The largest acquisition in the AIGC field: Databricks bought MosaicML for US$1.3 billion, with 60 employees after only 2 years of establishment

Text: Juny Editor: VickyXiao

Image source: Generated by Unbounded AI tool

This morning Western time, the big data giant Databricks announced that it has signed a definitive agreement to acquire MosaicML, a San Francisco-based artificial intelligence startup, for $1.3 billion.

After the acquisition, MosaicML will become part of the Databricks Lakehouse platform. MosaicML's entire team and technology will be brought under the banner of Databricks, providing enterprises with a unified platform to manage data assets and be able to use their own proprietary data to build, own and protect Own generative AI models.

It is worth noting that **MosaicML is a very young generative AI company. It was established in San Francisco in 2021. It has only publicly disclosed one round of financing and has only 62 employees. In the last round of financing, its valuation was 220 million US dollars, that is to say, the valuation of the acquisition of MosaicML directly jumped 6 times. **

The deal is the largest acquisition announced in the generative AI field so far this year. Not long ago, cloud computing giant Snowflake just announced the acquisition of another generative AI company, Neeva. After a few months of investment frenzy, a massive corporate wave of acquisitions of generative AI startups appears to be underway.

**Who is MosaicML? **

MosaicML was founded by Naveen Rao, former head of artificial intelligence products at Intel, co-founder of Nervanas, and Hanlin Tang, senior director of Intel AI Labs.

The founders of MosaicML, Hanlin Tang (first from left), Naveen Rao (second from left), pictures from MosaicML

Naveen Rao graduated from Duke University with a major in computer science in 1997, and later obtained a doctorate in neuroscience from Brown University. He has long been committed to the study and development of artificial intelligence neural networks. He worked as a researcher on neuromorphic machines at Qualcomm and founded the artificial intelligence company Nervanas in 2014.

Perhaps many people are not familiar with Nervana s now, but in the field of deep learning and AI chips, Nervana can be said to have had the same limelight. **Nervana's goal is to improve the computing efficiency of deep learning through its cloud services and hardware products. In 2015, it launched Neon, a super-performance deep learning underlying framework, which became a hit in the industry, and then launched it in 2016. Nervana Cloud deep learning cloud platform, and Nervana Engine dedicated hardware accelerator. **

Nervana said that after running the Neon framework on the Nervana Engine chip in the Nervana Cloud, the combination can achieve 10 times higher performance than the NVIDIA Titan X. Nervana's strong performance also attracted the attention of the chip overlord Intel at the time, and Intel's first big move in the field of AI was to acquire Nervana for $400 million. **

Report on Intel’s acquisition of Nervana in 2016, picture taken from Vox

After the acquisition, Naveen Rao became the person in charge of Intel's artificial intelligence products. The Neon architecture and Nervana related products were also integrated into Intel's product line. The Nervana Engine was named the Crest series.

Since then, from 2017 to 2019, Intel has repeatedly announced the progress of Nervana Lake Crest, and has repeatedly announced related chip products including the Nervana NNP-T series. But while everyone was waiting for the mass production of this series of chips, in 2020, Intel suddenly announced that it would replace the original Nervana server-side AI acceleration chip with the Israeli company Habana series products that it later acquired for $2 billion. The reason analyzed by the industry at that time was that Habana's technology and design were more scalable, and it already had mass production delivery capabilities.

**After Intel decided to "abandon" Nervana, Naveen Rao and Hanlin Tang, former core employees of Nervana, also left Intel, and the two founded MosaicML separately. **According to the information of LinkedIn, Hanlin Tang should be a Chinese, who obtained a Ph.D. in biophysics from Harvard University, and is currently the CTO of MosaicML.

So, what is the main business of MosaicML after leaving Intel?

MosaicML is still committed to helping companies improve AI efficiency, but this time they no longer invest too much energy in hardware, but focus on generative AI. **To put it simply, MosaicML provides a platform that allows enterprises of all types to easily train and deploy AI models in a safe environment, and helps enterprises reduce the overhead of AI systems. **

Their product portfolio portfolio includes open source, commercially licensed MPT Foundation series models and MosaicML inference and training services, providing a series of tools for enterprises.

For example, MosaicML Explorer can help developers explore and understand the time, performance and cost between different cloud services and hardware options to simplify and evaluate implementation options. Launched MosaicML Composer, an open-source deep learning library that provides 20 methods for computer vision and natural language processing, including models, datasets, and benchmarks. Launched the MosaicML AI development platform, which provides cost-effective model deployment and customized training, while ensuring data security, enabling users to have ownership of the model, etc.

Picture from MosaicML official website

Aiming at enterprise services, Databricks also uses generative AI to make moves

Looking back at the founding team of MosaicML, it can be said that their business choices have always been one step ahead of the trend. Make AI chips when everyone is still waiting and watching, and take the lead in exploring the commercialization of generative AI in the trough of the AI industry.

**Relying on a strong technical team background and industry experience, MosaicML received US$37 million in financing from well-known venture capital DCVC, Lux Capital, Future Ventures and other investors shortly after its establishment, and the total financing has climbed to US$64 million since then. **It is understood that Databricks' acquisition of MosaicML is mainly due to the commercialization capabilities of its generative AI model at the enterprise end.

Naveen Rao, CEO of MosaicML, previously stated that since 2018, the complexity of artificial intelligence models using large amounts of data for "training" has risen sharply, and training a model now costs at least millions of dollars. Small and medium-sized enterprises generally cannot afford it.

After this acquisition, the joint product of **Databricks' Lakehouse platform and MosaicML technology will enable enterprises to use their own proprietary data to train and build generative AI models simply, quickly and at low cost, allowing users to have Custom AI model development can occur without control and ownership of the data. **

Picture from MosaicML official website

MosaicML's automated optimization of model training promises 2–7x faster training than traditional methods, and near-linear scalability of resources ensures that models with billions of parameters can be trained in hours, not days of the past. **According to Databricks, with the platform and technical support of Databricks and MosaicML, the cost of training and using LLMs for enterprises will be significantly reduced, and it is expected to drop to around several thousand dollars. **

It is worth noting that before MosaicML joined, Databricks had developed a 12 billion parameter language model called Dolly-2 based on EleutherAI's Pythia-12b, and with the addition of MosiacML, Databricks will provide Dolly-2 and MosaicML MPT two leading large language models.

“Every organization should be able to benefit from the AI revolution and have more control over how its data is used. Databricks and MosaicML have an incredible opportunity to democratize AI and make Lakehouse the powerhouse of build generation The best place for artificial intelligence," said Ali Ghodsi, co-founder and CEO of Databricks, in a press release.

60 employees are happy to mention the "big gift package", AIGC's merger and acquisition wave kicked off

The acquisition of MosaicML is currently the largest publicly disclosed transaction in the field of generative AI. The acquisition amount of up to 1.3 billion is undoubtedly a "big gift" for MosaicML, which currently has only 62 employees.

According to Levels.fyi, the average salary of a software engineer at MosaicML was between $750,000 and $850,000. Whether the employee options will be discounted in cash or converted to Databricks options is unclear, but with the acquisition, the entire MosaicML team will join Databricks.

Image via Levels.fyi

In the current AI boom, the merger and acquisition of generative AI startups by large companies may have just begun.

Not long ago, Snowflake, the world's leading cloud data management company, announced that it had acquired Neeva, a generative AI search startup founded by two former Google employees, with an undisclosed transaction amount. Neeva's main business is to use generative AI for search, and it mainly focuses on the field of enterprise search. **After joining Snowflake, Neeva will help service enterprise customers use AI to quickly search and analyze data points, data assets, and gain data insights.

From the successive acquisitions of Snowflake and Databricks, we can see that large technology companies are gradually moving from independent research and development and strategic investment to the stage of mergers and acquisitions for generative AI technology. This also provides more opportunities for some start-up generative AI companies. If nothing else, we will see more similar acquisitions in the second half of this year.

In addition, whether it is the application direction of these two large acquisitions or the recent unicorns such as Cohere and Anthropic, their business focus is mainly on the enterprise-level application of generative AI technology. **

After more than half a year of excitement on the consumer side, generative AI technology has begun to aggressively march towards enterprise users.