Databricks acquires Electric, winds down its cloud, and keeps the local-first Postgres open source

Databricks has acquired Electric, the team behind the WASM Postgres project PGlite and the Electric sync engine, and is folding it into Neon, the serverless Postgres company Databricks bought last year in a deal widely reported around $1 billion. For developers the announcement splits cleanly in two: the hosted Electric Cloud is winding down, while the open-source projects stay open, and the direction of travel now points at a specific bet, that AI agents want their own local Postgres.

What Databricks actually bought

The asset that matters most is PGlite, a full Postgres compiled to WebAssembly that runs in the browser, in a Node process, or inside the sandbox an agent uses to execute code. Electric reports PGlite grew from roughly 1 million to 13 million weekly downloads over the past year, which is the kind of adoption curve that makes a small team acquisition-worthy on its own. Alongside it come the Electric sync engine, a real-time layer that keeps distributed clients aligned with a central Postgres record, plus Durable Streams and TanStack DB, the reactive client-side store Electric contributes to.

The team lands inside Neon, which is now the foundation of Databricks’ Lakebase managed Postgres service. IT Brief notes a detail the vendor posts skip past: PGlite’s technical lineage traces back to work by a Neon co-founder before Electric turned the idea into an embeddable product, so the deal reads less like two strangers merging and more like a reunion of a shared local-Postgres thread.

The split outcome for people already shipping on it

This is where the practical consequence lives. Electric Cloud, the hosted sync service, is winding down. Electric says existing customers have been contacted directly and will need to self-host the sync engine or move to another provider, with support promised through the transition. No shutdown date has been published, which is the one piece of uncertainty worth tracking: teams depending on Electric Cloud in production now have a migration on their roadmap without a firm deadline to plan against.

The open-source side is steadier. Electric states plainly that everything it has open sourced stays open source, covering the sync engine, PGlite, Durable Streams, and TanStack DB. For the large PGlite user base that is the reassurance that counts, because an embedded database is a dependency you compile into a shipped app, not a service you can swap in an afternoon. The honest caveat is that “stays open source” describes the license, not the roadmap, and the roadmap now answers to Databricks.

Why agents are the throughline

The strategic framing across all three announcements is the same. Databricks wants each AI agent to run its own local Postgres instance through PGlite while syncing state back to a central Lakebase database, so an agent keeps fast local context and still reconciles against the definitive cloud record. That is a concrete answer to a real problem in agent systems, where many short-lived workers need isolated state that does not corrupt a shared source of truth.

For a working developer, the near-term signal is less about Databricks’ platform and more about where local-first data tooling is being pulled. Sync engines and embedded databases spent the last few years positioned around offline apps and collaborative UIs. This deal reframes the same primitives as agent infrastructure, and it puts one of the most-downloaded pieces of that stack under a data-platform vendor rather than an independent startup.

What it means for teams

If you run Electric Cloud, treat this as a scheduled migration and start scoping the self-host path now rather than waiting for a shutdown notice. If you depend on PGlite, the sync engine, Durable Streams, or TanStack DB, nothing breaks today, but the governance question is worth a bookmark: an open license protects the code you already have, not the pace or direction of what ships next. And if you are weighing serverless Postgres options, this consolidates Neon and Lakebase as the local-first end of that market, against Supabase on the platform side and embedded engines like Turso on the edge side.

The larger point is that the local-first database category is being absorbed into the agent story faster than its original app-sync framing anticipated. The signal to watch is whether the open-source projects keep an independent release cadence under Databricks, or whether their roadmaps quietly narrow to what Lakebase needs. Stackmaven will check back on or around November 12.

Sources cited

1. Electric is joining Databricks (Electric blog) electric.ax
2. Electric is joining team Neon at Databricks (Neon blog) neon.com
3. Electric joins Databricks to bring WASM Postgres to AI agent sandboxes (Databricks blog) www.databricks.com
4. Databricks acquires Electric to give every AI agent its own Postgres database (The New Stack) thenewstack.io
5. Databricks buys Electric to power AI agent sandboxes (IT Brief) itbrief.co.nz