Nvidia challenger AI chip startup MatX raised $500M

MatX, a startup founded by former Google TPU engineers Reiner Pope and Mike Gunter, has secured a massive $500 million Series B funding round. The investment, led by Jane Street and Situational Awareness (a fund by ex-OpenAI researcher Leopold Aschenbrenner), signals strong confidence in the startup's ambitious goal: to build AI processors that are 10 times more effective than Nvidia's GPUs for training large language models (LLMs). This round, which also included investors like Marvell Technology and Stripe's co-founders, comes over a year after a $100 million Series A and positions MatX as a serious, well-funded challenger in the high-stakes AI hardware race dominated by Nvidia.

The company's technical pedigree is a key asset, with Pope having led AI software for Google's TPUs and Gunter serving as a lead TPU hardware designer. The new capital will be directed toward producing its first chips via TSMC, the world's leading semiconductor foundry, with a target to begin shipping in 2027. While the latest valuation wasn't disclosed, the round's size mirrors competitor Etched's recent $500 million raise at a $5 billion valuation, highlighting the intense investor appetite for alternatives to Nvidia. The success of MatX and similar startups could reshape the economics and availability of AI compute, potentially lowering costs and increasing innovation speed for companies building next-generation AI models.