Monday, May 11th 2026

Apple Strikes Deal with Intel Foundry Services for Chip Production in the U.S.

Apple and Intel Foundry Services have reached a preliminary agreement to onboard Apple as a foundry customer for manufacturing of Apple Silicon on its latest foundry nodes, according to a Wall Street Journal report. The move could see Apple end its near-total reliance on TSMC for production of its chips. Apple is looking to leverage the Intel 18A foundry node that offers comparable transistor densities and electric characteristics as TSMC's 2 nm-class nodes. Intel manufactures chips on the Intel 18A node in the state of Oregon, and is building new facilities in Ohio that will cater to future nodes. The company also has chip packaging facilities in states such as Arizona and Nevada. The move could improve Apple's contribution to the U.S. Government's efforts to onshore manufacturing of chips and other strategic commodities.
Source: Wall Street Journal
Add your own comment

14 Comments on Apple Strikes Deal with Intel Foundry Services for Chip Production in the U.S.

#1
kondamin
I wouldn't call 1/3 less dense similar
Posted on Reply
#2
vbq7qK68eyYAH4iR
Mere Mortal
I wonder what chip will get down first. I wouldn't be any of the large M based dies. Apple seems too cautious, so I'm not even sure Apple would put their A series on Intel. That leaves their many smaller series of chips. Maybe the Ultrawide band chip, as it is currently on TSMC 7nm. The modem is 4nm. the UWB seems like a low risk option. If Intel fails, they could stick with the 7nm another generation.
Posted on Reply
#3
Benedict ChenLi or Games4
kondaminI wouldn't call 1/3 less dense similar
Ehduehheb

Strategic commodity is this important?!

I know a lot of tech developed from war and military,

Dang

Computer chips strategic commodity that's mah focus.

1/3 less dense, uhhhh so which one is that?
Posted on Reply
#4
alawadhi3000
kondaminI wouldn't call 1/3 less dense similar
Whats your source of that info?
Posted on Reply
#6
jabbadap
alawadhi3000Whats your source of that info?
It just blind high density transistor count comparison, which does not take count of other differences of said processing nodes(source).
Posted on Reply
#7
Wirko
kondaminwww.tomshardware.com/tech-industry/intels-18a-and-tsmcs-n2-process-nodes-compared-intel-is-faster-but-tsmc-is-denser
That's the PPA, power-performance-area, that we tech freaks love to discuss. Then there's the other PPA, price-payment-availability, that we don't discuss because we don't get that data. The best we get are some analysts' informed guesses. Terms of payment, prepayment, financing - how much and when? High volume availability - how much and when? What are the long term commitments by each party? Customers base their decisions also on that.
(And there's that long 10% shadow of "force majeure" over Intel. INTC must grow. Period.)
Posted on Reply
#8
kondamin
jabbadapIt just blind high density transistor count comparison, which does not take count of other differences of said processing nodes(source).
Yes that means logic density is even higher as sram doesn't scale that good.
If you want more, let’s say shaders, and it seems most of us do. More transistors per area is a good thing.

Even if that comes at a cost of lower clocks.
WirkoThat's the PPA, power-performance-area, that we tech freaks love to discuss. Then there's the other PPA, price-payment-availability, that we don't discuss because we don't get that data. The best we get are some analysts' informed guesses. Terms of payment, prepayment, financing - how much and when? High volume availability - how much and when? What are the long term commitments by each party? Customers base their decisions also on that.
(And there's that long 10% shadow of "force majeure" over Intel. INTC must grow. Period.)
apple can't have the iphone 20 be worse than the 19 because the soc isn't as performant or the battery drains to quickly.
It matters even for the penny pinchers
Posted on Reply
#9
XeonsNOpterons
kondaminYes that means logic density is even higher as sram doesn't scale that good.
If you want more, let’s say shaders, and it seems most of us do. More transistors per area is a good thing.

Even if that comes at a cost of lower clocks.


apple can't have the iphone 20 be worse than the 19 because the soc isn't as performant or the battery drains to quickly.
It matters even for the penny pinchers
We shall see the true densities once they get to market on the big chips. Seems like Intel's Diamond Rapids-AP will feature 256-cores just like AMD's Venice, yet DMR-AP Panther Cove-X cores are likely much more complex as they are expected to have an IPC lead of 20% over Zen 6. Since DMR-AP is on 18A and Venice is on TSMC 2, we shall see then when both are launched their die areas. I have a feeling the TSMC numbers are not quite what they claim, but again, only real chips will show it. At the very least, they will be very comparable in overall characteristics, which is great as Intel needs to succeed with nodes too. Having just TSMC is not only a national security risk for the US, but the entire world as the rest of the world relies on them too.

And no, the clock speed does matter too. It is a combination of the transistor density as well as achievable clock speed range and power per watt across that range.
Posted on Reply
#10
Fouquin
20A/18A took the spot of 3nm on Intel's old foundry road map, and we already know 18A is right behind N3E on density metrics. It is however quite fast, and Apple could rely on it for fast switching logic that doesn't need an especially massive amount of SRAM. Apple already has 50%+ of TSMC's preliminary A14 capacity so nothing particularly next-gen or cutting edge is going to be on Intel 18A. Most likely second sourcing for legacy SoCs still produced at other TSMC facilities, plus they'll want to do a lot of trials with Intel's packaging. Apple co-designed and built TSMC's InFO and WMCM packaging and assembly over a decade of partnership. That won't just magically translate over.
Posted on Reply
#12
kondamin
alawadhi3000Thats an educated guess/estimation based on an unrealistic best case scenario, it won't translate to the real world.
isn't that true for both and pretty correct in both of the right and wrong ness?
Posted on Reply
#13
alawadhi3000
kondaminisn't that true for both and pretty correct in both of the right and wrong ness?
Not in this case, a modern chip isn't all logic with HD libraries, its a mixed chip.
TSMC real world density advantage versus Samsung and Intel in the 2nm class should be, on average "only" ~10%-15%.
Posted on Reply
#14
XeonsNOpterons
Fouquin20A/18A took the spot of 3nm on Intel's old foundry road map, and we already know 18A is right behind N3E on density metrics. It is however quite fast, and Apple could rely on it for fast switching logic that doesn't need an especially massive amount of SRAM. Apple already has 50%+ of TSMC's preliminary A14 capacity so nothing particularly next-gen or cutting edge is going to be on Intel 18A. Most likely second sourcing for legacy SoCs still produced at other TSMC facilities, plus they'll want to do a lot of trials with Intel's packaging. Apple co-designed and built TSMC's InFO and WMCM packaging and assembly over a decade of partnership. That won't just magically translate over.
Completely off topic, but I recognize the art of your thumbnail and I know the artist. I love her artwork!
Posted on Reply
Add your own comment
Nov 17th, 2026 13:14 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts