Databricks

Databricks is a data lakehouse platform for SQL warehousing, data engineering, machine learning and AI agents on open table formats, sold pay as you go.

Databricks puts a company's data pipelines, SQL analytics and AI applications on one copy of its data, held in open formats in cloud object storage. It is the Data + AI Platform of Databricks Inc., the independent San Francisco company founded in 2013. The design is a lakehouse, where compute is decoupled from storage. Warehouses resize or scale out without moving the data underneath them. Data, analytics and AI workloads run in one place, across the clouds a company already uses. Data engineering runs on Lakeflow Jobs, Lakeflow Pipelines and Lakeflow Connect, which orchestrate processing and analytics pipelines. Databricks SQL serves BI through SQL warehouses in Classic, Pro and Serverless forms, with materialized views, streaming tables and a result cache. Unity Catalog governs the tables, applying row filters and column masks at query time. Lakebase adds a managed Postgres database for applications that serve data and features. Agent Bricks, Model Serving, AI Functions and AI Search cover building and serving models and agents, and Genie answers questions asked in plain language. The platform is built for data engineers, analysts, data scientists and machine learning engineers who want to share one governed set of tables. It fits organizations whose data already lives in cloud storage and that want it kept in Delta Lake or Apache Iceberg tables. One copy of the files serves both Delta and Iceberg clients, so outside engines can read the same data. Application developers use Lakebase and Databricks Apps to build on company data next to the analytics work. Platform owners track compute through account budgets that email named people when a threshold is crossed. A common pattern streams events from Apache Kafka or Azure Event Hubs into streaming tables and shapes them with Lakeflow Pipelines. Analysts then query the results from a SQL warehouse and publish dashboards in Tableau. Data scientists train and serve models on those governed tables, and teams that model data in dbt point it at the same warehouses. OpenSharing, formerly Delta Sharing, lets partners and SAP Business Data Cloud read live tables without a copy. When a bad load lands, time travel restores a table to its earlier version. Databricks suits a company that wants its data warehouse, its pipelines and its AI agents governed in one place over open data.