I'm really surprised AMD hasn't pushed much harder into the space given their aggressive targeting of the data center with Epyc. I too use mostly Macs during the day and any ML projects are always relegated to Nvidia machines in my rack / amazon.