Discussion Leading Edge Foundry Node advances (TSMC, Samsung Foundry, Intel)

Page 41 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DisEnchantment

Golden Member
Mar 3, 2017
1,626
5,908
136
TSMC's N7 EUV is now in its second year of production and N5 is contributing to revenue for TSMC this quarter. N3 is scheduled for 2022 and I believe they have a good chance to reach that target.

1587737990547.png
N7 performance is more or less understood.
1587739093721.png

This year and next year TSMC is mainly increasing capacity to meet demands.

For Samsung the nodes are basically the same from 7LPP to 4 LPE, they just add incremental scaling boosters while the bulk of the tech is the same.

Samsung is already shipping 7LPP and will ship 6LPP in H2. Hopefully they fix any issues if at all.
They have two more intermediate nodes in between before going to 3GAE, most likely 5LPE will ship next year but for 4LPE it will probably be back to back with 3GAA since 3GAA is a parallel development with 7LPP enhancements.


1587739615344.png

Samsung's 3GAA will go for HVM in 2022 most likely, similar timeframe to TSMC's N3.
There are major differences in how the transistor will be fabricated due to the GAA but density for sure Samsung will be behind N3.
But there might be advantages for Samsung with regards to power and performance, so it may be better suited for some applications.
But for now we don't know how much of this is true and we can only rely on the marketing material.

This year there should be a lot more available wafers due to lack of demand from Smartphone vendors and increased capacity from TSMC and Samsung.
Lots of SoCs which dont need to be top end will be fabbed with N7 or 7LPP/6LPP instead of N5, so there will be lots of wafers around.

Most of the current 7nm designs are far from the advertized density from TSMC and Samsung. There is still potential for density increase compared to currently shipping products.
N5 is going to be the leading foundry node for the next couple of years.

For a lot of fabless companies out there, the processes and capacity available are quite good.
 

Ajay

Lifer
Jan 8, 2001
15,754
7,992
136
Utilization rate for Fab18 for N5 family is expected to fall below 70%. This additional to the drop in the utilization rate for Fab14/Fab15 for N7 family.

TSMC already cut the planned investment of 44B to 36B.
Well, that pretty much tells us all we need to know. The largest foundry in the world facing this degree of retrenchment means the whole semiconductor supply chain it going to see tough times for a while. Wonder how this will affect ASML given their very high book to bill ratio.
 

Doug S

Platinum Member
Feb 8, 2020
2,368
3,776
136
Well, that pretty much tells us all we need to know. The largest foundry in the world facing this degree of retrenchment means the whole semiconductor supply chain it going to see tough times for a while. Wonder how this will affect ASML given their very high book to bill ratio.

Nobody is going to cancel ASML orders they already have in the queue. What it probably means for them is the waiting list will become shorter, but we're years away from it going away.

Also remember these "retrenchments" are from the greatly increased levels of investment during the heady days of 2021 and 2022 when it looked like foundries would be fully booked for years and governments were throwing wads of cash at anyone ready to spend a few billion building a fab in their borders. More regression to the mean, in other words.
 

BorisTheBlade82

Senior member
May 1, 2020
667
1,022
136
Well, that pretty much tells us all we need to know. The largest foundry in the world facing this degree of retrenchment means the whole semiconductor supply chain it going to see tough times for a while. Wonder how this will affect ASML given their very high book to bill ratio.
Definitely severely. The fact that they are sitting on a monopoly (although a deserved one) might help them sail through it a bit softer - relatively speaking.
I am sure they have some brutal conditions on order cancellation - including being left out in the cold when the market picks up again.
 

Ajay

Lifer
Jan 8, 2001
15,754
7,992
136
Nobody is going to cancel ASML orders they already have in the queue. What it probably means for them is the waiting list will become shorter, but we're years away from it going away.

Also remember these "retrenchments" are from the greatly increased levels of investment during the heady days of 2021 and 2022 when it looked like foundries would be fully booked for years and governments were throwing wads of cash at anyone ready to spend a few billion building a fab in their borders. More regression to the mean, in other words.
Yep. Everyone forgets about the cyclical nature of the semiconductor market when times are good. Even companies at the core of that market. One would think they would be better at their forecasts by now.
 

Exist50

Platinum Member
Aug 18, 2016
2,445
3,043
136
Utilization rate for Fab18 for N5 family is expected to fall below 70%. This additional to the drop in the utilization rate for Fab14/Fab15 for N7 family.

TSMC already cut the planned investment of 44B to 36B.
I have to think there's something lost in translation here. An actual 70% utilization would be catastrophic. I'd think it's something like "70% of expected" or "70% of expected growth".
 

Doug S

Platinum Member
Feb 8, 2020
2,368
3,776
136
Yep. Everyone forgets about the cyclical nature of the semiconductor market when times are good. Even companies at the core of that market. One would think they would be better at their forecasts by now.

Do they forget, or do they know getting board approval for increasing investments by billions can only happen when times are good? Even if they know they will have to cancel some of that investment when the cycle turns, they've got some projects far enough along it is too late to turn back.

If they invested conservatively during good times, that's no guarantee they would be allowed to continue to invest at similar levels when the company is losing money during a bad downcycle.
 

BorisTheBlade82

Senior member
May 1, 2020
667
1,022
136
I have to think there's something lost in translation here. An actual 70% utilization would be catastrophic. I'd think it's something like "70% of expected" or "70% of expected growth".
It is not too far fetched...
Will the TAM of 5nm shrink to 70% of what they were planning for capacity in the crazy market of the last 2-3 years? I mean, seemingly everyone was heavily wrong in predicting this decline.
And if you can get some good government money by building Fabs, you might factor in some under-utilization as well.
Heck, in Germany the industry gets huge amounts of tax money in order to keep their employees paid in such situations - this quite likely being an important factor for Intel and TSMC in their decision making process.

Of course you can lower lead times and a lot of TSMC's customers would be happy to get some orders brought forward. But the sensible thing for most in such a situation is to level out throughput over time instead of going full throttle until the cliff.
 

Exist50

Platinum Member
Aug 18, 2016
2,445
3,043
136
It is not too far fetched...
Will the TAM of 5nm shrink to 70% of what they were planning for capacity in the crazy market of the last 2-3 years? I mean, seemingly everyone was heavily wrong in predicting this decline.
And if you can get some good government money by building Fabs, you might factor in some under-utilization as well.
Heck, in Germany the industry gets huge amounts of tax money in order to keep their employees paid in such situations - this quite likely being an important factor for Intel and TSMC in their decision making process.

Of course you can lower lead times and a lot of TSMC's customers would be happy to get some orders brought forward. But the sensible thing for most in such a situation is to level out throughput over time instead of going full throttle until the cliff.
I'm still very skeptical... I could see a significant reduction in short term demand, but translating that to fab utilization is tricky. It's been a while since I've heard the numbers, but iirc ~95% is ideal. 70%, they'd be spending hundreds of millions of dollars, if not billions, on maintaining the fabs without any revenue to match. If that number is even close to realistic, TSMC's margins will crater.
 

Doug S

Platinum Member
Feb 8, 2020
2,368
3,776
136
Some perfect examples of normalcy bias here.

Versus exceptionalism - claims that this time things are different, like gold bugs who scream "fiat money is dead hyperinflation is starting!" every time the stock market drops, the dollar is down, or gas prices surge.
 

maddie

Diamond Member
Jul 18, 2010
4,780
4,759
136
Versus exceptionalism - claims that this time things are different, like gold bugs who scream "fiat money is dead hyperinflation is starting!" every time the stock market drops, the dollar is down, or gas prices surge.
Yep, except the broken clock is regularly correct. Don't lump all crazy claims in the same bag as you'll always find someone who claims some outrageous future. By the way, a nice trick to discredit opponents, compare them to a crazy and then say "see".

Let's revisit this by year end, if we even remember :), and compare present views.
 

Ajay

Lifer
Jan 8, 2001
15,754
7,992
136
Do they forget, or do they know getting board approval for increasing investments by billions can only happen when times are good? Even if they know they will have to cancel some of that investment when the cycle turns, they've got some projects far enough along it is too late to turn back.

If they invested conservatively during good times, that's no guarantee they would be allowed to continue to invest at similar levels when the company is losing money during a bad downcycle.
Fair point. Thanks.
 

Doug S

Platinum Member
Feb 8, 2020
2,368
3,776
136
Yep, except the broken clock is regularly correct. Don't lump all crazy claims in the same bag as you'll always find someone who claims some outrageous future. By the way, a nice trick to discredit opponents, compare them to a crazy and then say "see".

Let's revisit this by year end, if we even remember :), and compare present views.

Feel free to mark Jan 24, 2024 in your calendar and tell me "I told you so" if you're right so long as you will also tell me "I was wrong, I admit it" if you aren't :)
 
  • Like
Reactions: maddie

jpiniero

Lifer
Oct 1, 2010
14,732
5,363
136
Remember that Meteor Lake (and who knows what else made by Intel coming out in 2023/4+) is mostly TSMC, even if the CPU tile is still made internally. TSMC said that they expect business to pick up in the second half.... picking up 'share' might be a part of that.

Now how big of a volume Meteor Lake is of course TBD.
 

jerrytw

Junior Member
Apr 12, 2022
3
10
51
From the Samsung investor forum, Samsung has said their SF2 process offers a 5% density improvement over SF3 with an 8% performance boost or 16% lower power consumption. And only SF2P has BSPDN. For the SF1.4 process, the area will reduce by more than 30% compared to the SF2 process. Any thought?
 

Doug S

Platinum Member
Feb 8, 2020
2,368
3,776
136
From the Samsung investor forum, Samsung has said their SF2 process offers a 5% density improvement over SF3 with an 8% performance boost or 16% lower power consumption. And only SF2P has BSPDN. For the SF1.4 process, the area will reduce by more than 30% compared to the SF2 process. Any thought?


Samsung is good at announcing new nodes and making it sound like they are even with TSMC, but they are terrible at delivering working processes ever since EUV arrived.
 

Geddagod

Golden Member
Dec 28, 2021
1,163
1,047
106
From the Samsung investor forum, Samsung has said their SF2 process offers a 5% density improvement over SF3 with an 8% performance boost or 16% lower power consumption. And only SF2P has BSPDN. For the SF1.4 process, the area will reduce by more than 30% compared to the SF2 process. Any thought?
Samsung's own claims of density or performance are often contradictory when comparing their own nodes. Kinda annoying.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,659
14,659
136
Intel IFS's president is stepping down. Replacement already announced.

Also of note: another exit and news:
Additionally on Tuesday, Gelsinger said Raja Koduri, who was leading Intel's (INTC) initiative in graphic chips, would leave the company at the end of the month.

Intel (INTC) shares fell more than 3% in mid-day trading on Tuesday,
while Taiwan Semiconductor (TSM) and GlobalFoundries (GFS) both saw modest gains.
 

Exist50

Platinum Member
Aug 18, 2016
2,445
3,043
136
  • Like
Reactions: Tlh97 and Hitman928