Image Credits:Naveen Rao

Naveen Rao, who was formerly the head of artificial intelligence at Databricks, is in talks to raise $1 billion at a $5 billion valuation for a new startup, called Unconventional, Inc., that’s building a new type of computer, according to four people familiar with discussions. 

Andreessen Horowitz has agreed to lead the investment, and Lightspeed and Lux Capital are also participating in the deal, two of the sources said, although none of those VC firms responded to our request for comment. Databricks is also said to be investing in Rao’s new company, Bloomberg reported last month. 

Rao has already raised hundreds of millions and plans to begin building the startup without waiting for the full $1 billion round to close, our sources said. He plans to raise the rest of the capital in installments, a funding approach commonly referred to as a “tranched” round. 

Rao declined to comment, although he did publicly acknowledge the new startup on X last week, confirming its name and describing its hoped-for product as “rethinking the foundations of a computer to build a new substrate for intelligence that is as efficient as biology. Brain Scale Efficiency without the biological baggage!”

Databricks acquired Rao’s previous startup, MosaicML in 2023, for $1.3 billion. Founded by Rao in 2020 with a focus on training and deploying large AI models, MosaicML raised $33.7 million from investors, including Lux Capital, DCVC, Playground Global, and Samsung Next, according to PitchBook. Prior to MosaicML, Rao co-founded the machine learning platform Nervana Systems, which Intel Corp. acquired in 2016 for reportedly more than $400 million. 

Rao, who spent over two years as VP of AI at Databricks (now valued at $100 billion and generating $4 billion in ARR), left the company last month to focus on his new startup, according to his LinkedIn profile and earlier reporting by Bloomberg.

Our sources say Rao’s vision will essentially compete with Nvidia by designing a novel AI machine that includes both custom silicon chips and server infrastructure.