Info TSMC, Samsung and IFS's Quarterly Financial Reports

[Markfw]

六大客戶訂單加持 台積電3奈米旺到年底 | 產業熱點 | 產業 | 經濟日報

台積電今年在蘋果、輝達、英特爾、聯發科、高通及博通等六大客戶的人工智慧（AI）、高效運算（HPC）訂單動能大力加持下，產...

money.udn.com

Here is the source, yeah AMD is not listed. Anyhow the information is for this year, AMD should have more tape out for 3nm which contribute more. N4P is relatively cheaper, maybe that's why...

Maybe its lost in the chinese to english translation, as its not there either. In fact look here: They are 3rd.

(link https://exploresemis.substack.com/p/tsmcs-top-10203040-customers-who)

 

[Tigerick]

Maybe its lost in the chinese to english translation, as its not there either. In fact look here: They are 3rd.

(link https://exploresemis.substack.com/p/tsmcs-top-10203040-customers-who)

View attachment 93679

Yeah, I have reread the article, it seems to list top 6 OEMs for 3nm revenue, then it is making sense. Anyhow it is unofficial list, do you have list for 2023?

PS: I have removed top 6 list due to different context.

 

[Markfw]

Yeah, I have reread the article, it seems to list top 6 OEMs for 3nm revenue, then it is making sense. Anyhow it is unofficial list, do you have list for 2023?

PS: I have removed top 6 list due to different context.

Google listed that one, I did not look down the list further.

 

[Tigerick]

	2023	2024	2025	2026	2027	2028
IFS	Intel 3	Intel 18A
Density	143	195
Samsung	SF3E	SF3	SF2	SF2P	1.4 nm
Density	159	178	187	198	243
Perf / Power	6.56	7.79	9.62	10.55
TSMC	N3B	N3E	N2	A16		A14
Density	283	273	313	349		392
Perf / Power	9.3	10.26	16.92

TechInsights has presented their comparison table between IFS, SF and TSMC at 2024 SEMI International Strategy Symposium. IFS's process density as suspected is pretty low, that's why they don't win any major contract. I have calculated performance/power at the table above.

N3B's density is in fact denser than upcoming N3E due to multi-patterning solution. And that's why Apple will continue using it for M4 series cause M4 Max will likely pass 100 billions transistor threshold.

 

[Tigerick]

https://www.reuters.com/technology/intel-discloses-financials-foundry-business-2024-04-02/

During a presentation for investors, Chief Executive Pat Gelsinger said 2024 would be the year of worst operating losses for the company's chipmaking business and that it expects to break even on an operating basis by about 2027.

Here is the BS from Pat:

Gelsinger said the foundry business was weighed down by bad decisions, including one year ago against using extreme ultraviolet (EUV) machines from Dutch firm ASML. Partially as a result of the missteps, Intel has outsourced about 30% of the total number of wafers to external contract manufacturers such as TSMC, Gelsinger said. It aims to bring that number down to roughly 20%.

Anyhow, I have created a table to compare yearly revenue from 3 foundry. We should see how IF going to turn things around in 2030:-

	Intel Foundry	+/- Margin	Report Loss Business to TSMC's N3B		Samsung Foundry		Total IF + SF Revenue		TSMC's Advanced Process (N7 & Below)
2022	$27.49 billions	- $5.2 billions			$22.53 billions		$50 billions		$40.22 billions
2023	$18.9 billions	- $7 billions			$16.77 billions		$35.67 billions		$40.16 billions
2024	$17.6 billions	- $13.4 billions	- $4 billions		$19.31 billions		$36.91 billions		$62.33 billions
2025		- -	-$10 billions
2026		- -	NVL-S N2
2027		- -
2028
2029
2030	> SF

Based on revenue figures of IFS in 2023 ($18.9 billions), IFS should overtake SF already which has earned $16.77 billions in 2023. The reason behind is Intel going to order N3B wafers from TSMC this year. There is report about Intel going to spend $4 billions this year and $10 billions next year for N3B wafers. That's mean this year IFS's revenue should drop sequentially...

It also means the worst year of Intel Foundry has not come yet...

 

[H433x0n]

View attachment 95493

	2023	2024	2025	2026	2027	2028
IFS	Intel 3	Intel 18A
Density	143	195
Samsung	SF3E	SF3	SF2	SF2P	1.4 nm
Density	159	178	187	198	243
Perf / Power	6.56	7.79	9.62	10.55
TSMC	N3B	N3E	N2	N2P		14A
Density	283	273	313	349		392
Perf / Power	9.3	10.26	16.92

TechInsights has presented their comparison table between IFS, SF and TSMC at 2024 SEMI International Strategy Symposium. IFS's process density as suspected is pretty low, that's why they don't win any major contract. I have calculated performance/power at the table above.

N3B's density is in fact denser than upcoming N3E due to multi-patterning solution. And that's why Apple will continue using it for M4 series cause M4 Max will likely pass 100 billions transistor threshold.

This is comparing the theoretical max dense library. So it’s comparing 2-1-fin cells against a 3 fin cell. A more proper comparison would be comparing TSMC’s 2-2 fin cell library that they use as standard for HPC against Intel’s 3 fin library used in Intel 3. When using that metric TSMC is still considerably more dense 182.5mtr/mm2 vs 143mtr/mm2 but it’s comparing apples/apples. These are the cells that would be used in AMD processors and Nvidia GPUs.

 

[Tigerick]

TSMC Process	Features	HVM Date	Actual HVM's Date	First Device Shipped	Product
N3E	Base	H2 2023	Q4 2023	Q2 2024	iPad Pro M4
N3P	Density +4%	H2 2024	Q4 2024	Q3 2025	iPhone 17 A19 Pro
N3X	Density +~14%	H2 2025	Q4 2025	Q3 2026 ?	Zen 6 CCD
N2	Performance +15% Density +15% Base w/ GAA	H2 2025	Q4 2025	Q2 2026 - WWDC ?	M5 Pro and Max ?
N2P	Performance +10%	H2 2026		2027
A16	N2P with SPR: Density +10%	H2 2026		2027
N2X	+10% Fmax	2027		2028
A14	2nd Gen GAA Performance +15% Density +23%	2028		2029
A14 with SPR	Density +10% ?	2028		2029

I have compiled upcoming 2nm process from TSMC. There are 3 variants of N2 family, once you know the differences, it will be easier to understand each node and delivery date. That's why the slide below seems contain lots of misinformation:-

 

[MrTeal]

Jesus christ @ Bitmain. I know they're the 1000 lbs gorilla of Bitcoin mining, but the same number of 2nm wafers in 2025 as Apple is nuts.

It funny for as much as the goal of BTC was decentralization, pretty much every single transaction on the blockchain runs through equipment sole sourced from one Chinese company.

 

[FlameTail]

View attachment 112264

Qualcomm is missing in that Trendforce chart. Any idea why?

Intel, AMD, Apple, Nvidia, and MediaTek tipped to use TSMC's cutting-edge 2 nm node; Qualcomm notably absent

TSMC's upcoming 2 nm node has plenty of high-profile customers such as AMD, Apple, Intel, Nvidia, MediaTek, Broadcom and Bitmain. The chips will be used in a wide array of devices, ranging from smartphones to cryptocurrency mining ASICs.

www.notebookcheck.net

 

[MrTeal]

Qualcomm is missing in that Trendforce chart. Any idea why?

Intel, AMD, Apple, Nvidia, and MediaTek tipped to use TSMC's cutting-edge 2 nm node; Qualcomm notably absent

TSMC's upcoming 2 nm node has plenty of high-profile customers such as AMD, Apple, Intel, Nvidia, MediaTek, Broadcom and Bitmain. The chips will be used in a wide array of devices, ranging from smartphones to cryptocurrency mining ASICs.

www.notebookcheck.net

Most sources have the same speculation as the article you linked; Qualcomm pivoting to Samsung.

 

[FlameTail]

Most sources have the same speculation as the article you linked; Qualcomm pivoting to Samsung.

Or to Intel Foundry perhaps?

https://forums.anandtech.com/threads/qualcomm-snapdragon-thread.2616013/post-41346575

 

[Doug S]

Jesus christ @ Bitmain. I know they're the 1000 lbs gorilla of Bitcoin mining, but the same number of 2nm wafers in 2025 as Apple is nuts.

It funny for as much as the goal of BTC was decentralization, pretty much every single transaction on the blockchain runs through equipment sole sourced from one Chinese company.

If that chart is correct then TSMC will only be making M5s for Apple in 2025. They sell more than 10 iPhones for every Mac they sell, so obviously that M5 volume would be a faction of their later volume with A20.

 

[Tarkin77]

TSMC Process	Features	HVM Date	Actual HVM's Date	First Device Shipped	Product
N3E		H2 2023	Q4 2023	Q2 2024	iPad Pro M4
N3P		H2 2024		2025
N2	Base w/ GAA	H2 2025		2026
N2P	Performance +10%	H2 2026		2027
A16	N2P with SPR: Density +10%	H2 2026 ?		2027

I have compiled upcoming 2nm process from TSMC. There are 3 variants of N2 family, once you know the differences, it will be easier to understand each node and delivery date. That's why the slide below seems contain lots of misinformation:-


View attachment 112264

Zen 6 and mi400 taped out already??

 

[FlameTail]

Most sources have the same speculation as the article you linked; Qualcomm pivoting to Samsung.

Pivoting to Samsung for flagship products sounds like a blunder to me. The last time Qualcomm did so, it didn't end well (888, 8 Gen 1).

If Qualcomm isn't using TSMC N2 (and that's a big if), then I think it's more likely that Qualcomm is pivoting to Intel Foundry than Samsung Foundry.

Samsung's nodes are lagging behind TSMC in all PPA metrics, and they also have horrible yield issues on top of that. On the other hand, it seems Intel 18A-P could match TSMC N2 in atleast some PPA metrics. There are doubts about Intel Foundry too, but it's looking much more hopeful than Samsung.