You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Ming-Chi Kuo on Apple-Intel partnership, with TSMC lurking in the back
▌Below are my latest industry checks on the Apple-Intel partnership, which help interpret it at a deeper level:
1. Apple has kicked off low-end/legacy iPhone, iPad, and Mac processors at Intel on the 18A-P series (using Foveros packaging).
2. The order mix is roughly 80% iPhone, mirroring Apple's end-device sales mix.
3. Apple's wafer plans at Intel reflect the technology lifecycle of the 18A-P series: small-scale testing in 2026, ramp in 2027, continued growth in 2028, and decline in 2029.
4. Apple is also actively evaluating Intel's other advanced-node technologies.
5. Intel's mass production timeline and shipment scale remain unclear, and assemblers/EMS have yet to see any shipment schedules.
6. Intel's 2027 yield target is to first stabilize at 50–60% or higher.
7. Even if Intel's initial shipments go smoothly, TSMC will still retain over 90% of supply share.
8. Internally, sentiment at Intel toward Apple’s orders appears mixed.
9. Apple began discussions with Intel well before TSMC's advanced-node capacity became tight.
10. Apple recognizes that TSMC's resources will continue tilting toward AI.
Building on these findings, this analysis examines the three parties’ strategic responses to the structural shift in advanced-node manufacturing. TSMC stands out here, as the industry leader and a passive party in this event. What it appears able to do is limited, but the situation deserves deeper reading.
▌Apple is systematically cultivating Intel to become a long-term key supplier:
..
Didn't Ian Cutress suggest several months ago that in his opinion, due to the increased heat from BSPD, he doubted or said it wasn't a good idea to use 18A or 14A fabbed AP/SoCs for phones? Now they're talking about fabbing the A21 or whatever A series SoC on 18A-P. Who's wrong here? Or what's the story now?
Didn't Ian Cutress suggest several months ago that in his opinion, due to the increased heat from BSPD, he doubted or said it wasn't a good idea to use 18A or 14A fabbed AP/SoCs for phones? Now they're talking about fabbing the A21 or whatever A series SoC on 18A-P. Who's wrong here? Or what's the story now?
Well there is 50% thermal conductivity improvement promised with BSPDN but I would not take what Ming Chi Kuo says at face value his yield number Is inaccurate without the die size.
Didn't Ian Cutress suggest several months ago that in his opinion, due to the increased heat from BSPD, he doubted or said it wasn't a good idea to use 18A or 14A as a phone AP/SoC? Now they're talking about fabbing the A21 or whatever A series SoC on 18A-P. Who's wrong here? Or what's the story now?
Apple truly drove the modern TSMC (and the Asian electronics supply chain in general).
Can they do the same with Intel and domestic supply chain? Not sure the economic and political fundamentals of the US can support it… but the US needs to desperately try.
Didn't Ian Cutress suggest several months ago that in his opinion, due to the increased heat from BSPD, he doubted or said it wasn't a good idea to use 18A or 14A fabbed AP/SoCs for phones? Now they're talking about fabbing the A21 or whatever A series SoC on 18A-P. Who's wrong here? Or what's the story now?
nothing in here says A21. it is legacy products mostly targetted at specific markets.
Also TSMC does not package phone processors. a separate company packages them with DRAM, correct?
This seems like a reasonable checkout of Intel foundry by Apple.
I have been a supplier to Apple from multiple companies. I would not recommend it to anyone I care about LOL. It provides some unique advantages but I would not list profitbility as one of them.
nothing in here says A21. it is legacy products mostly targetted at specific markets.
Also TSMC does not package phone processors. a separate company packages them with DRAM, correct?
This seems like a reasonable checkout of Intel foundry by Apple.
I have been a supplier to Apple from multiple companies. I would not recommend it to anyone I care about LOL. It provides some unique advantages but I would not list profitbility as one of them.
My comment about A21 was just my own speculation based on the time when the Apple silicon from Intel could possibly appear. In terms of release dates, a low-end (non-pro) Intel fabbed iPhone chip would probably be the A21 or A22. Apple doesn't design new chips for legacy products (I actually can't think of a circumstance where they've done that; new products just use previously designed chips from TSMC), if I'm understanding you correctly. Though, there's alway a first time for everything. The two options for an Intel-based iPhone AP would be (1) monolithic chip (high certainty) or (2) a packaged-chiplet based (highly unlikely; but, it has the advantages of better yield/mix+match for individual chiplets at different nodes; but would be a first for a phone AP). Opinions?
My comment about A21 was just my own speculation based on the time when the Apple silicon from Intel could possibly appear. In terms of release dates, a low-end (non-pro) Intel fabbed iPhone chip would probably be the A21 or A22. Apple doesn't design new chips for legacy products (I actually can't think of a circumstance where they've done that; new products just use previously designed chips from TSMC), if I'm understanding you correctly. The two options for an Intel-based iPhone AP would be (1) monolithic chip (high certainty) or (2) a packaged-chiplet based (highly unlikely; but, it has the advantages of better yield/mix+match for individual chiplets at different nodes; but would be a first for a phone AP). Opinions?
I think you highlight all the correct challenges and options. I can't imagince Apple or Intel confirming details unless there was a Political reason to do so .
What I do know is that is will be a small portion of what Apple makes as a test case to start. What I also know is that it is VERY difficult to make money selling to Apple. TSMC has a special relationship with Apple on the leading edge so they are different.... and TSMC is obviously much better that any other manufacturer.
I could see Apple bifurcating their chips by node like the way AMD has done (e.g., TSMC's N4 vs. N3). In Apple's case, it could be Intel's 18A-P (non-pro) vs. TSMC's N2P (pro). 18A-P could also be used to fab the silicon they need for the Apple Watch and modems since they chips for those devices are usually a node or two behind.
What I do know is that is will be a small portion of what Apple makes as a test case to start. What I also know is that it is VERY difficult to make money selling to Apple. TSMC has a special relationship with Apple on the leading edge so they are different.... and TSMC is obviously much better that any other manufacturer.
