I think any price spikes in DRAM will be short lived simply because the tool set needed to make both DRAM and NAND flash is so similar Samsung, Hynix, and Micron can easily switch capacity back and forth between the two. So as prices for NAND parts soften more capacity will be converted over to DRAM production.
I know in the past that it was not an "easy" transition to move from producing DRAM to flash and vice versa. Micron at one time mentioned that it took about 6 months for the transition and a little more time to achieve peak yields again. There was also some additional investment as well. Moreover, you are not producing product on the fab line during the transition which will have a negative impact on sales.
Yes, the rising DRAM ASPs could have a negative effect on the PC market. The increasing DRAM prices typically don't increase the price of the PC (the PC suppliers take lower margins) but they make it more difficult to lower the price of the PC. Currently, many PC manufacturers are looking to lower the price of their ultrabooks to try to spur increased market acceptance and rising DRAM prices make it more difficult to do this.
There really is the "elasticity of demand" effect in the DRAM market. History has shown that when DRAM ASPs are collapsing, unit volumes oftentimes surge. However, when DRAM ASPs climb, DRAM unit shipments are oftentimes weak. This is likely to hold true in 2013 as DRAM ASPs are forecast to jump 40% this year (with the DRAM market up 28%) but DRAM unit volume shipments are forecast to decline 8%.
?? I would have thought if the PC price went up 40% due to DRAM prices it would have an impact, but by the looks of things that will only change 1-2%, hardly worth a mention given that life's inflation rate is 2-3%. In relative terms a PC might have actually gone down in price.
Isn't part of the problem that DRAM has essentially hit the feature size wall for charge storage devices - the same wall everyone panics about Flash reaching, but DRAM reached it without fuss in the past year or two since no-one was much paying attention? After all, who needed more DRAM?
Now we have a steady state product (well, going sideways, low power mobile, vertical stacking, etc.) and when consumption ticks up there is no slack in the system.
Personally I think the long scaling law in DRAM where we needed ever more RAM for bigger applications was artificially driven by bad compilers os OS bloating. There are very few applications that need to have the large data sets that 3GB of RAM supports, the only ones I can think of are 3D modeling, CGI and a few games.
All of the rest are just because the compilers & OS are very cavalier with memory. It seems now the standard is that if one routing in a library is used the whole 100 MB of code in the library is added to the program.
If you look at the executable with a hex editor you often see MB's of text strings stored in there that have nothing to do with the applciation.
The other thing that seems to have happened is that this bloat slows a lot of programs down seemingly requiring a faster CPU too.