# Databricks Mosaic

Databricks Mosaic is a generative AI startup acquired by Databricks in 2023 for $1.4 billion, specializing in efficient large language model training and deployment. Its technology underpins Databricks' Data Intelligence Platform.

Databricks Mosaic refers to the generative AI company MosaicML, which was acquired by Databricks in June 2023 for $1.4 billion. MosaicML specialized in building efficient systems for training and deploying [large language models](https://www.wikiprompt.org/wiki/large-language-model) (LLMs), offering a platform that enabled organizations to create custom AI models using their own data. The acquisition integrated MosaicML's technology into Databricks' lakehouse architecture, forming the foundation of its Data Intelligence Platform and enhancing its [generative AI](https://www.wikiprompt.org/wiki/generative-ai) capabilities.

MosaicML was known for its open-source models, including MPT (Mosaic Pretrained Transformers), which were designed to be cost-effective and accessible for enterprises. The company's tools allowed users to train, fine-tune, and serve LLMs at scale, addressing the growing demand for private and customizable AI solutions. Following the acquisition, Databricks leveraged MosaicML's expertise to develop products like Dolly, an open-source language model with 6 billion parameters, and to integrate third-party models from providers such as [OpenAI](https://www.wikiprompt.org/wiki/openai) and [Anthropic](https://www.wikiprompt.org/wiki/anthropic) into its platform.

## Background and Founding

MosaicML was founded in 2021 by Naveen Rao and Hanlin Tang, both of whom had prior experience in AI and hardware. Rao previously co-founded Nervana Systems, a deep learning hardware startup acquired by Intel in 2016. Tang had worked at Intel and held a PhD in neuroscience. The company was headquartered in San Francisco, California, and quickly gained attention for its focus on making LLM training more efficient and affordable.

The startup raised significant funding from venture capital firms, including a $64 million Series B round in 2022 led by Tiger Global Management. MosaicML's platform was built on [machine learning](https://www.wikiprompt.org/wiki/machine-learning) principles, emphasizing distributed training and optimization techniques that reduced the computational resources required for large-scale models.

## Technology and Products

MosaicML's core offering was a managed service for training and deploying LLMs, which supported popular architectures like the [Transformer](https://www.wikiprompt.org/wiki/transformer). The platform included tools for data preparation, model training, and inference, all designed to run on cloud infrastructure from providers like [Amazon Web Services](https://www.wikiprompt.org/wiki/amazon-web-services), [Microsoft Azure](https://www.wikiprompt.org/wiki/azure), and [Google Cloud](https://www.wikiprompt.org/wiki/google-cloud).

One of MosaicML's key innovations was its open-source model family, MPT, which included MPT-7B and MPT-30B. These models were released in 2023 and were notable for their competitive performance relative to larger proprietary models, achieved through efficient training techniques. MosaicML also developed Composer, a PyTorch-based library for neural network training, and Streaming, a dataset streaming library that facilitated efficient data loading.

## Acquisition by Databricks

In June 2023, Databricks announced its intention to acquire MosaicML for $1.3 billion, with the deal closing later that year. The acquisition was part of Databricks' broader strategy to expand its [artificial intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence) capabilities, particularly in the realm of generative AI. Databricks integrated MosaicML's technology into its lakehouse platform, enabling customers to build and deploy custom LLMs alongside their data analytics workloads.

The acquisition also led to the development of Databricks' Data Intelligence Platform, which combines data management with AI model training and serving. MosaicML's tools were rebranded as Databricks' foundation model training and serving capabilities, and the company continued to release open-source models under the MPT name.

## Impact and Legacy

MosaicML's acquisition by Databricks was a significant event in the AI industry, highlighting the growing importance of efficient LLM training and the trend of data platforms integrating generative AI. The technology contributed to Databricks' ability to offer a secure environment for enterprises to train models on proprietary data, a key differentiator against competitors like [Google DeepMind](https://www.wikiprompt.org/wiki/google-deepmind) and [OpenAI](https://www.wikiprompt.org/wiki/openai).

Under Databricks, MosaicML's legacy continued through products like Dolly and the integration of models from [Anthropic](https://www.wikiprompt.org/wiki/anthropic) and [OpenAI](https://www.wikiprompt.org/wiki/openai) into the Databricks platform. The acquisition also influenced Databricks' financial growth, contributing to its valuation reaching $188 billion by July 2026, following a $3 billion funding round led by Coatue Management.

MosaicML's focus on open-source models and efficient training methods has had a lasting impact on the AI community, encouraging broader access to LLM technology and fostering innovation in model optimization.

---
Source: https://www.wikiprompt.org/wiki/databricks-mosaic
License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/)
Last updated: 2026-09-07T02:31:30.57773+00:00