Databricks, the renowned Data and AI company, has announced its definitive agreement to acquire MosaicML, a leading generative AI platform. This strategic move aims to make generative AI accessible to organizations of all sizes, empowering them to develop, control, and secure their generative AI models using their own data. The acquisition, valued at approximately $1.3 billion, including retention packages, positions Databricks as a key player in driving the democratization of AI.
MosaicML has gained recognition for its advanced MPT large language models (LLMs). Their MPT-7B model has already garnered over 3.3 million downloads, and the recent release of MPT-30B has demonstrated the efficacy of enabling organizations to quickly build and train their own state-of-the-art models in a cost-effective manner. Prominent customers such as AI2 (Allen Institute for AI), Generally Intelligent, Hippocratic AI, Replit, and Scatter Labs have harnessed MosaicML’s platform for a diverse range of generative AI use cases.
Ali Ghodsi, Co-Founder and CEO of Databricks, expressed excitement about the joint efforts of Databricks and MosaicML in democratizing AI and establishing the Lakehouse as the premier platform for building generative AI and LLMs. He emphasized their shared vision, which is rooted in transparency and a history of open-source contributions, and highlighted the value this collaboration will bring to customers during the ongoing computing revolution.
The integration of Databricks’ Lakehouse Platform with MosaicML’s technology will offer customers a straightforward and efficient means of maintaining control, security, and ownership over their valuable data without incurring exorbitant costs. MosaicML’s automatic optimization of model training has shown 2x-7x faster training compared to standard approaches. Coupled with the near-linear scaling of resources, organizations can now train multi-billion-parameter models within hours instead of days. The collaboration between Databricks and MosaicML will make training and utilizing LLMs accessible at a fraction of the previous cost, reducing expenses from millions to thousands of dollars.
Databricks’ unified Data and AI platform, bolstered by MosaicML’s generative AI training capabilities, will provide a robust foundation to serve the largest organizations globally while remaining flexible enough to cater to a wide range of AI use cases.
Following the completion of the acquisition, the entire MosaicML team, including its renowned research division, is expected to join Databricks. MosaicML’s specialists in machine learning and neural networks have been at the forefront of groundbreaking AI research, focusing on improving model training efficiency. The team has contributed significantly to popular open-source foundation models, including the widely used MPT-30B, as well as developing the training algorithms powering MosaicML’s products.
Over time, MosaicML’s platform will be integrated, supported, and scaled within Databricks’ infrastructure, creating a seamless unified platform where customers can build, own, and secure their generative AI models. This collaboration will provide customers with greater flexibility in developing their models, training them using their unique data, and generating differentiating intellectual property for their businesses.
Naveen Rao, Co-Founder and CEO of MosaicML, expressed his enthusiasm for the acquisition, underscoring the shared belief in empowering individuals to build and train models infused with their own opinions and viewpoints. Rao sees the collaboration with Databricks as an opportunity to address the current generative AI wave and drive accessibility for many. Both Databricks and MosaicML share a similar mission, having transitioned from researchers to entrepreneurs, and are committed to continuing their journey together with the broader AI community.
You must log in to post a comment.