Databricks Announces Acquisition of Generative AI Platform MosaicML for $1.3 Bn

Databricks, has announced its definitive agreement to acquire MosaicML, a generative AI platform, in a transaction valued at approximately $1.3 billion. The acquisition aims to make generative AI accessible to organisations, allowing them to "build, own and secure generative AI models with their own data." The CEO of Databricks, Ali Ghodsi, emphasised the goal of democratising AI and making the "Lakehouse the best place to build generative AI and LLMs."

MosaicML was co founded by Naveen Rao and Hanlin Tang in San Franscico in 2021. Rao expressed enthusiasm about the acquisition, stating that joining forces with Databricks will help make their vision of empowering everyone to build and train their own AI models a reality. The proposed acquisition is subject to customary closing conditions, including regulatory clearances.

MosaicML is recognized for its state-of-the-art MPT large language models, which have been downloaded over 3.3 million times. It has demonstrated how organisations can quickly build and train their own models using their data in a cost-effective manner. Notable customers such as AI2, Generally Intelligent, Hippocratic AI, Replit, and Scatter Labs leverage MosaicML for various generative AI use cases.

The Databricks Lakehouse Platform, combined with MosaicML's technology, will provide customers with a "simple, fast way to retain control, security, and ownership over their valuable data without high costs." MosaicML claims that its automatic optimization of model training enables 2x-7x faster training compared to standard approaches. Furthermore, the scalability of resources allows for training multi-billion-parameter models in hours, rather than days. The collaboration between Databricks and MosaicML is expected to significantly reduce the cost of training and using LLMs, making it accessible to a wider range of organizations.

The entire MosaicML team, including its research team, is expected to join Databricks after the transaction closes. MosaicML's platform will be integrated and scaled over time to provide a unified platform where customers can "build, own and secure their generative AI models." This collaboration aims to offer organisations greater choice in building their own models, training them with their unique data, and creating intellectual property specific to their businesses. The acquisition aligns with the shared vision of Databricks and MosaicML to democratise AI and give organisations more control over their data. 

MosaicML, had earlier, in the month of May launched a new service called MosaicML Inference to compete with OpenAI and other similar platforms on price. The offering aims to assist enterprise customers in training and deploying generative AI-powered solutions. As large models like ChatGPT and Stable Diffusion gain popularity, organisations are increasingly interested in leveraging these technologies.

MosaicML highlights that while existing solutions are suitable for those unconcerned with cost and willing to share data, building and hosting their own models can be a more cost-effective option for organizations with strict data privacy requirements. The company claims to make its technology accessible to all organizations at a significantly lower price, up to 15 times cheaper than its competitors.

CEO Naveen Rao mentioned that MosaicML has optimized its offerings to achieve cost savings. By offering the service at a much lower price, the company still maintains profitability. MosaicML has developed foundational models to support its new service, similar to the technology used in chatbot-enabled products by Microsoft and Google.

While MosaicML, like its competitors, provides its technology for rent, it differentiates itself by also offering its code to customers. This allows customers to run the code on their own hardware, ensuring the confidentiality of their data from MosaicML. This approach appeals to corporate customers who prioritize data privacy, as the value of an AI system is heavily dependent on the training data used.