iPhone 17 Pro was originally planned to be the first to use TSMC's 2nm process: Apple postponed it because it was too expensive

[Daniel Nenni]

According to media reports on January 1, Apple originally planned to use TSMC's 2nm processor chips in the iPhone 17 Pro and iPhone 17 Pro Max, but due to TSMC's high 2nm cost and limited production capacity, Apple postponed the commercialization of 2nm to 2026.

The report pointed out that TSMC has started trial production of the 2nm process at its Baoshan plant in Hsinchu. The initial yield is 60%, which means that nearly 40% of the wafers cannot be used and the cost of each wafer is as high as US$30,000.

Therefore, Apple has postponed the commercialization of 2nm by one year. It is expected that the 2026 iPhone 18 Pro series will be the first to launch TSMC's 2nm chip, which means that the iPhone 17 series will continue to use TSMC's 3nm process.

Earlier at the IEDM 2024 conference, TSMC disclosed key details of the 2nm process. Compared with 3nm, the transistor density of 2nm increased by 15%, the performance increased by 15% at the same power consumption, and the power consumption decreased by 24-35% at the same performance.

In addition, TSMC's 2nm process introduces gate-all-around (GAA) nanosheet transistors for the first time, which helps to adjust the channel width and balance performance and energy efficiency. The nanosheet transistors using the new process can achieve significant energy efficiency improvements at a low voltage of 0.5-0.6V.

​

iPhone 17 Pro原计划首发台积电2nm工艺：苹果嫌贵 推迟了

快科技1月1日消息，据媒体报道，苹果公司原本计划在iPhone 17 Pro和iPhone 17 Pro Max应用台积电2nm处理器芯片，但由于台积电2nm成本高、产能有限，苹果将2nm商用时间推迟到了2026年。报道指出

news.mydrivers.com

 

[Daniel Nenni]

Cost is not the problem. Apple funds TSMC development and is granted first access. Yield is not the problem. N2 is yielding ahead of plans. If Apple is concerned sales will outpace supply they can use both N2 and N3 with different phone models, which they have before. If Apple did skip N2, QCOM and MediaTek will gladly take up the slack but given Apple's massive technology ego I seriously doubt that will happen. Just my opinion of course.

 

[Barnsley]

Cost is not the problem. Apple funds TSMC development and is granted first access. Yield is not the problem. N2 is yielding ahead of plans. If Apple is concerned sales will outpace supply they can use both N2 and N3 with different phone models, which they have before. If Apple did skip N2, QCOM and MediaTek will gladly take up the slack but given Apple's massive technology ego I seriously doubt that will happen. Just my opinion of course.

Maybe the innovation hasnt been there this year to warrant such a move.

Or do you think Apple have seen some things on the Android Phones and Huawei mobile devices that they feel they need to incorporate into their phones and have that pumped by fanbois as the latest and greatest?

 

[nghanayem]

A little early for an April 1st joke news.mydrivers.com. N2 isn't supposed to enter HVM until the very end of the year (TSMC's own words). If Apple was actually going to use N2 for this year (especially for the non pro) TSMC would have needed to have started HVM back in December or maybe at the latest January with how long modern CMOS processing times are. Cost had nothing to do with it and the Iphone 17 would have never been intended to use an N2 chip. Nor does one just change which node a chip is going to use less than a year out from systems using the chip launching. It is simply impossible. Since TSMC never committed to a late 2024 HVM start I see no reason why anyone would call Apple not using a process not ever intended to be HVM this early as "postponing adoption". Unless we want to say Apple also postponed using A7 CFET for the M5 iPad this year . I never really got the people saying N2 iphones in 2025. TSMC claims N3 started HVM in Dec'22 and pro only Iphones didn't drop until September 23. Then you have TSMC publicly say N2 will not enter HVM until late Q4'25 to early Q1'26 and people somehow get it in their heads that Apple will launch finished iPhones in the millions 1-2 quarters before TSMC even started making the bloody things. The whole thing annoys me because TSMC worked real hard on N2 and it is a really strong technology that by all indications looks very healthy; and yet people call foul. Stupid...

 

[Maximus]

For Pro/ProMax models:

2023: N3B
2024: N3E
2025: N3P
2026: N2

N2 HVM is aiming for the second half of 2025 (likely Q4), which doesn't align with the 2025 iPhone timeline. There was never a plan to use N2 for the 2025 iPhone, so there's no 'delay' involved.

 

[Daniel Nenni]

A little early for an April 1st joke news.mydrivers.com. N2 isn't supposed to enter HVM until the very end of the year (TSMC's own words). If Apple was actually going to use N2 for this year (especially for the non pro) TSMC would have needed to have started HVM back in December or maybe at the latest January with how long modern CMOS processing times are. Cost had nothing to do with it and the Iphone 17 would have never been intended to use an N2 chip. Nor does one just change which node a chip is going to use less than a year out from systems using the chip launching. It is simply impossible. Since TSMC never committed to a late 2024 HVM start I see no reason why anyone would call Apple not using a process not ever intended to be HVM this early as "postponing adoption". Unless we want to say Apple also postponed using A7 CFET for the M5 iPad this year . I never really got the people saying N2 iphones in 2025. TSMC claims N3 started HVM in Dec'22 and pro only Iphones didn't drop until September 23. Then you have TSMC publicly say N2 will not enter HVM until late Q4'25 to early Q1'26 and people somehow get it in their heads that Apple will launch finished iPhones in the millions 1-2 quarters before TSMC even started making the bloody things. The whole thing annoys me because TSMC worked real hard on N2 and it is a really strong technology that by all indications looks very healthy; and yet people call foul. Stupid...

TSMC's official words are N2 HVM 2H 2025 and N2P is late 2026. N2 already has tape-outs and I can assure you one of them is Apple and another is probably QCOM since N2 is SoC optimized. I know MediaTek is another N2 partner but they are generally not first out of the gate.

TSMC N3 will live on to be the best node for TSMC with 99% of the tape-outs. Hopefully Samsung and Intel Foundry get some N2 business.

 

[nghanayem]

TSMC's official words are N2 HVM 2H 2025 and N2P is late 2026. N2 already has tape-outs and I can assure you one of them is Apple and another is probably QCOM since N2 is SoC optimized. I know MediaTek is another N2 partner but they are generally not first out of the gate.

TSMC N3 will live on to be the best node for TSMC with 99% of the tape-outs. Hopefully Samsung and Intel Foundry get some N2 business.

I never said N2 didn't have tape outs, but tape outs don't equal HVM. If they did, then we could say absurd things like N3 entered HVM one year after N5 HVM (in 2021). TSMC even went so far to say the first production wafers won't ship to customers for revenue until late Q1 to early Q2 2026. How can Apple launch millions of iPhones before the first production lots even reach EOL (or even start). It simply is not possible. But credit where credit is due, I did misremember Dec'25. TSMC said Q4'25 is when they expected to start the first production material. Now, maybe it does end up starting HVM in Dec just like N3 did, but TSMC hasn't given the public any granularity beyond Q4'25.

From Q1'24 earnings:

C.C. Wei -- Chief Executive Officer


Randy, the N2's ramp profile we say is very similar to N3 because of, look at the cycle time, we start the N2 production in the second half of 2025, actually in the last quarter of 2025. And because of the cycle time and all the kind of back-end process, and so we expect the meaningful revenue will start from the end of the first quarter or beginning of the second quarter of 2026. That's what we mean that is the profile is very similar to N3. Now your second question is there have been a lot of engagement and the tape-out will be higher, and do we see a very steep kind of a production? Well, we do expect that, but let me say again, N2 is a very complicated work or a very complex technology node.


So my customer, they also take a little bit longer time to prepare for the tape-out. So that's why they all engage with TSMC in the early stage. And -- but for their product ramp-up, they will have their own product road map and their own business consideration. However, we still say that N2 will be a very, very big node for TSMC.

 

[siliconbruh999]

TSMC's official words are N2 HVM 2H 2025 and N2P is late 2026. N2 already has tape-outs and I can assure you one of them is Apple and another is probably QCOM since N2 is SoC optimized. I know MediaTek is another N2 partner but they are generally not first out of the gate.

TSMC N3 will live on to be the best node for TSMC with 99% of the tape-outs. Hopefully Samsung and Intel Foundry get some N2 business.

Intel should at least keep their business in house same with Samsung.
This is a poor showing. If you don't use your own fabs why would another company use yours at least Intel is back in-house with 18A massively

 

[IMakeICsAtHomeJK]

At what point does more powerful technology reach a stage of diminishing returns? (The graph of the logarithmic function comes to mind.) More specifically, digital devices ostensibly perform similar (if not the exact) social functions as a decade previous. One could say: "But, the same could be said of 2015." However, mobile phones were decidedly not similar in social function when comparing 2005 to 2015. A reasonable wager would be that for the vast majority of users of flagship smart-phones, the computational resources idle on stand-by; and in the case of activation rarely, if ever, exploit the peak capabilities available. Exemplorum gratia: "FaceBook-machine", "FaceTime-machine", "CandyCrush-machine", and so on.
Of course, the updated technology is said to be more energy efficient: ["More performant for longer durations."] Light is more effectively algorithmically processed when passed through the camera lens. Video games benefit from increased graphics potentials. Data transfer rates and textual analysis benefit from enhanced or supplementary hardware modules. Yet these incremental improvements are examples of the macro-trend as whole: at best, marginally relevant gradualism.
Binary logarithm plot with grid.png (Rus, 2021)

Earlier at the IEDM 2024 conference, TSMC disclosed key details of the 2nm process. Compared with 3nm, the transistor density of 2nm increased by 15%, the performance increased by 15% at the same power consumption, and the power consumption decreased by 24-35% at the same performance.

Absolutely incredible achievement.

Cost is not the problem. Apple funds TSMC development and is granted first access. Yield is not the problem.

Excuse the question if the answer is obvious, but what then is "the problem"? Personal guess: the 'bang-for-buck' ratio is unjustifiable to Apple Inc.

N2 isn't supposed to enter HVM until the very end of the year (TSMC's own words). If Apple was actually going to use N2 for this year (especially for the non pro) TSMC would have needed to have started HVM back in December or maybe at the latest January with how long modern CMOS processing times are. Cost had nothing to do with it and the Iphone 17 would have never been intended to use an N2 chip. Nor does one just change which node a chip is going to use less than a year out from systems using the chip launching. It is simply impossible. Since TSMC never committed to a late 2024 HVM start I see no reason why anyone would call Apple not using a process not ever intended to be HVM this early as "postponing adoption".

Wow. nghanayem's post reads quite reasonable in its assertions. The statement "N2 isn't supposed to enter HVM until the very end of the year (TSMC's own words)." (nghanayem) pointedly answers my question of " . . . what then is 'the problem'": there is no problem. Web search confirms this to be the case. One source provided out of many below.
TSMC: 2nm Chips Arriving in 2026 "TSMC's N2 schedule is on track to enter high-volume manufacturing in late 2025."

For Pro/ProMax models:

2023: N3B
2024: N3E
2025: N3P
2026: N2

N2 HVM is aiming for the second half of 2025 (likely Q4), which doesn't align with the 2025 iPhone timeline. There was never a plan to use N2 for the 2025 iPhone, so there's no 'delay' involved.

Continuing the thread, Maximus succinctly concurs no such "problem" exists. Moreover, let's grant 振亭's (the article's author) declaration of an approximate year-long delay. In the grand scheme of things, the hold-up towards the decelerated march of slight innovation is inconsequential. There is a reason German automobiles (Porsche, Bavarian Motor Works, Mercedes-Benz) and California-designed computers (Apple) are a status symbol among even the Chinese in the PRC. The reason is not absolute performance superiority. The reason has to do with intangibles. Aesthetic. User Experience. Affective engagement. Evoked mindstate. I am reminded of an individual putting forth the case that when the iPhone was first released, there was nothing 'special' about it. The individual argued that numerous other mobile phones, at the time: (1) had more powerful hardware, (2) an operating system with more features than the iPhone, (3) a greater functional ease-of-use for taking pictures or listening to music and so on, and (4) more potent status symbols (for example, Nokia 8800, Vertu Signature—a price tag of thousands of dollars). The individual's thesis was that the iPhone was more parsimonious—anxiolytic even. The fact of its being simpler than other devices at the time was interestingly its prime value-add. At a time of increasing digital complexity, it aided its user in making sense of the world.
Apologies for the long-winded point, it's just that this article on 'news.mydrivers.com' is bereft of worthwhile analysis—Hell, posters have ascertained its 'characterizations' (if we're being gracious) to be categorically false. And if one were to play Devil's Advocate to entertain 振亭's claim of lapsed timelines — and in this post writer's conspiratorial opinion: 振亭's implication of Apple's stumble — the claim/implication is still devoid of consumer impact; even shareholder relevancy, by extension. At this current right-most point of the logarithmic curve of the graph above, one gets the feeling that only a technological transposition to the exponential curve below with respect to progress would result in consumer reconsiderations.
Logarithm inversefunctiontoexp.svg.png (Stpasha, 2011)

This poster is quite skeptical, yet the present focus upon machine learning is a possible vehicle escape from deceleration.

[Post written on $400.00 USD 2022 HP laptop running GNU/Linux (Fedora). Poster is not solely pro-Apple.]

 

[Xebec]

Here's the process node evolution of Apple Silicon. I took the "Evolution of Apple A series chips" diagram from wikipedia, and added the nodes identified in Wikipedia for each chip. May not be exactly accurate, but it looks like Apple has never used a node more than twice so far. (A11 Bionic used "10nm" which was a fairly mild step but seems on paper significant vs. "14nm" and "7nm" class).

Note A9 was a time where they used both Samsung 14nm and TSMC 16nm nodes, hence S/T.

Diagram source: https://en.wikipedia.org/wiki/Apple_silicon

 

[Xebec]

At what point does more powerful technology reach a stage of diminishing returns?

One factor to consider is Apple is still beholden to shareholders. It's possible wall street would punish them if say they got stuck on a TSMC N3(x) for 5 years. Even though you're right it doesn't necessarily matter that much for the product (costs or performance), percieved stagnation could cause stock price drops.

 

[Daniel Nenni]

Here's the process node evolution of Apple Silicon. I took the "Evolution of Apple A series chips" diagram from wikipedia, and added the nodes identified in Wikipedia for each chip. May not be exactly accurate, but it looks like Apple has never used a node more than twice so far. (A11 Bionic used "10nm" which was a fairly mild step but seems on paper significant vs. "14nm" and "7nm" class).

Note A9 was a time where they used both Samsung 14nm and TSMC 16nm nodes, hence S/T.

View attachment 2638
Diagram source: https://en.wikipedia.org/wiki/Apple_silicon

The A8 (iPhone 6) and A8x are TSMC 20nm, the first Apple SoC on TSMC.

TSMC first used double patterning on 20nm then added FinFETs for 16nm. This is an example of the low risk approach TSMC adopted when Apple signed on.

TSMC 16nm did not yield as expected so Apple used both Samsung 14nm and TSMC 16nm for the iPhone 6s. Apple went back to TSMC after that and has not left. The Apple/TSMC agreement is one for the ages. It completely disrupted the foundry business for the greater good, absolutely.

 

[Daniel Nenni]

Excuse the question if the answer is obvious, but what then is "the problem"? Personal guess: the 'bang-for-buck' ratio is unjustifiable to Apple Inc.

Cost is not the issue. Apple is not just a TSMC customer, Apple is a very close development partner. Time-to-market (capacity) is the big issue for Apple and as I said before Apple can use both N2 and N3 for the next iPhone launch if they wanted to. TSMC R&D is making N2 wafers. TSMC Fab 20 is making N2 wafers. It all depends on what the yield is and how many die per wafers. I will look more into this now that the holidays are over but I still think it is possible for Apple to announce an N2 based product this year.