MatX Raises $500 Million to Challenge Nvidia in LLM Chips

Overview

• The Series B was led by Jane Street and Situational Awareness, with Marvell, NFDG, Spark Capital, and Stripe co-founders Patrick and John Collison participating.
• Founded by former Google TPU engineers Reiner Pope and Mike Gunter, the startup is building accelerators tailored to large language models.
• MatX plans to finalize its chip design in 2026, manufacture with TSMC, and begin shipping in 2027, using the funding to lock in scarce memory and fab capacity.
• The company asserts roughly 10× performance gains over Nvidia GPUs and says internal tests top Nvidia’s upcoming Rubin Ultra on compute density, though these claims are unverified.
• MatX employs around 100 people and is hiring aggressively, with a go-to-market strategy focused on a small set of leading AI labs.