Apple "A20" SoC Linked to TSMC "N3P" Process, AI Aspect Reportedly Improved with Advanced Packaging Tech

[T0@st]

Over a year ago, industry watchdogs posited that Apple was patiently waiting in line at the front of TSMC's 2 Nanometer GAA "VVIP queue." The securing of cutting-edge manufacturing processes seems to be a consistent priority for the Cupertino, California-headquartered fabless chip designer. Current generation Apple chipsets—at best—utilize TSMC 3 nm (N3E) wafers. Up until very recently, many insiders believed that the projected late 2026 launch of A20 SoC-powered iPhone 18 smartphones would signal a transition to the Taiwanese foundry's advanced 2 nm (N2) node process. Officially, TSMC has roadmapped the start of 2 nm mass production around the second half of 2025.

According to Jeff Pu—a Hong Kong-based analyst at GF Securities—the speculated A20 (2026) chipset could stick with N3P. Leaks suggest that aspects of Apple's next in line "A19" and "A19 Pro" mobile SoCs could be produced via a 3 nm TSMC process. MacRumors has picked up on additional inside track whispers; about Apple M5 processors (for next-gen iPad Pro models) being based on N3P—"likely due to increased wafer costs." Pu reckons that Apple's engineering team has provisioned a major generational improvement with A20's AI capabilities, courtesy of TSMC's Chip on Wafer on Substrate (CoWoS) packaging technology. This significant upgrade is touted to tighten integration between the chip's processor, unified memory, and Neural Engine segments. Revised insider forecasts have positioned A21 chip designs as natural candidates for a shift into 2 nm GAA territories.



View at TechPowerUp Main Site | Source

 

[kondamin]

Probably don’t feel like getting an other n3b generation where they have to toss most of the wafer

 

[bonehead123]

Sounds ok I guess, but while we're busy talkin about 2 & 3nm, I'm sure both TSMC & the fruity boys are hard at work on the next biggest (er smallest) thing, like sub-nm chips, hahahahahaha

 

[alwayssts]

This is a potential bummer, if true, as another year of 3nm for Apple means another year of those fabs not being occupied by things, well, mostly non-Apple.

I have a feeling Apple is playing hardball for pricing, as they have before, and will use 2nm in 2026 (A20), but IDK. It makes absolutely zero sense to have those 2nm fabs operational if nobody is making chips there.
Apple attempts to negotiate, but really is not in the position to do so (given they demand process leadership as part of their business model). That's why these stories are floated but generally eventually inaccurate.

TBQH though, this does line up with people whom feared that 3nm would not go truly wide until it was manufactured in the US (in ~2027), which I was kind of hoping would not be the case. We'll see.

Even if so, it's not to say we won't get 3nm GPUs/CPUs before that point, only that it may be slightly later, have lower supply at first, and/or come at a larger premium until then...which sucks.

If it is true, then the true value (and perhaps performance due to needing every possible chip to yield) from 3nm products may not be seen until a refresh generation from AMD/nVIDIA in ~2028 (on n3p).

In some (if not many) ways this does make sense, as the continued cost of 2nm are likely due to TSMC seeing the road-ahead is difficult/expensive, and this buys them time (a couple years) to make a better process.

The continued high costs will likely continue until Samsung etc can compete on that approximate level (N2P and/or BSPD), while the margins until then could fund TSMC's fulfillment of something better.

 

[kondamin]

This is a potential bummer, if true, as another year of 3nm for Apple means another year of those fabs not being occupied by things, well, mostly non-Apple.

I have a feeling Apple is playing hardball for pricing, as they have before, and will use 2nm in 2026 (A20), but IDK. It makes absolutely zero sense to have those 2nm fabs operational if nobody is making chips there.
Apple attempts to negotiate, but really is not in the position to do so (given they demand process leadership as part of their business model). That's why these stories are floated but generally eventually inaccurate.

TBQH though, this does line up with people whom feared that 3nm would not go truly wide until it was manufactured in the US (in ~2027), which I was kind of hoping would not be the case. We'll see.

Even if so, it's not to say we won't get 3nm GPUs/CPUs before that point, only that it may be slightly later, have lower supply at first, and/or come at a larger premium until then...which sucks.

If it is true, then the true value (and perhaps performance due to needing every possible chip to yield) from 3nm products may not be seen until a refresh generation from AMD/nVIDIA in ~2028 (on n3p).

In some (if not many) ways this does make sense, as the continued cost of 2nm are likely due to TSMC seeing the road-ahead is difficult/expensive, and this buys them time (a couple years) to make a better process.

The continued high costs will likely continue until Samsung etc can compete on that approximate level (N2P and/or BSPD), while the margins until then could fund TSMC's fulfillment of something better.

Some ai chips design house or crypto bro will foot the bill if apple doesnt.as long as the market doesn’t correct it self there is a lot of stupid money that can be thrown at it.

remember Altman throw a trillion at making fabs for ai?

 

[Wirko]

remember Altman throw a trillion at making fabs for ai?

He better start making nucular reaktors first, one reaktor every morning until he spends one half of that trillion. Or is he planning to make chips that produce electricity instead of consuming it?

 

[kondamin]

He better start making nucular reaktors first, one reaktor every morning until he spends one half of that trillion. Or is he planning to make chips that produce electricity instead of consuming it?

Humanity needs them with or without AI.
Electricity needs to be a lot cheaper and constant if they want us to do away with natural gas for heating
Never mind the amount of power needed to desalinate water.

 

[alwayssts]

Some ai chips design house or crypto bro will foot the bill if apple doesnt.as long as the market doesn’t correct it self there is a lot of stupid money that can be thrown at it.

remember Altman throw a trillion at making fabs for ai?

While this is possible, I wonder if the process is currently fit for something like an AI chip of any practical use. Remember that the A-series is (comparatively-speaking) relatively simple silicon. Generally <100mm2.
That's typically what early nodes and their yields are designed towards. Small size due to early yields, low-voltage, low-leakage, high density, while exceeding the clock potential of the former process at it's maximum.

While it could be used for something like mining chips, I wonder if those companies would be willing to pay that level of premium. While it's possible, it's likely doubtful on any large scale IMHO.

They could throw as many 2/4/8-bit or whatever ops on a chip they could get to yield, but I don't know if that's really practical versus buying something larger/more-capable clocked lower on the former process.

 

[kondamin]

While this is possible, I wonder if the process is currently fit for something like an AI chip of any practical use. Remember that the A-series is (comparatively-speaking) relatively simple silicon. Generally <100mm2.
That's typically what early nodes and their yields are designed towards. Small size due to early yields, low-voltage, low-leakage, high density, while exceeding the clock potential of the former process at it's maximum.

While it could be used for something like mining chips, I wonder if those companies would be willing to pay that level of premium. While it's possible, it's likely doubtful on any large scale IMHO.

They could throw as many 2/4/8-bit or whatever ops on a chip they could get to yield, but I don't know if that's really practical versus buying something larger/more-capable clocked lower on the former process.

Well there is that wafer scale design house (cerebras) that uses software to ignore the bad parts of the wafer/chip

I doubt there is much holding others back from doing something similar