AI hardware company Etched has raised $800 million across four undisclosed funding rounds to develop low-latency AI inference clusters. The company is integrating chip, rack, software, and manufacturing design to enhance large-model inference throughput, latency, and energy efficiency. Etched's first A0 chip, produced on TSMC's N4P process, is currently being validated with customers. Etched has secured over $1 billion in orders for its inference clusters and is working with a team of over 400 experts from companies like NVIDIA, TPU, and Broadcom. Strategic investors, including VentureTech Alliance, are supporting Etched's efforts to improve compute energy efficiency at the rack and cluster levels.