Matx

MatX is an AI semiconductor company headquartered in Mountain View, California, that designs custom, high-throughput chips and hardware architectures purpose-built for large language models (LLMs) and frontier AI workloads. Founded by former Google TPU engineers, the company develops silicon optimized for machine learning training, reinforcement le... arning, and inference tasks including both prefill and decode stages. Its flagship MatX One chip is engineered to deliver industry-leading FLOPS per unit area, low-latency decode, long-context support, and high-scale interconnect capable of supporting clusters with hundreds of thousands of chips.Read More