Nah, a GTX 1060 6GB for $100 + any 10-year old i5/i7 is still surprisingly capable for messing about with ML. It's not fast but it gets the job done. Also, getting free compute for messing around in, say, Google Cloud is still pretty easy. If you get to the point where those 2 options become a bottleneck, you're probably informed enough to find work in the field and afford something nicer.