MatX vs Nvidia

MatX, founded by ex-Google TPU engineers, designs chips optimized specifically for large language models, with first silicon due in 2027.

Nvidia designs GPUs and AI accelerators that power the vast majority of frontier model training and inference workloads.