Databricks

Databricks sells a cloud data and AI platform built around the lakehouse concept—a single architecture that combines the flexible storage of data lakes with the reliability and performance of data warehouses. The core offering, Databricks Lakehouse, runs on a customer’s existing AWS, Azure, or Google Cloud account and layers open-source technologies such as Delta Lake, Apache Spark, and MLflow with proprietary optimization, governance, and collaboration tools. Data engineering teams use it for ETL and streaming, data scientists use it for model training and feature stores, and—increasingly—software developers use Mosaic AI and DBRX to embed generative AI into business applications.

The platform monetizes primarily through usage-based subscriptions tied to compute and storage consumed on the lakehouse, alongside premium capabilities in Unity Catalog governance and Mosaic AI model serving. Databricks targets large enterprises and fast-growing digital natives that need to unify analytics and machine learning on one copy of data rather than siloing them across separate warehouses and lakes. By shipping its own open-weights model and acquiring MosaicML, the company is trying to move up the stack from infrastructure provider to full-stack AI platform vendor.

Databricks was founded in 2013 by a group of University of California, Berkeley researchers—Ali Ghodsi, Matei Zaharia, Ion Stoica, Reynold Xin, Patrick Wendell, and Andy Konwinski—who had created Apache Spark in the AMPLab. The company began as a commercial support and hosting provider for Spark, but soon pivoted to a fully managed cloud platform that abstracted cluster management and notebook collaboration for data teams. This academic-to-commercial lineage gave Databricks deep credibility in the open-source community and early enterprise traction with advanced analytics workloads.

The defining strategic shift came with the coining of the 'lakehouse' architecture in the late 2010s, positioning Delta Lake and the Databricks runtime as a replacement for legacy data warehouses. The model gained rapid enterprise adoption as companies sought to reduce data duplication between lakes and warehouses. In 2023, Databricks acquired MosaicML, signaling a decisive bet on generative AI model training and serving. The subsequent release of the DBRX open-weights model in 2024 and a landmark $10 billion Series J round at a $62 billion valuation cemented its status as the largest private data+AI platform vendor.

Databricks’ primary rival is Snowflake, which offers a simpler, SQL-first data cloud but has been racing to add AI and ML capabilities. Databricks generally wins when enterprises need advanced analytics, custom model training, or open-source flexibility; Snowflake wins when the buyer is a SQL-centric analyst team prioritizing ease of use and native data sharing. Against hyperscalers, Databricks competes by being multi-cloud and workload-agnostic, whereas AWS, Azure, and Google prefer to lock customers into native stacks. The MosaicML acquisition and DBRX release give Databricks a credible story for enterprises that want to own their models rather than rent APIs from OpenAI or Anthropic.

The platform’s biggest competitive vulnerability remains complexity. Implementing a lakehouse demands skilled data engineers and platform teams, creating an opening for simpler competitors and managed Spark alternatives. Databricks is investing heavily in serverless autoscaling and natural-language interfaces to close that gap. If it succeeds, the $62B valuation will look prescient; if not, large customers may default to the 'good enough' AI services bundled by their cloud provider.