[avivweinstein]

I work at Groq. We arent using GPUs at all. This is a novel hardware architecture of ours that allows this high throughput and latency. Nothing sketchy about it.