(2) When do you guys think Qualcomm will adopt chiplets? Intel and AMD are aggressively pursuing it. Apple does too (in a limited sense) for their gargantuan M Ultra chips.
Qualcomm will not adopt chiplets for PCs anytime soon. Chiplets are still the future in some sense, and we’ll probably see variants become more suitable for very low power operation for sure, e.g. some GPU+IO tile and a CPU tile or whatever on a very low power interposer might happen eventually.
Apple only does chiplets for their desktop Ultra chips. The M3 Pro/Max do not use chiplets even if in theory they’d benefit cost wise due to their size — probably in part because they don’t think there’s anything suitable yet for the hit to power and they can afford the die size costs and sell directly to customers.
Another thing at that, RE: die size, is that Qualcomm is selling a mainstream premium part designed for power and energy efficiency. They aren’t selling an M3 Pro/Max with a huge GPU because they wouldn’t have the bus width for it (much less the use currently what with games on Windows or compute for Adreno, honestly what would you do with it?) and that market segment is too limited for Qualcomm/AMD/Intel right now save for games, whereas Apple can grab that from creators etc.
With LPDDR6 they (AMD/Intel/QC) could throw a bigger GPU onboard without using a bigger bus, but by that point they’ll have another logic shrink and speed boost with N3E/P if not N2 (smaller shrink there though) so it sort of levels out.