MatX, an AI chip business formed by Google alumni, has raised $300 million or more in Series A funding

Three sources claim that less than a year after obtaining $25 million for its seed round, MatX, a business that designs processors that support huge language models, has raised a Series A of about $80 million.

An individual familiar examined the transaction told TechCrunch that Spark Capital spearheaded the investment and valued the business at a pre-money valuation in the mid-$200 million area and a post-money valuation in the low $300 million range.

Mike Gunter, a former Google employee who worked on the creation of the tech giant’s AI processors known as Tensor Processing Units (TPUs), and Reiner Pope, who also came from Google’s TPU team and built AI software, co-founded MatX two years ago.

Pope and Gunter want to alleviate the scarcity of processors made to manage AI workloads. According to them, AI workloads with “at least” 7 billion and “ideally” 20 billion or more activated parameters are the sweet spot for their processors. Additionally, MatX’s website boasts that its chips offer superior performance at more reasonable costs. According to the startup, MatX’s sophisticated interconnect—also known as the communication channels that AI chips utilize to exchange data—makes its chips especially adept at scaling to big clusters.

The business wants its processors to be ten times more effective than Nvidia’s GPUs at training LLMs and producing results, the two told Bloomberg. MatX was seeking to raise $75 million to $100 million for the round, according to a report published by The Information last month.

High-profile AI angel investors Daniel Gross, who formerly oversaw search and AI at Apple after Apple acquired his firm, Cue, in 2013, and Nat Friedman, the former CEO of GitHub, led the startup’s seed round, which was revealed last December. Friedman and Gross often work together as angel investors. Together with Ilya Sutskever, the former chief scientist at OpenAI, Gross has now co-founded a new AI business called Safe Superintelligence.

The AI boom and the soaring demand for Nvidia’s processors have heightened investor interest in chip design companies. Jonathan Ross, a former TPU engineer, launched the semiconductor business Groq, which saw its valuation nearly treble to $2.8 billion in August from its initial $1 billion valuation in April.

Leave a Reply

Your email address will not be published. Required fields are marked *