MatX Raises $500M Series B to Build TSMC‑made AI Chips Aiming to Rival Nvidia

MatX, an AI chip startup founded by former Google TPU engineers, closed a $500 million Series B on Feb 24, 2026. Led by Jane Street and Situational Awareness, the round included Marvell Technology, NFDG, Spark Capital and Stripe co‑founders Patrick and John Collison. The capital follows a $100 million Series A and will fund chip development and manufacturing at TSMC with planned initial shipments in 2027. MatX aims to deliver processors that offer up to 10x training performance for large language models versus current Nvidia GPUs, leveraging founders’ TPU software‑hardware experience and full‑stack co‑design. Technical details are limited, but analysts expect approaches such as tensor‑optimized cores, advanced memory hierarchies, chiplet layouts and novel numerical formats. The raise signals growing investor appetite for GPU alternatives amid Nvidia’s ~80% market share, supply constraints and rising training costs. Strategic investors and partnerships (notably TSMC and Marvell) may ease manufacturing and go‑to‑market hurdles, but MatX still faces the challenge of building compilers, libraries and customer adoption against entrenched GPU ecosystems. Traders should note the funding’s potential to accelerate competition in AI hardware, possibly affecting firms exposed to AI compute demand, chipmakers, and cloud GPU prices as alternative accelerators approach production.