Local AI Hardware: Break Even in 2.6 Years?

In short, running a $3,299 GMKtek EVO-X2 (Ryzen AI Max+395 with 198 GB) 24/7 with the Gemma 4 26B-A4B model, being as generous as possible, only saves you $1,279.07/year in inference costs. (120 t/s for output tokens at $0.34/M.)

So, that's how you get a break even at 2.6 years...