Naveen Rao, the AI head at Databricks, is leaving the company to launch a new venture focused on developing a novel type of computer to address the rising costs of AI computing. Databricks has confirmed that Rao will transition to an advisory role and plans to invest in his new company. The amount of Databricks' investment has not been disclosed. Rao has also reportedly begun early talks with other potential investors.

Rao's departure comes at a time when Databricks is riding high on a $100 billion valuation, following a recent $1 billion funding round. Databricks' CEO, Ali Ghodsi, has publicly announced his support for Rao's new venture.

The Rising Costs of AI Computing

Rao's new company aims to tackle the increasing expenses associated with AI computing. AI training compute needs have been doubling every 3.4 months since 2012, a significant acceleration compared to the historical rate of every two years between 1959 and 2012. This surge in compute usage could lead to large AI experiments costing as much as $200 billion within 5 to 6 years if current trends persist. While traditional computing power per dollar historically improved rapidly, that rate has slowed significantly in recent years, creating a market opportunity for specialized computing solutions.

Rao's Background and Databricks' AI Strategy

Naveen Rao is a serial entrepreneur with a strong track record. He previously sold MosaicML, a data and AI analytics startup, to Databricks in 2023 for $1.3 billion. Before that, he co-founded Nervana Systems, a machine intelligence platform acquired by Intel in 2016 for approximately $350 million. At Databricks, Rao played a crucial role in shaping the company's generative AI tools, including the development of open-source models like MPT and DBRX.

Databricks has been actively expanding its AI capabilities. At the Data + AI Summit in June 2025, Databricks introduced Agent Bricks, a product for building production AI agents optimized for enterprise data, and Lakebase, a new category of operational databases built on open-source Postgres and optimized for AI agents. Databricks will use the new capital to accelerate its AI strategy, expand Agent Bricks, launch the new Lakebase category, and fuel global growth.

Implications and the Future of Databricks

Rao's exit raises questions about the sustainability of Databricks' ambitious AI roadmap, particularly its emphasis on democratizing access to large language models through its data lakehouse platform. His expertise in efficient AI training, honed from his time at Intel and Nervana Systems, helped Databricks compete against rivals like Snowflake and Amazon Web Services. Without Rao, there may be gaps in pushing boundaries on efficient inference and training at scale. However, analysts suggest that Databricks could mitigate risks by doubling down on its open-source ethos and fostering community-driven innovation.

Databricks' AI strategy hinges on its ability to integrate acquisitions like MosaicML seamlessly and advance projects such as Delta Lake and Unity Catalog. The company's recent funding is earmarked for accelerating AI initiatives, including enhanced model serving and governance tools.

Next-Generation Computing

Next-generation computing companies are developing a new age of intelligent computing. These companies are committed to creating cutting-edge solutions that expand computing's capabilities. Artificial intelligence and blockchain technology are two of the main topics that next generation computer companies are concentrating on. Companies specializing in next-generation computers also focus on forging new, more natural, intuitive user interfaces.