Skip to content
Companies

MatX

AI accelerator chips built for large language models

Raised to date

$527.9M

of $557.9M offered across 1 Form D filing

Still open
$30M
Investors
23
Last filed
Mar 10, 2026

About

MatX is an artificial intelligence hardware startup founded in 2022 by former engineers from Google's Tensor Processing Unit (TPU) team. The company designs and develops high-throughput AI accelerator chips and systems specifically optimized for large language models (LLMs). It aims to address the market's chip shortages and provide a more powerful, cost-effective alternative to general-purpose GPUs for both training and inference tasks.

The company's key differentiators include its founding team's expertise from Google's TPU project and its focus on advanced interconnect technology, which it claims will deliver 10x better performance for the largest AI models by enabling superior scaling across large clusters.

Funding

  • EquityFiled Mar 10, 2026Form D
    $527.9M/ $557.9M

Product

AI Accelerator Chips and Systems

AI Accelerator Hardware

MatX's core offering consists of specialized AI accelerator chips engineered for transformer-based large language models. The hardware is designed to support both the training and inference phases for models starting at 7 billion parameters, with an optimal performance target for models of 20 billion parameters or more.

  • High-throughput processing optimized for LLMs
  • Advanced interconnects for superior scaling to large clusters
  • Designed to offer high performance at a lower cost than general-purpose GPUs

Built for

Frontier AI research and development labs

People on file