Confirming recent reports, Databricks Inc. today announced that it’s acquiring Neon Inc. to enhance its cloud data platform, in a deal the Wall Street Journal reported is worth about $1 billion.

San Francisco-based Databricks received a $62 billion valuation after its most recent funding round last year. The company’s namesake platform enables organizations to store data, analyze it and use it to power artificial intelligence workloads. The software has been adopted by more than 60% of the Fortune 500.

Neon is a venture-backed startup with a managed cloud database. It raised more than $130 million in funding prior to the acquisition from Microsoft Corp.’s M12 fund and other backers.

Neon’s platform is based on PostgreSQL, one of the industry’s most popular open-source relational databases. The company has replaced PostgreSQL’s storage engine, the component that holds users’ records, with an internally developed technology. Neon says its customized implementation of the database is easier to manage and performs certain tasks faster.

The company’s storage engine is powered by a component of PostgreSQL called the write-ahead log, or WAL. If an outage occurs while users try to add a record to a PostgreSQL database, the record might be saved in an incomplete form or not at all. The WAL module is designed to avoid such technical issues. 

When users add a record to a PostgreSQL database, the record is not sent directly to the database but rather saved to the WAL first. This creates a backup copy of the record. If a crash interrupts a data write operation, the affected data can be recovered from the WAL.

Neon’s custom storage engine implements the WAL as a cluster of servers spread across multiple data centers. When users add a record to Neon, the record is first sent to the WAL, which saves it to disk. The file is subsequently moved to servers called Pageservers that hold it in an immutable format, which means it can’t be edited. The Pageservers are responsible for making files available to the applications that use a Neon database.

Neon’s architecture makes it possible to launch new database instances in under a second. That capability lends itself well to powering AI agents, which was one of the factors behind Databricks’ decision to swoop in. The company will integrate Neon’s technology into its cloud data platform to help customers run AI agents.

Many of the tasks for which programming-optimized AI agents are used require launching a database. Usually, the process takes several minutes, which means generating prompt responses takes that much longer. Neon’s ability to launch database instances in under a second allows AI agents to respond to user requests faster.

The platform also eases certain related tasks. When multiple AI agents use information from the same database instance, the database can become overwhelmed with requests. Developers can use Neon to quickly create a separate copy of a database’s contents for each AI agent.

“By bringing Neon into Databricks, we’re giving developers a serverless Postgres that can keep up with agentic speed, pay-as-you-go economics and the openness of the Postgres community,” said Databricks co-founder and Chief Executive Officer Ali Ghodsi.

Databricks expects to complete the acquisition by July 31.

Image: Neon