FPG, I'm with you that this forum is likely underestimating what TSMC will deliver on 20nm. Qualcomm will ramp their 20nm modems fast this year (>100 Mu), and Apple will have a 20nm A8 in the second half, as well (>100 Mu).
nVidia's Maxwell is also 20nm, though that is a discrete GPU, not a mobile SOC (and it will have 10s of millions of units, if not >100 Mu). So that's at least several hundred million units coming from the TSMC 20nm process this year.
I think most of the mobile SOCs will come either late this year (too late to matter), or in early 2015 (in time for CES/MWC'15). But I think by the time they ramp, it will be very fast, and TSMC will support a broad product deployment.
By early 2015 (when Intel plans Cherry Trail, the Bay Trail follow-on), competitors will be shipping 20nm in pretty high volume, IMO. The opportunity for Intel to lead here will be if 14nm density improves super-linearly (as they claim), and the graphics architecture overhaul actually succeeds to moving faster than at least half a dozen competitive architectures, who are also planning to advance their architectures.
The chorus on Intel's impending mobile dominance is getting muted to my ears.
- Wait for Intel's 32nm SOC process...
- Wait for Bay Trail...
- Wait for Merrifield...
- Wait for Cherry Trail...
- Wait for Broxton...
It's kind of silly. Qualcomm, Samsung, nVidia, and Apple... none of these are standing still. Intel needs a dramatic breakthrough to leap (and stay) ahead. Yet for the past few years, they've been chronically behind, and making silly priority calls, like dual-boot systems before they even have Android native systems shipping on the latest platforms.
And even if they earn some MSS this year (by selling Bay Trail deeply discounted to cover the BOM shortfalls), I don't see how Intel moves to profitability later. How do they raise prices, if mobile SOC's are not going to give up their business easily, and the PC business is still losing volumes.
It's a business model shift that I don't think people here understand (the analysts do). And shifting further going forward, into sub-dollar SOC's that get integrated into baby clothes, will only accelerate the shift to a low margin business - let alone when we can expect the "Internet of things" to ramp to a billion units, where it could even conceivably replace some of Intel's lost business. I haven't seen anything from Intel in terms of business model shifts to accommodate the "new" new world.
I sold half my shares already, on the hope that they can at least solve the problems that they have historically been good at solving - on the process and architecture side. I'm still not convinced they know how to solve the problems on the business side.