2 min Analytics

Databricks acquires Neon: serverless Postgres database for AI agents

Databricks acquires Neon: serverless Postgres database for AI agents

Neon’s technology can spin up a Postgres instance in less than 500 milliseconds, which is crucial for AI agents’ fast working methods.

The official announcement of the acquisition follows reports last week that Databricks was in advanced talks to acquire Neon for more than $1 billion. The deal is part of Databricks’ acquisition strategy to expand its AI capabilities.

According to Nikita Shamgunov, CEO of Neon, the company started four years ago to build the best Postgres solution for the cloud. The acquisition will enable Neon to accelerate this mission with the support of Databricks. For Databricks, the technology is an asset in addressing traditional database limitations.

Neon’s serverless architecture offers three key advantages. First, developers can launch a fully isolated Postgres database within 500 milliseconds. Second, the complete separation of compute and storage makes costs proportional to actual usage. Finally, the solution is fully compatible with the existing Postgres ecosystem.

Vision on database architecture

The combined vision of Databricks and Neon focuses on breaking through traditional database limitations. Conventional databases require computing power and storage to be scaled up simultaneously, which is inefficient for AI workloads. Neon’s complete separation between compute and storage is better suited to the needs of modern AI applications.

Integrating Neon’s serverless Postgres architecture with the Databricks Data Intelligence Platform will help development teams build more efficient AI agent systems. This should prevent performance bottlenecks with thousands of agents running simultaneously, simplify infrastructure, and reduce costs.

The Neon team will join Databricks upon completion of the acquisition. The planned acquisition is subject to customary closing conditions, including regulatory approvals.

Tip: Databricks moves from lakehouse to data intelligence