Groq calls their inference chips “Language Processing Unit”: https://groq.com/blog/the-groq-lpu-explained