Industry Analyst Walks Back Claim about Apple A20 SoC Using N3P, Repredicts TSMC 2 nm

[T0@st]

Earlier in the week, Apple specialist press outlets picked up on a noted industry analyst's technological forecast for a future iPhone processor design. Jeff Pu—of GF Industries, Hong Kong—predicted that the next-generation A20 SoC would be produced via a TSMC 3 nm (N3P) nodes process. Despite rumors of Apple gaining front row seats at the "2 nm ballgame," the partnership between fabless chip designer and foundry could potentially revisit already covered ground. The A19 chipset was previously linked to N3P (by insiders), with Pu expressing the belief that A20 would utilize the same fundamental lithographic underpinnings; albeit enhanced with TSMC's Chip on Wafer on Substrate (CoWoS) packaging technology (for AI improvements).

This morning, MacRumors followed up on their initial news article—they reported that "wires were crossed" at GF Industries, regarding projections for the (2026) iPhone 18 generation. The publication received direct feedback from the man of the hour: "Jeff Pu (lead Apple analyst) has since clarified that he believes the A20 chip will be manufactured with the N2 process, so the information about the chip using the N3P process should be disregarded. Earlier reports had said the A20 chip would be 2 nm, so rumors align again. This is ultimately good news, as it means the A20 chip should have more substantial performance and power efficiency improvements over the A19 chip." Cutting-edge smartphone processor enthusiasts expressed much disappointment when A20 was (regressively) linked to N3P; the latest revisement should instill some joy. According to industry moles, TSMC is making good progress with its cutting-edge 2 nm node process—mass production is expected to start at some point within the second half of 2025.



View at TechPowerUp Main Site | Source

 

[Daven]

I think the analyst just didn’t know how to count:

A17 Pro 2023 3 nm
A18 2024 3 nm
A19 2025 3 nm
A20 2026 something bad would have to happen for another 3 nm after three years

 

[phints]

Crazy to think almost the whole industry waits for TSMC 2N now to see progress. Otherwise we will just keep getting more TDP increase/nothing generations like Zen 5, RTX 50, etc.

 

[TumbleGeorge]

Crazy to think almost the whole industry waits for TSMC 2N now to see progress. Otherwise we will just keep getting more TDP increase/nothing generations like Zen 5, RTX 50, etc.

Today's engineers, the all-electronic Gen Z and beyond, are incapable of inventing or truly improving chip architecture. It's not just about pushing more gigahertz and transistor counts and shrinking the size. The devices we're constantly staring at have eaten away at our intelligence to the point where we can't think outside the box, literally.

 

[alwayssts]

I think the analyst just didn’t know how to count:

A17 Pro 2023 3 nm
A18 2024 3 nm
A19 2025 3 nm
A20 2026 something bad would have to happen for another 3 nm after three years

It's simply that Apple uses every major process varient, and only for a year.
And TSMC makes each process varient because they can't make something better until after that point. So from the point we knew there was a N3B, N3E, N3P, we knew Apple would be on 3nm for three years.
And that 2nm likely wouldn't be truly ready for mass production and/or yields (even if hidden under high costs to the point nobody would use it) until that point.

Similarly, we can predict with almost-certainty Apple will use N2, N2P, and A16 (or N2P w/ BSPD if you wish) over the next three years, and TSMC should have something better by 2030.

On a similar note, we know since same fabs are used for the similar processes, as Apple moves their designs to a better process node, and hence different fabs, then companies like AMD/nVIDIA will use them.
For general consumer products; obviously there are outliers like Turin etc (that those companies use the smaller processes earlier for either small pipecleaner chips and/or extremely high-margin products).
And too some others (like Intel's use of N3B after Apple started on N3E), but generally-speaking this is how it goes (due to wafer allocation and cost) AFAIK. It's been generally-true for a very long time.

So, likely, one can assume AMD/nVIDIA will likely use N3P in 2028, and in 2030 will use N2P when Apple transitions fabs to whatever is after A16 (2nm BSPD). Also, those companies perhaps A16 in 2032.
Then in 2034, Apple will transition to whatever is two full process node from now, and AMD/nVIDIA (etc) will use that future process.

Obviously I don't know how many variant nodes that future process could have, or if they are even nodes. Maybe they are COWOS/BSPD improvements sold as nodes (etc), but they will likely be labeled as such.
And I don't know there is three, there might be more, but you get the point. They are general generations that allow transitions like that to occur.

I remain curious if AMD will ever take the initiative to port some of their designs to a smaller process ahead of the curve (beyond something like Turin) given some of their chiplets are very, very small.

For instance, imagine they make a GPU chiplet and that chiplet is only ~50mm2, or just large enough for a 64-bit memory controller (with minimal cache). Even if yields aren't great, it might give them a large advantage in certain designs to transition to that process node quickly even if more expensive. Obviously that would depend on volume and/or allocation to other companies (like Apple), but it's possible.
This could be something that could truly help in the competitive atmosphere, although they can't be too quick to do it either given costs and the need to stretch out their product improvements as well.

We've heard rumors of early transitions to 3nm designs for certain AMD products (that didn't ultimately happen) and the almost-certainty of MI-series chips/COWOS (stacking) doing it, so it's possible in the future.

This is why I've been watching Samsung. If they can get something decently yielding in volume (>N3P/~N2P) before TSMC's transition to something better than A16 for Apple.
Perhaps a company like AMD might use it (2028/2030), especially if cost-competitive.
If even in certain products and/or use-cases. Obviously a lot of variables there dependent upon what Samsung is able to accomplish, their further roadmap, and/or how entrenched certain designs are at TSMC.

But those aspects about the only exciting things to contemplate outside the explained and generally-expected cadence, hence why I find their potential kind of interesting, and champion the competition.

 

[AggressiveCombo]

Crazy to think almost the whole industry waits for TSMC 2N now to see progress. Otherwise we will just keep getting more TDP increase/nothing generations like Zen 5, RTX 50, etc.

Zen 5 is literally the opposite of a TDP increase, and I don't think it's a nothing generation, there's actual generational improvements in places outside of gaming

 

[Fatalfury]

WHETHER TSMC stays at 3nm for another 5 years or not...

there's some time for samsung and intel to atleast try to catchup...

hope intel 18A sees the light of the day.

 

[alwayssts]

WHETHER TSMC stays at 3nm for another 5 years or not...

there's some time for samsung and intel to atleast try to catchup...

hope intel 18A sees the light of the day.

They won't, but most mass-produced stuff probably will until we get the grand fight with a little better. Much stuff maybe longer than that, given the current (and even perhaps future) relative expense of TSMC 2nm.
1.28x increase is not woth 50% more cost to most, and it gets worse from there. Almost nobody wants to use BSPD because it's incredibly expensive for little added benefit. I think that's Samsung's window TBH.
Maybe they come in with BSPD, and it's not as great as TSMC or maybe even Intel, but perhaps similar-enough performance (especially to N2P w/o BSPD or Intel 18A) and lower cost.

It truly does depend on if Samsung/Intel (maybe even Rapidus and/or others targeting similar markets over time) become viable (performance/volume/price) alternatives, and when. Earliest probably 2028.

I hate the node names because is there a difference between N3P, Samsung 2nm, and 18A? I mean, there is...but they're not *that* different in reality. Different relative strengths as always, but pretty close.
Is transitioning from one to the other really a giant leap from starting out at N3E? I would argue it's probably not, and still call it a similar-enough node that you could call them all '3nm'.
Maybe N2P, Samsung 2nm with BSPD, and whatever comes after 18A actually 2nm (and that might be the real fight in ~2030 for things most of us care about). Obviously mobile/high-margin will use this stuff earlier.

I hope they are competitive and available ASAP as well. I'd love if there was greater competition by 2028, if not 2030. There might be, and there might not be. I'm watching the same yield updates you are.