Databricks

• 1. Introduction

  ›

  Databricks gave one of its 2023 products a name borrowed from a sheep. In March of that year, the technology world was reacting to the runaway popularity of OpenAI's ChatGPT. Databricks answered with an open source language model called Dolly, named after Dolly the sheep. It gave outside developers a free way to build their own chatbots. That was a small release from a company that had started a decade earlier as a research project at a public university. This documentary traces how that project became one of the most valuable private software companies in the world. It also asks why rival cloud providers and rival AI labs all eventually signed partnerships with the same firm. And it asks what, exactly, its engineers spent over a decade building.

• 2. From Research Project To Company

  ›

  The AMPLab project at the University of California, Berkeley set the stage for Databricks. Researchers there were building Apache Spark, an open-source distributed computing framework built on the Scala language. Seven of them, Ali Ghodsi, Andy Konwinski, Arsalan Tavakoli-Shiraji, Ion Stoica, Matei Zaharia, Patrick Wendell, and Reynold Xin, founded Databricks in 2013. The new company set up its headquarters in San Francisco.

  By 2017, Microsoft Azure had integrated the platform directly into its own cloud, launching it as Azure Databricks. In February 2021, Databricks partnered with Google Cloud to integrate with the Google Kubernetes Engine and Google's BigQuery platform. That same month, the company reported having more than 5,000 customers. The tools drawing in those customers rested on an architecture the company had already started calling a lakehouse.

• 3. The Lakehouse Platform

  ›

  The platform runs natively across three cloud providers at once: Amazon Web Services, Microsoft Azure, and Google Cloud Platform. An open data marketplace sits on top of it, built on a protocol called Delta Sharing. Databricks calls its core architecture a lakehouse, a design that combines elements of data warehouses and data lakes to manage both structured and unstructured data. Underneath that architecture sits Delta Lake, an open-source project the company maintains that adds ACID transaction support to data lakes.

  Apache Spark forms the technical foundation underneath the lakehouse, letting analysts run queries on semi-structured data without a traditional database schema. In October 2022, the Lakehouse product received FedRAMP authorization, clearing it for use by the U.S. federal government and its contractors. Databricks also built MLflow and Koalas, open-source projects spanning data engineering, data science, and machine learning.

  In June 2020, Databricks launched Delta Engine, a fast query engine built to work with Delta Lake, Apache Spark, and MLflow. Five months later, in November 2020, the company introduced Databricks SQL, initially called SQL Analytics, for running business intelligence and reporting directly on data lakes. Analysts could query it with standard SQL or connect it to outside tools including Holistics, Tableau, Qlik, Sigma, Looker, and ThoughtSpot. The same platform also supported machine learning, data storage and processing, and streaming workloads.

  In June 2024, Databricks open sourced Unity Catalog, a unified governance offering, under the Apache Spark 2.0 license. The company also co-organized massive open online courses on Spark and ran a community conference called the Data + AI Summit, previously known as Spark Summit. That yearly gathering would later become the venue where Databricks unveiled its move into building AI agents.

• 4. Racing Into Artificial Intelligence

  ›

  In early 2024, Databricks released a set of tools called Mosaic for customizing, fine-tuning, and building AI systems. The set included AI Vector Search for building retrieval-augmented generation models. It also included AI Model Serving, for deploying and monitoring models, and AI Pretraining, for building custom large language models.

  In March 2024, Databricks released a foundation model called DBRX under its own Databricks Open Model License. DBRX used a mixture-of-experts architecture built on the open-source MegaBlocks project, and cost $10 million to create. It carried 136 billion parameters in total, but used only 36 billion of them, on average, to generate a response. According to the company, DBRX outperformed other open-source models available at the time, including beating Llama 2 at logic puzzles and general knowledge questions. Databricks said companies could use DBRX as a foundation to build their own customized AI models on proprietary data.

  Databricks folded its newest AI capabilities into a broader release it called the Data Intelligence Platform. That release combined the existing lakehouse architecture with generative AI features gained through a recent acquisition.

  At the 2025 Data + AI Summit, Databricks introduced three new products at once. Agent Bricks became a development platform for building AI agents. Lakebase arrived as a transactional, or OLTP, database. Databricks One launched as a no-code AI business intelligence platform. At the same event, the company said its Databricks SQL product was on track to reach a $1 billion revenue run rate. Behind several of these launches sat a string of acquisitions Databricks had made over the preceding years, starting with a tool called Redash.

• 5. Buying Into New Markets

  ›

  In June 2020, Databricks bought Redash, an open-source tool for building data visualizations and interactive dashboards. A year later, in 2021, it acquired 8080 Labs, a German no-code company whose product, called bamboolib, let people explore data without writing any code.

  In May 2023, Databricks bought a data security group called Okera, extending its data governance capabilities. The following month, it acquired the open-source generative AI startup MosaicML for $1.4 billion. In October of that year, the company bought a data replication startup called Arcion for $100 million.

  In 2024, Databricks bought a data-management startup called Tabular for more than $1 billion. The following year, it acquired a serverless database startup named Neon for around $1 billion. That buying spree paralleled another kind of spending: in early March 2025, the company said it would invest $1 billion in San Francisco's downtown.

• 6. Chasing A Bigger Valuation

  ›

  In September 2013, Databricks raised $13.9 million from Andreessen Horowitz, saying it wanted to build an alternative to Google's own MapReduce system. New Enterprise Associates led two more rounds, $33 million in 2014 and $60 million in 2016. Andreessen Horowitz then led three further rounds: $140 million in 2017, and two more in 2019 worth $250 million and $400 million. Microsoft was also a noted investor in that 2019 Series E round, though the company did not disclose the amount.

  In February 2021, Databricks raised a $1 billion Series G round led by Franklin Templeton Investments, valuing the company at $28 billion. Its total funding stood at $1.9 billion, with additional backing from Amazon Web Services, CapitalG, a growth equity firm under Alphabet, and Salesforce Ventures. In August 2021, Databricks closed an eighth funding round led by Morgan Stanley, raising $1.6 billion and valuing the company at $38 billion.

  In September 2023, Databricks raised a $500 million Series I round backed by Capital One Ventures and Nvidia. A year later, in December 2024, the company announced $10 billion in new financing at a $62 billion valuation, led by Thrive Capital.

  In August 2025, Databricks announced a $1 billion Series K round led by Thrive Capital and Insight Partners, pushing its valuation past $100 billion. Just three months later, in December 2025, the company completed a $4 billion Series L round led by Insight Partners, Fidelity, and J.P. Morgan, valuing Databricks at $134 billion. It was the company's third major fundraise in less than a year. Coverage of the round described it this way: 'It comes as the company focuses on building products that address the needs of the AI revolution: a database for AI agents, an AI agent platform, and apps that let companies build and deploy data and AI applications.' The company's expanding balance sheet was about to fund partnerships with three of the most prominent players in artificial intelligence.

• 7. Partnering With Rivals

  ›

  In March 2025, Databricks signed a five-year partnership with Anthropic. The deal, valued at $100 million, folded Anthropic's AI products directly into the Databricks platform. The following month, in April 2025, Databricks was named to the Forbes AI 50 list. By June 2025, the company had entered a four-year partnership with Alphabet, agreeing to bring Google's Gemini models into its platform. Three months after that, in September 2025, Databricks struck a similar deal with OpenAI, agreeing to incorporate the company's large language models into its platform for another $100 million.

  In December 2024, Databricks joined Wiz and Workday in running their products on AWS through a new purchasing button called 'Buy with AWS.' Separately, in June 2025, Databricks also announced a partnership with Google Cloud to integrate its platform with Google Cloud's own services. All of those partnerships fed a business now generating billions of dollars a year, a scale the company would put a number on in early 2026.

• 8. Revenue And What Comes Next

  ›

  Databricks reported $1.6 billion in revenue for its 2023 fiscal year. By the 9th of February 2026, that figure had grown into a $5.4 billion revenue run rate, up more than 65 percent year over year. The company also said it had delivered positive free cash flow over the trailing twelve months.

  In March 2026, Databricks released Genie Code, an AI agent built for data science and engineering tasks. The same month, the company also announced Lakewatch, an agentic security platform designed to automate threat detection and response. With Lakewatch, Databricks extended its reach from managing a customer's data to defending it, a natural next step for a platform built to hold nearly everything a company runs on.