Yes. We have a quad MI300A server and run several inference models on it. For $107k it has saved us so much money on tokens already and it's a heck of a lot faster than cloud services.