In a significant development within the competitive landscape of data management, Databricks has officially unveiled its latest offering, Lakebase, during the Data + AI Summit 2025. This fully managed Postgres database is designed to seamlessly integrate with Databricks’ Data Intelligence Platform, specifically catering to the burgeoning needs of AI-driven application development. Currently in public preview, Lakebase introduces an operational layer to the lakehouse ecosystem, empowering developers and enterprises to construct AI applications and agents on a unified, multi-cloud platform.
Ali Ghodsi, the CEO of Databricks, emphasized the pivotal role of Postgres in the open-source database arena, stating, “Postgres has essentially won as the lingua franca for open source databases.” He highlighted the extensive ecosystem of extensions that have been cultivated by the community, which enable a wide array of functionalities, from vector processing to key-value storage.
Recent data from Stack Overflow’s Developer Surveys for 2023 and 2024 indicates a notable surge in PostgreSQL’s popularity, with the database now surpassing MySQL as the preferred choice among developers. Databricks is strategically positioning Lakebase within the operational database market, which is valued at over 0 billion. The architecture of Lakebase, powered by Neon, separates compute from storage, allowing for independent scaling that ensures low latency, high concurrency, and robust availability for transactional requirements.
“We’re creating a new category in the database market: a modern Postgres database, deeply integrated with the lakehouse and today’s development stacks. With Lakebase, we’re giving [enterprises] a database built for the demands of the AI era,” Ghodsi remarked. Traditional operational databases, he noted, often rely on outdated architectures that struggle to meet the rapid access and integration needs of modern AI applications. Lakebase addresses these challenges by automatically syncing with lakehouse tables, facilitating AI model serving through an online feature store, and integrating with Databricks Apps and Unity Catalogue.
Built on open-source Postgres, Lakebase benefits from a vibrant developer community and maintains compatibility with existing tools and extensions. A notable feature is its ability to create copy-on-write clones of databases, which is particularly advantageous for agent-based development and testing. The system is designed for rapid deployment, launching in under a second, and offers a usage-based pricing model.
Better Than Legacy Databases
Reynold Xin, co-founder of Databricks, pointed out that contemporary OLTP databases, whether proprietary like Oracle or open-source like MySQL and Postgres, have seen little architectural evolution since the 1990s. He elaborated on how Lakebase innovatively separates compute and storage into three layers to meet OLTP demands. Data is stored in scalable object stores, with a caching layer that minimizes latency and efficiently manages the write-ahead log (WAL). Ephemeral Postgres nodes are positioned on top to handle read-write operations, enabling fast and elastic compute capabilities.
Xin asserted that Lakebase can deliver single-digit millisecond latency at scale, a significant advantage for enterprises. Nikita Shamgunov, founder of Neon, expressed his belief that we are on the cusp of an era where AI-generated software will dominate, with AI agents increasingly taking the reins in application creation, including the databases they depend on. “In a couple of years, I think 99% of all the databases on the platform will be created by AI agents,” he stated, underscoring the transformative potential of AI in software development.
Shamgunov also highlighted how Neon’s architecture, featuring capabilities like branching and isolation, is particularly suited for AI agents that require safe, disposable environments to experiment and learn. “Lakebase removes the operational burden of managing transactional databases,” commented Anjan Kundavaram, chief product officer at Fivetran, allowing customers to concentrate on application development rather than the intricacies of provisioning, tuning, and scaling.
Jelle Van Etten, head of the global data platform at Heineken, noted that their analytical data platform is evolving into an operational AI data platform, necessitating the delivery of insights to applications with minimal latency. Databricks has reported that numerous enterprises have engaged in Lakebase’s private preview, exploring applications across various sectors, including retail personalization, healthcare workflows, and agent-based experiences.
Better Than Snowflake?
The launch of Lakebase comes shortly after Snowflake’s acquisition of Crunchy Data, which aims to enhance its offerings with a Postgres solution. Snowflake has positioned Postgres as a top choice for developers, citing its flexibility, cost-effectiveness, and native AI features, such as vector support (pgvector). This initiative builds on Snowflake’s previous foray into transactional data with Unistore, which integrates both transactional and analytical workloads within a single framework.
Vivek Raghunathan, Snowflake’s SVP of engineering, remarked, “We’re tackling a massive 0 billion market opportunity and a real need for our customers to bring Postgres to the Snowflake AI Data Cloud.” The industry appears to be gravitating towards PostgreSQL, with both Snowflake and Databricks actively acquiring smaller players. This trend underscores a broader strategy focused on preparing for AI, real-time data processing, and the evolving demands of businesses.
