"I just don't see the market turning on a dime any more, and I'm less confident that it's as simple as Intel just leveraging their manufacturing and design prowess. It might take years, or it might never happen."
[I think that you are ignoring the economic situation that will be extremely critical in the next few years. Bottom line is that the economics of the ARM foundries appear ready to crash and burn.
TSMC's dividends have exceeded their free cash flow for all of the last four years. This is a situation that cannot be perpetuated. Taken as a whole, the ARM foundries are investing a level of CapEx that as a percentage of sales cannot be supported in anything other than the very short term. They are having to make huge investments in an attempt to get to 20nm and then FinFET. Literally company-crushing levels of investments. A big problem in and of itself but those investments aren't getting them any improvements in economics. They aren't even staying flat.
Intel on the other hand is spending much, much less as a percentage of sales, can easily manage it's CapEx needs and is getting improved economics at each new node and is managing to maintain margins at the same time. It's doing this by taking leadership in density.
I don't see how TSMC even gets the half-step to FinFET much less to 10nm without changing its business model or eliminating its dividends.
In short, it's too little, too late and there is nothing the ARM world can do to close the gap. Intel is going to continue to pull away while the ARM foundries move slower and slower with their cash constraints.
I've heard people suggest that Apple or Qualcomm will simply jump in and throw money at the foundries but if that were seriously suggested you would see Apple and Qualcomm shareholders turning over furniture and lighting cars on fire. They adopted the fabless model specifically to get rid of these costs.
ARM's fabrication is a train wreck and in the next few years it is going to be obvious to everyone. The fabless model has sucked all of the funding for R&D and building fabs out of the foundries. This is why the number of state-of-the-art fabs has shrunk consistently and now is really down to Samsung and TSMC, with TSMC likely to drop to tier 2 status in the near future. There isn't the time, money or technology to reverse this trend. Intel can accelerate this trend by taking the top paying customers away from TSMC.
I don't see any way that the economics don't play out decidedly in Intel's favor in the next few years. From here on out I think it is as simple as Intel just leveraging their manufacturing and design prowess, their superior economics and their huge capacity.]
AI
