Recently, MatX, an AI chip company founded by two former Google hardware engineers, announced that it has completed a $500 million Series B funding round. This round was jointly led by Jane Street, a renowned quantitative trading giant, and the investment fund Situational Awareness, founded by Leopold Aschenbrenner, a former OpenAI researcher.
MatX's founding team has an impressive background: CEO Reiner Pope was responsible for the AI software development of Google's TPU and played a key role in optimizing the efficiency of Google's PaLM model; CTO Mike Gunter was the chief designer of Google's TPU hardware and has nearly 30 years of experience in chip architecture.
Targeting NVIDIA: Aiming for a Tenfold Performance Leap
Differing from NVIDIA GPUs, which are designed to be versatile for gaming, graphics, and scientific computing, MatX takes an extremely specialized approach. Its core goal is to develop a processor optimized specifically for large language models (LLMs). The company claims that by removing redundant computing modules found in GPUs, its chips could be up to ten times faster than current NVIDIA GPUs when processing Transformer architecture models.
This "less is more" strategy aims to address the current cost bottleneck in AI computing power. This round of funding also included participants such as Marvell Technology, NFDG, Spark Capital, and the co-founders of Stripe, Collison brothers.
Official Shipment in 2027, Valuation Approaching $5 Billion
Although MatX has not yet publicly disclosed its latest valuation, its valuation exceeded $300 million during its Series A funding round in 2024. Referring to the market performance of its direct competitor Etched, which recently completed a similarly sized financing round with a valuation of $5 billion, the industry generally holds a positive view of MatX's market potential.
Currently, MatX has established a production cooperation relationship with TSMC and plans to use this new funding to advance the chip's prototyping and mass production, with an expected official shipment to global AI laboratories in 2027
