Databricks acquires Electric to give every AI agent its own Postgres database
Databricks on Tuesday announced that it’s acquiring Electric, the startup behind the WASM-based Postgres project PGlite and the Electric sync engine, as The post Databricks acquires Electric to give every AI agent its own Postgres database appeared first on The New Stack .
Databricks has announced the acquisition of Electric, the company behind the Postgres project PGlite and the Electric sync engine, as agentic applications transform how developers utilize databases. The Electric team will join Neon, a serverless Postgres company Databricks acquired for around $1 billion last year, forming the foundation of its Lakebase database service.
The companies have not disclosed the deal's terms. PGlite, a complete Postgres database in WebAssembly (WASM), is now a part of Databricks' portfolio. It runs in various environments, including the browser, Node.js processes, and sandboxes used by agents to execute code. PGlite supports dynamic extension loading, including pgvector, a popular Postgres vector extension.
The sync engine at the core of Electric plays a significant role in Databricks' interest in the company. This engine maintains a central Postgres database that can be synced in near real-time with browser tabs, mobile apps, or agents, creating a multiplayer model akin to Figma or Google Docs but applied to Postgres and the agents that utilize it.
This sync engine is what makes real-time sync challenging to build from scratch. Electric's founders, James Arthur and Valter Balegas, state that all previously open-sourced components, such as the sync engine, PGlite, Durable Streams, and TanStack DB, will remain open source after the acquisition. However, the hosted Electric service will be discontinued, and cloud users will need to self-host or migrate to another provider.
This acquisition is part of Databricks' growing database acquisitions, which include Neon and the transactional processing startup Mooncake. Databricks argues that traditional non-agentic applications share a single database among many clients, while agent workloads create numerous databases and branches. For instance, agents now create about four times more databases than human users on Lakebase, with the average project carrying around 10 database branches, some running over 500 iterations deep.
As a result, database compute for some applications on Lakebase now exists for under 10 seconds, as agents frequently branch databases similarly to how they branch code. This phenomenon has led to a surge in the creation of applications, as creating a database branch is a relatively cheap operation.
Written by urgent.news from The New Stack's reporting — not their text. Machine-written — it may contain errors, so check the original before relying on it.
