“Etched was able to get its silicon running inference in a rack within 40 days of receiving it, compared to 10 months for another well-known AI chip company.”