14nm, 16nm, 1nm - what gives?
The talk of 14nm, 16nm, or even 1nm, is interesting...
Marketing can blur the lines between for example, routing being on say a 16nm basis, and the main structures that make up a transistor being on a larger one, say 20 or 22nm.
The issue of whether there is enough demand to produce enough chips, and spreading across 14nm and 16nm to cope with this perhaps a 'red herring' - with TSMC and others moving from 300mm to 450mm wafers, significantly more devices can be made per wafer, even more so if they are made at 14nm rather than say, 22nm. So the demand issue may be less negative than it seems - the true demand is there, and growing perhaps, but can be met by a smaller number of fabs... which would really reduce operating costs...
The physics of Moore's Law running out at around 5nm may be a non-sequitur in that the financials become the dominating issue around 7nm. The cost of the equipment, fab, and then mask sets for 7nm devices are now so expensive when spread across the devices being built, that it is MORE expensive to build at 7nm say, than at 10-12nm. The development costs at the 7nm node are huge in comparison to at the 20nm node, and 22nm has proven a challenge for Intel, so...
We are probably seeing the practical end of 50+ years of shrinking... and the falsehood of lower costs year after year (lower costs are not the same as lower prices!).
