Intel Could Manufacture Apple's Next-Generation A20 SoC for iPhone

[AleksandarK]

Apple is reportedly considering diversifying its chip manufacturing strategy with a new silicon manufacturer: Intel. While the upcoming iPhone 17 series, expected next year, will likely feature A19 chips produced by TSMC, a recent rumor from Chinese leaker Fixed Focus Digital hints at a potential switch to Intel for the A20 chipsets powering the 2026 iPhone 18 series. The A18 and A18 Pro chipsets debuted alongside the iPhone 16 series in September 2024, manufactured using TSMC's N3E node. Apple's A19 chips are expected to upgrade to TSMC's N3P node. According to the source, Apple is seeking an Intel 20A node. However, since the A20 node is canceled in favor of 18A, Apple could be an Intel Foundry customer for either 18A or 14A nodes.

Despite the buzz, skepticism persists. Intel has historically struggled with process node transitions and even outsourced production of its Arrow Lake CPUs to TSMC, raising questions about its readiness to deliver on Apple's demands. On the other hand, alternative reports suggest Apple might stick with TSMC's yet-unnamed 2 nm node for the A20, maintaining continuity in its supply chain. As the iPhone 18 series remains two years away, much can change. For now, we are left speculating whether this rumored collaboration with Intel represents a new chapter in Apple's chipset innovation or just a rumor with little substance. If the US government mandates more domestic production, chip designers could be looking at some of the more local manufacturing options, like Intel does on US soil. That could force Apple, NVIDIA, AMD, and Qualcomm to look into Intel's offerings.



View at TechPowerUp Main Site | Source

 

[tetrapak]

Yeah, errrm no, never again They can't even produce their own sh**, for sure apple won't trust them their top tier CPU.

 

[john_]

They will have to provide a manufacturing node that is MORE efficient than TSMC's. Intel can't even provide a working manufacturing node at the moment.
I could believe a rumor about a second grade SOC that will be going to cheaper models. There Apple could accept a less than excellent product.

 

[RimbowFish]

is this a joke

 

[Darmok N Jalad]

The irony would be if Apple managed to succeed on an Intel node before Intel does. It would certainly tell us where the problem is. Aren’t Meteor Lake yields rumored to not be that great even with TSMCs help?

 

[usiname]

Even if we assume that Intel can reach acceptable yields, how their nodes that are at least 3 years behind will compete with the TSMC's one? Their best node Intel 3 just started to be used and only for low clocked server CPUs. It is shrink to the Intel 4 which is 123 mt/mm^2, little better than N6 of TSMC. Its not surprise it was not used for Arrow Lake or Lunar Lake, even Intel know this node is garbage, by the time A20 is manufactured, TSMC will be ready with their N2P node which probably will be close to 300 mt/mm^2, while Intel's A18 will be 180 mt/mm^2 at most

 

[sudothelinuxwizard]

@AleksandarK shouldn't A20 in the fourth line be 20A? That aside, interesting!

 

[mechtech]

Will they use the 14nm++++++ line??

 

[Darmok N Jalad]

Even if we assume that Intel can reach acceptable yields, how their nodes that are at least 3 years behind will compete with the TSMC's one? Their best node Intel 3 just started to be used and only for low clocked server CPUs. It is shrink to the Intel 4 which is 123 mt/mm^2, little better than N6 of TSMC. Its not surprise it was not used for Arrow Lake or Lunar Lake, even Intel know this node is garbage, by the time A20 is manufactured, TSMC will be ready with their N2P node which probably will be close to 300 mt/mm^2, while Intel's A18 will be 180 mt/mm^2 at most

Maybe Apple’s approach is different enough to matter. They aren’t pushing high clocks or high power. The process might not be an issue for what they intend to produce. Intel has been pushing their desktop designs in both clocks and power. Their nodes may not be failing to produce working chips, but rather failing to perform at desired targets that are simply unsustainable. Raptor Lake is failing due to Intel pushing their design too far. Arrow Lake is failing to compete because the high thermals and clocks strategy was played out 2-3 generations ago. Intel can’t compete against its own previous generation that was pushed too hard. Almost seems like an architectural issue that the competition isn’t having.

 

[Onasi]

Apple basically financed TSMCs node advances. That’s the reason why they are a uniquely VIP customer and get first dibs every time, no questions asked. It makes absolutely no sense for them to go for an unproven, potentially highly inferior node from a foundry service that hasn’t really established itself as a viable partner for their style of chip designs. Unless the US government forces their hand it doesn’t seem likely.

 

[Redwoodz]

Only way this happens or even makes sense is if they actually bought Intel first.

 

[R0H1T]

They'll manufacture the chips for the most profitable lineup Apple has? I think there's day dreams & then there's this

Their nodes may not be failing to produce working chips, but rather failing to perform at desired targets that are simply unsustainable.

And they can't afford any let ups with SD elite or Dimensity 9400 especially QC with their Oryon cores ~

Oppo Find X8 Pro review

The Find X8 Pro is running Color OS 15 on top of an Android 15 core. The software update policy that Oppo has in store for the Find X8 Pro includes 5 OS...

www.gsmarena.com

 

[AnotherReader]

Even if we assume that Intel can reach acceptable yields, how their nodes that are at least 3 years behind will compete with the TSMC's one? Their best node Intel 3 just started to be used and only for low clocked server CPUs. It is shrink to the Intel 4 which is 123 mt/mm^2, little better than N6 of TSMC. Its not surprise it was not used for Arrow Lake or Lunar Lake, even Intel know this node is garbage, by the time A20 is manufactured, TSMC will be ready with their N2P node which probably will be close to 300 mt/mm^2, while Intel's A18 will be 180 mt/mm^2 at most

I don't think that transistor density metric is quite accurate. First of all, you should be using the same product ported to two different processes. In the absence of that, we know that Golden Cove in Intel 7 has similar SRAM density to Zen 3 built using TSMC's N7. This suggests that N7 and Intel 7 are rather comparable in density. Intel 3 is slightly behind TSMC's N5 in SRAM density, but should have similar logic density. With power via, also known as backside power delivery, 18A should have higher density than N3. Of course, all this only matters if Intel ships on time.

 

[londiste]

Even if we assume that Intel can reach acceptable yields, how their nodes that are at least 3 years behind will compete with the TSMC's one? Their best node Intel 3 just started to be used and only for low clocked server CPUs. It is shrink to the Intel 4 which is 123 mt/mm^2, little better than N6 of TSMC. Its not surprise it was not used for Arrow Lake or Lunar Lake, even Intel know this node is garbage, by the time A20 is manufactured, TSMC will be ready with their N2P node which probably will be close to 300 mt/mm^2, while Intel's A18 will be 180 mt/mm^2 at most

Are you comparing TSMC's HP library densities or HD one?

 

[CyberPomPom]

Wait, didn't Intel skip A20 as a node for their customers in order to release A18 faster?

 

[phanbuey]

Apple will get a massive tax break to do it - it will happen even if Intel isn’t the best

 

[usiname]

I don't think that transistor density metric is quite accurate. First of all, you should be using the same product ported to two different processes. In the absence of that, we know that Golden Cove in Intel 7 has similar SRAM density to Zen 3 built using TSMC's N7. This suggests that N7 and Intel 7 are rather comparable in density. Intel 3 is slightly behind TSMC's N5 in SRAM density, but should have similar logic density. With power via, also known as backside power delivery, 18A should have higher density than N3. Of course, all this only matters if Intel ships on time.

Intel 7 is similar to n7, just exactly what suggest the density of both nodes:
N7 - 91.2
Intel 7 - 100.76
Intel 3 is also slightly behind N5:
Intel 3 (same density as 4) - 123.4
N5 - 138.2
But A18 is build on N3E - 216 and intel's 18A definitely won't be any close to this even if it was existing node today, let alone after 2 years when A20 will be on N2 or N2P.
Their 20A node has 15% higher perf/watt than Intel 3 so this node is waaaaay behind even the current TSMC nodes

Maybe Apple’s approach is different enough to matter. They aren’t pushing high clocks or high power. The process might not be an issue for what they intend to produce. Intel has been pushing their desktop designs in both clocks and power. Their nodes may not be failing to produce working chips, but rather failing to perform at desired targets that are simply unsustainable. Raptor Lake is failing due to Intel pushing their design too far. Arrow Lake is failing to compete because the high thermals and clocks strategy was played out 2-3 generations ago. Intel can’t compete against its own previous generation that was pushed too hard. Almost seems like an architectural issue that the competition isn’t having.

The problem for apple is that they will have two versions of their SOC, with one inferior with 15-20% worse battery life and terrible termals. They won't allow that

 

[LabRat 891]

If Apple is competing with themselves and everyone else for TSMC's fabrication, this doesn't seem too crazy.
I know 'porting' uArchs between nodes/processes isn't simple, but it seems worst-case would be revised silicon made @ TSMC (after senescing their older SoCs/room opens up at TSMC.)

Apple will get a massive tax break to do it - it will happen even if Intel isn’t the best

'Didn't think about that.
Both, the outgoing and incoming admins were-are pro U.S. chips manufacturing. No matter who'the decision is made under, financially incentivizing domestic production is plausible.

-Though, I'm not sure if Intel has a domestic fab built yet for their latest process node.

 

[kondamin]

Someone sent apple an email about intel 18a and the receiver at apple opened the email and sent them an opening notification.

with that apple looked in to intel foundry business with enough truth to it to make up a rumour.

 

[Prima.Vera]

Will they use the 14nm++++++ line??

No, the 14nm+++++++ line.

 

[Readlight]

The phone, who they can block from anywhere.

 

[JohH]

I won't believe it until I see it.
It is more likely if Apple has dealings with Intel it's for ancillary products like chips for headphones or watches.

 

[Nanochip]

If this is true (I’m skeptical because I don’t see why Apple would switch from tsmc), then Intel had better move heaven and earth to get it done.

 

[hsew]

I won't believe it until I see it.
It is more likely if Apple has dealings with Intel it's for ancillary products like chips for headphones or watches.

I could also see Apple shifting older A and M chip prod to Intel to save $, only using TSMC for their newest lineup…

Unless, of course, hell has actually frozen over and Intel has regained their long-lost node lead.

 

[Daven]

I won't believe it until I see it.
It is more likely if Apple has dealings with Intel it's for ancillary products like chips for headphones or watches.

Or chips to run the bidet toilets at Apple HQ.

Wait, didn't Intel skip A20 as a node for their customers in order to release A18 faster?

A20 is the Apple SoC. 20A is the Intel fab node which is cancelled.