Databricks in Talks to Acquire Neon, a Serverless Database Startup, for Around $1 Billion

Databricks is in advanced negotiations to acquire Neon, a San Francisco-based startup that builds a serverless, PostgreSQL-compatible cloud database optimized for modern and AI-driven applications, according to people familiar with the matter. The potential deal is expected to be valued at approximately $1 billion, though the total amount could exceed that once employee retention packages are included. While some insiders consider the deal close to final, the discussions remain active and could still fall through.

Neon has raised more than $130 million in funding from investors including Microsoft’s M12 venture fund. Its platform is built around a serverless architecture that separates compute and storage, allowing each to scale independently based on real-time usage. The system automatically adds or removes resources as workload requirements change, helping developers avoid over-provisioning and manage costs more efficiently.

This approach is particularly aligned with use cases where workloads are short-lived and demand fluctuates rapidly, such as those seen in AI-based systems. Neon’s platform supports the storage of vector embeddings, which are commonly used by machine learning models to represent data. It also offers the ability to provision a new PostgreSQL instance in about one second, a feature designed to support high-frequency, on-demand interactions.

These characteristics have made Neon relevant for developers building AI coding assistants and task-based AI agents, which frequently create and discard databases as part of their operation. In addition to technical performance, Neon’s pricing model is usage-based, charging customers by the second rather than through pre-allocated compute blocks or long-term provisioning, making it more cost-effective for dynamic and fast-executing applications.

Databricks’ reported interest in Neon follows its recent acquisition of Fennel AI, a startup focused on real-time data pipeline infrastructure. This comes after last year’s $1.3 billion acquisition of MosaicML, a company that develops large language models, and an earlier acquisition of Lilac AI, which specializes in dataset management. These deals reflect Databricks’ broader effort to position itself as a comprehensive platform for enterprise AI development and deployment.

The company has also partnered with Anthropic to integrate Claude language models into its platform, a move aimed at helping customers build AI agents capable of executing multi-step business processes such as planning, document processing, or customer support workflows. Acquiring Neon would add transactional database capabilities that complement these efforts, enabling low-latency, ephemeral data operations as part of AI workloads.

Neon’s technical design includes features such as connection pooling to improve application performance and reduce latency, as well as granular access controls down to the row level, important for data governance and security. These features, while increasingly expected in modern database environments, are packaged within a fully managed, cloud-native service, offering appeal to both startups and enterprise developers.

If the deal closes, it would give Databricks direct ownership of a modern, AI-aligned transactional database and further deepen its infrastructure stack. While Neon remains early-stage, with ongoing product evolution and customer acquisition, the Databricks platform could give it broader market access and operational scale.

Led by CEO Nikita Shamgunov, previously the co-founder of MemSQL (now SingleStore), Neon has positioned itself as an alternative to monolithic database systems by focusing on developer experience, flexible deployment, and pricing efficiency. Its combination of serverless architecture and AI-oriented features makes it a strategic fit within Databricks’ expanding ecosystem.