Yeah, I didn't want to spend more than 50k for local inference stack. I can amortize it in my taxes so it's not a big deal but beyond it would start eating into my other allocations. I might still get M5 Ultra if it pops up and benchmarks look good, possibly selling M3 Ultra.