This is the beginning of the commoditization of computer processing. Right now, the only thing that gives you a lead is what you can have fabricated... Apple an

This is the beginning of the commoditization of computer processing. Right now, the only thing that gives you a lead is what you can have fabricated... Apple and AMD using TSMCs fabs have proven this to be extremely true taking the top CPU spots from Intel the past ~3 years. While we will probably see a small spike in "custom dogshit CPUs" I don't think it will be long lived (maybe ~5 years that suck starting around the end of this decade). It certainly won't be worse than the current variety of processors in phones or random other small electronics. Not to mention there's little incentive to use hardware that has poor software support. Eventually (~25 years), the fabs will go from billions to millions to maybe a few tens of thousands of dollars. Purchasing a CPU will (eventually in like 40-50 years) simply be deciding how much power, heat, and physical space you have to budget... The reduced cost from mastering processes will mean companies are competing on price; which, is typically good for consumers and awful for the businesses involved. If anything, as we really push physics to its limits, the largest companies will likely be forced into cooperation to make any reasonable gains via something like RISC-V because of diminishing returns. I believe "Machine Learning" will be what leads the way for eventual commoditization of processor fabrication. Pretty much what Relativity has done with 3D printing rockets: using ML to correct, predict, and/or even utilize, what would now-a-days be considered physical flaws to produce extremely perfect parts... but instead of rocket parts, it'll be lithography and CPU fabs one day.