As we been saying - Monolithic 3D is finally practical, so lets start to scale up, as scale down is no longer easy and full of benefits.
And it looks that the first vendors to reach this conclusion are the NAND NV vendors, as had been presented across this IEDM.
The problem for EUV is it has only one application space and that window is shrinking. In addition, the higher energy associated with its higher resolution wavelength produced some fundamental new effects, e.g., ionization, shot noise, etc., which will require more time to be grasped.
On the other hand, 193 nm is more familiar, and the temptation to merely increase patterning steps is there. Although such increases will be more complex and costly, a detailed calculation, factoring in number of layers, volume, performance, power, etc. is needed to determine the overall cost impact to the particular product. There won't be a unique "right" answer unfortunately.
But as you implied, leakage gets worse, and interconnect performance will also get worse. So a breakout from this trend is needed, soon.
What is the performance/leakage penalty at a given feature size for a given available wavelength?
If EUV allows feature shrink without performance/leakage penalty compared to UV used now for comperable cost EUV wins hands down.
Chasing moores law min feature size only and not performance/leakage metrics is not a solution
The facts are in process shrink below 1/10th of wavelength will require multiple patterning.
where increased cost of EUV equipment break even with double, triple patterning costs still being played out with rapid development in the EUV space.
If they would just hire me it could go much faster.
GlobalFoundries already gave up on EUV it seems:
“10nm will be optical,” he said. “We have evidence that we can do 7nm with immersion.”