I'm sure tons of companies would, but that's not what I meant.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Secures Half of TSMC's 2nm Production Capacity for iPhone 18
- Thread starter MacRumors
- Start date
- Sort by reaction score
Digitimes is questionable at times, but they didn't put it into the present tense. They said it 'will go into production in 4th quarter'.
Well, the Apple chips are not some special "one of a kind" unicorns with magic mojo, they are just optimized original ARM architecture processors. Nearly every ARM64 based Linux distribution runs fine on Apple Silicone without modification - go figure. 😎
[edit] Most probably Windows ARM64 would also run fine on AS, but surely not wanted.
Last edited:
No, next comes zero nm and then the big breakthrough -1 nm.
Well maybe not. But the point is that at some point in the next few years, shrinking feature size will no longer be possible. A silicon atom is 0.2 nm. They can't go smaller than single atoms
It has been a long road. I remember when getting below 1um was a big deal, and now we well soon be a 1nm. A thousand times smaller. But there is clearly a limit coming.
There is no upside for TSMC to participate in a 'blocking' strategy for the rest of their customers. Before Hauwei gotten banned they used to get new process nodes about the same time as Apple did. Mediatek shipped on 4nm before Apple did. AMD got the first. 2nm wafer ( go back to post 25 in this thread ).
Performance of local AI on iPhones/iPads is in an "overpowered" state. If going to put more 'smart' into a 'smartphone' then being smarter is likely going to require more 'horse power' . They may have plateaued in legacy apps but looking backwards most of the time is the kind of thing that got Intel into trouble.
Also if Apple is going to shift to more chiplet design philosophy 2nm could enable making smaller, focused function chips that assemble into a package that isn't much bigger. If Apple is going to not layer the RAM chips on top of the Apple Silicon chip (and place them side by side) then a smaller Apple Silicon die would limited the 2D footprint bloat.
(the volume might go down , but 2D footprint would have to go up. ). If a smaller chip then can get more dies out of a single wafer which could offset the cost increase. ( for last several iterations iPhone die sizes have gotten larger and larger. At A18 they split the regular die (90 mm^2) off from the 'Pro' ( 105mm^2 ) . As the wafer costs very substantively increase keeping the die size the same means the die keeps getting more expensive to make.
There is a reoccurring theme that macrumors front page article tend to spin about Apple being "special" and access to TSMC process nodes are hindered/blocked because Apple is so 'great'. That is mostly smoke. It churns comments in and ads in these threads , but it isn't how TSMC is running their business.
There was a time when Apple had more cash burning a hole in their pockets than several of the other competitors to wafer reservations ( e.g., AMD one foot in the grave. ). That isn't the case anymore. The AI mainia of throwing billions at hardware vendors is trickling down to several players that are more than willing to throw a chunk of that at TSMC. TSMC isn't going to turn them down or allow Apple to 'cock block' them without punishment. Also doesn't make any sense for Apple to pay penalty fees just to block others. ( that will just buy them more enemies that will complain to TSMC and governments which will just bring more problems over the long term. )
Last edited:
Not necessarily. Financial quarter is whatever the company adopts. The US Gub'ment, for example, Oct 1st is the start of the 1st quarter which would mean July - September is the 4th quarter. Apple's 4th quarter ends on the last Saturday of September, so for Apple, now is the 4th fiscal quarter.
Apple has ordered almost half of TSMC's initial 2nm production capacity for the iPhone 18 as the pure foundry begins mass production of its next-generation chip process.
According to DigiTimes, TSMC's 2nm process went into production in the fourth quarter of 2025 as planned, with the foundry setting substrate pricing at a premium $30,000 per unit. Despite the steep costs, chip manufacturers are said to be scrambling to secure production slots, with Apple leading the charge alongside Qualcomm for the largest allocations.
The Taiwan-based semiconductor manufacturer expects to produce between 45,000 to 50,000 wafers monthly across its Baoshan and Kaohsiung facilities by the end of 2025, with capacity scaling to over 100,000 wafers per month in 2026.
TSMC's 2nm process promises a major leap forward from current 3nm technology, with up to 15% faster performance and 30% better power efficiency compared to the A19 chips expected in iPhone 17 models. The advanced node allows for higher transistor density, which should translate to improved processing capabilities and battery life for Apple's A20 chips. Industry analysts Ming-Chi Kuo and Jeff Pu have said the A20 chip in iPhone 18 models will be manufactured with TSMC's first-generation 2nm (N2) process, so it seems all but confirmed that Apple will adopt the more advanced silicon wafer technology.
Beyond Apple and Qualcomm, TSMC's 2nm customer list will expand in 2027 to include NVIDIA, Amazon's Annapurna, Google, and more than ten other major clients. TSMC is accelerating capacity expansion plans as a result, with full utilization expected in 2026.
In a change to Apple's usual iPhone launch cycle, iPhone 18 Pro models are expected to launch in fall 2026, with the base iPhone 18 and entry-level iPhone 18e scheduled for release the following March.
Article Link: Apple Secures Half of TSMC's 2nm Production Capacity for iPhone 18
Only 2 nm!!! Oh no, Apple's going to run out of nms soon! Doomed... 🤣
Stupid games like that will backfire.
Japan's Rapidus plans to have 2nm class mass production available in 2027.
As a very new enterprise, it's unclear whether they will follow the TSMC model (under promise and over deliver year after year) or the Intel/Samsung model (promise the world then repeatedly fail); but one's stereotypical bias is that being Japanese they will deliver something impressive.
This is also an interesting project to keep an eye on in terms of geopolitics. Intel is trying hard to squeeze whatever they can out of the US taxpayer with their "oh but what about the big scary Chinese" act; but if Rapidus deliver well, the US government might be better off contracting with them to build a US fab than trying to work further with the pathological deviants at Intel.
All these devices we throw away... anyone ever think about how many are in landfills now?
Personally I bought 50% of Apple's 50% of TSMCs wafers just so I could throw them all away.
Personally I bought 50% of Apple's 50% of TSMCs wafers just so I could throw them all away.
No. They hare just going to count slower and smaller. There are 10 digits after the deciman point. 10.0 , 10.1, 10.2 ... 10.9. There are 100 two digits back 10.00 ... 10.98, 10.99. Three digits 1,000 10.000 ... 10.999 . There are an infinite set of numbers between 10 and 11 is just keep ging right of the decimal point.
The n
These process label names are really "nm" anymore. The next shift will be Angstroms and an 'A' in front. 2nm is 20A. Instead of play deeper into the right of the decimal point they just shift to a base that is small than 'nanometer'.
It really isn't 'nm' in the first place at this point. So the physical dimensions of the atomic don't really count because not counting anything physical anyway. What will come eventually is that 'problem' that 'nm' is a 2D measure and the chips will go more 3D. As long as strictly stick with a 2D measurement to ecode a 3D issue there will be a number of quirky artifacts that show up in the marketing number produced.
If stack the circuits on a single die 2 or 4 deep then they'll claim that it takes 1/2 or 1/4 the amount of 2D space. If stack high enough it would shrink past what the individual 2d sheet is.
A limit on sales marketing folks making up a new insanely great label to spin to customers. Don't hold your breath.
The blocking issue coming up though is more so making it affordable as opposed to making it smaller. If smaller gets to be too expensive it will slow (if not stop).
Well, the Apple chips are not some special "one of a kind" unicorns with magic mojo, they are just optimized original ARM architecture processors. Nearly every ARM64 based Linux distribution runs fine on Apple Silicone without modification - go figure. 😎
[edit] Most probably Windows ARM64 would also run fine on AS, but surely not wanted.
This is patently false. Apple designs a 100% custom architecture that runs the ARM ISA.
Your statement is like saying AMD processors are nothing more than an optimized Intel x86 architecture.
The reason they can run Linux is because they are ARM compatible the same way AMD processors can run Windows because they are x86 compatible
Apples processor designs are simply superior and do in fact have “magic mojo”.
