Trending post
MatX, a startup specializing in the design of chips for large language models (LLMs), has reportedly closed a Series A funding round worth about $80 million. This follows a seed round of $25 million obtained barely a year ago. This already speaks to renewed interest from investors in this company’s innovative approach toward AI hardware. The Series A round was led by Spark Capital, and sources close to the deal revealed that the company was valued at a pre-money valuation in the mid-$200 million range with a post-money valuation reaching the low $300 million.
Just two years old, MatX was founded by experienced AI chip designers Mike Gunter and Reiner Pope. Gunter, who has a background in working for Google on the development of Tensor Processing Units, or TPUs, and Pope, who developed AI software for the same group, are positioning the company to address the growing shortage of specialized chips required for AI workloads. Their mission is to design high-performance chips that will be able to handle more complex AI tasks, specifically focusing on large-scale models with billions of parameters.
MatX claims that their chips outperform in processing AI workloads of “at least” 7 billion activated parameters, ideal focus models at 20 billion or more. The company also points out that at a much lower price than other market leaders, including Nvidia’s GPUs, high performance results are delivered. The other differentiator of MatX’s technology is its advanced interconnect, a system designed to improve the communication pathways between AI chips; this is critical for scaling large clusters necessary to train expansive AI models.
The long-term vision of the startup is ambitious: MatX aims to make its processors up to 10 times more effective at training and delivering results for large language models compared to Nvidia’s GPUs. As such, the company has been working to present a cheaper alternative to expensive, in-demand chips that Nvidia has sold for years, dominating the AI hardware market. A boom in demand of AI chips has flooded the markets in recent years due to the rapid advances in AI, particularly in natural language processing and other large-scale AI tasks.
The Series A round marks an important milestone in MatX’s journey. This latest round follows a successful seed round the company raised last December, led by well-known angel investors in the AI space, including former GitHub Chief Executive Nat Friedman and Daniel Gross, who previously oversaw search and AI at Apple. Both Friedman and Gross have been vocal about supporting novel AI ventures, and they continue to work together, with Gross recently co-founding a new AI firm called Safe Superintelligence, along with former OpenAI chief scientist Ilya Sutskever.
The AI chip sector has emerged as a hot area of interest for investors, given the surging demand for specialized hardware for deep learning tasks and other applications with an AI connection. Over recent months, several other startups have experienced a valuation windfall. Groq, one such startup founded by former Google TPU engineer Jonathan Ross, saw its valuation spike from $1 billion to $2.8 billion at the start of this year, underscoring the significant promise of this space.
MatX has not issued a comment on the funding round announced at present. But the involvement of the significant funding from Spark Capital is a very positive statement on the potential the investors have in disrupting the AI hardware market. With clear focus on improving the performance of AI chips and reducing costs, MatX is establishing itself as a very serious player in the field of AI infrastructure, which is rapidly expanding.
As AI continues revolutionizing industries, competition between companies designing specialized hardware for LLMs and other AI models will only get more intense. MatX, led by its ambitious and ambitious goals and visionary leadership, fits that category, ready to challenge the natural order of dominance in the AI chip market held by Nvidia. If successful, the company might play a pivotal role in shaping the future of AI hardware development by producing powerful, affordable chips for use by researchers, developers, and businesses.
source: techcrunch
