodes, and most remain small—yet even a serverless database that scales to zero establishes a floor on what the smallest viable application costs to operate.
The acquisition reunites two pieces of work that began at the same company. PGlite originated as an experiment by Neon co-founder Stas Kelvich, who compiled Postgres to WebAssembly to test whether it could function client-side. Electric adopted the project and developed it into a production-ready system, with that repository becoming the foundation of PGlite. The Neon team writes that as coding agents drive creation costs to zero, "the number of applications explodes, and most of them are small," while noting that "you can't have an age of abundance if every app requires a fixed minimum of compute." The deal extends Databricks' database acquisition streak, which includes Neon itself and the transactional processing startup Mooncake. The combined technologies position Databricks to support database architectures where ephemeral, agent-driven compute replaces long-lived, human-managed infrastructure as the dominant pattern. The convergence of local-first sync and serverless branching suggests database architecture may soon privilege throwaway experimentation over durable persistence, challenging assumptions that have shaped data infrastructure for decades.