Success at 20nm is the whole shooting match atm....
that will bring into play being able to address the data bottleneck that AI is causing between the storage cache (L1,L2 and L3) and DRAM by specifically mitigating the deficiencies in dram function
this is a major performance limitation that results from the disparity between the speed of data processing needed and the speed of the memory access ....they work against each other at present causing delays and major inefficiencies as the data flow increases - seems 4DS 20nm will provide a promising solution
AI itself?
total rubbish atm in my opinion but as it searches for its real use and applications, data will just grow exponentially and those processing deficiencies increase accordingly
name one compelling use application for AI ? ....but it will arrive eventually
https://themarket.ch/english/chinas-massive-build-out-of-semiconductor-manufacturing-capacity-is-scary-ld.1189720nm success is the real game imo as it addresses that niche application - ie processing inefficiencies that exist between DRAM and storage cache / NAND etc...
follow on question is what value is that niche application?
there are many other considerations that come into play once it works at 20nm .....till then they are of no value to the conversation
just one view of many
glah
(20min delay)