Can Groq Really Take On Nvidia?

Groq LPU based systems provide very low latencies for LLM inference processing. But only for smaller models.