How AI describes Etched

Etched is building frontier inference clusters, a new category of AI hardware that co-designs chips, racks, software, and manufacturing to run frontier models with best-in-class throughput, latency, and power efficiency for prefill and decode workloads. It introduces Low Voltage Inference and Cluster Scale Memory to achieve high FLOPs density and SRAM-like memory speeds, enabling trillion-parameter MoEs with low latency across a scalable interconnect. The company is vertically integrated, shipping its first racks, opening a Taiwan factory, pursuing over $1B in customer contracts, and sourcing a team from top semiconductor and AI firms to scale to gigawatt-level inference.