Databricks Acquires Electric to Bring Local Postgres to AI Agent Sandboxes

Databricks is bringing the team behind database startup Electric into the company as it looks to strengthen the data infrastructure powering the next generation of AI agents.

"Agents need super-fast Postgres and this team built an amazing WASM (WebAssembly) implementation of Postgres that runs in your browser but can sync back with Postgres instances asynchronously. Exactly what blazing-fast AI agents today need. Excited to supercharge our 𝐋𝐚𝐤𝐞𝐛𝐚𝐬𝐞 𝐏𝐨𝐬𝐭𝐠𝐫𝐞𝐬 offering with these capabilities," Ali Ghodsi, Databricks CEO, said.

Databricks plans to integrate the technology with its Lakebase Postgres offering, extending database capabilities from the cloud and lakehouse to the edge.

Electric has developed data tools designed specifically for distributed, agentic applications, including PGlite, a lightweight Postgres database that can run directly inside an application, browser, or AI agent sandbox.

PGlite is built using WebAssembly, allowing Postgres to run locally without requiring a separate database server. Electric said PGlite's weekly downloads have grown from 1 million to 13 million over the past year, reflecting growing developer interest in embedded databases for distributed applications.

Electric has also developed a real-time synchronisation engine designed to keep data generated by distributed agents synchronised with centralised infrastructure. Databricks said the technology will allow agents to maintain fast local context while synchronising information with Lakebase, giving organisations centralised control over data.

The companies share a foundation in Postgres, which has increasingly become an important database technology for AI applications. Electric's PGlite builds on earlier WebAssembly Postgres work associated with Neon co-founder Stas Kelvich.

Databricks said combining Lakebase Postgres and PGlite could allow developers to give individual agents lightweight databases while maintaining a shared, synchronized data layer.

For developers, the integration is intended to support collaborative agentic applications, local Postgres databases inside agent sandboxes and real-time synchronisation between groups of agents.

Databricks said the Electric team will continue developing these capabilities as part of the company, with the goal of helping customers build and deploy agentic applications at scale.

Last month, Databricks signed a term sheet for a fresh funding round that values the company at $188 billion, marking a significant jump from its $134 billion valuation earlier this year.