Discussion Leading Edge Foundry Node advances (TSMC, Samsung Foundry, Intel) - [2020 - 2025]

[DisEnchantment]

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.


N7 performance is more or less understood.


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

AnandTech Forums: Technology, Hardware, Software, and Deals

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

www.anandtech.com

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.

AnandTech Forums: Technology, Hardware, Software, and Deals

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

www.anandtech.com

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.

---------------------------------------------------------------------------------------------------------------------------------------------------


FEEL FREE TO CREATE A NEW THREAD FOR 2025+ OUTLOOK, I WILL LINK IT HERE

 

[SiliconFly]

.

 

[SiliconFly]

.

 

[jdubs03]

Good chat about IFS (and LNL, other stuff):

Starts at the 2:28:40 mark.

 

[dttprofessor]

For long term ,ifs is base of intel.
For short term,making more money is Top priority to intel now.

 

[Hitman928]

I know it's not leading edge fabs, but this is the closest thread for this news. TI announced they are receiving $1.6B in funding from the CHIPS act and ultimately expects to receive between $6B - $8B in additional tax credit to help finance the building/expanding of 3 fabs. The fabs nodes are in the 28 nm - 130 nm range which may not be very exciting, but these nodes are still in very commonly used in analog/RF/IoT designs.

Texas Instruments signs preliminary agreement to receive up to $1.6 billion in CHIPS and Science Act proposed funding for semiconductor manufacturing in Texas and Utah

NEWS HIGHLIGHTS: The U.S. Department of Commerce has proposed up to $1.6 billion for TI in direct funding through the CHIPS and Science Act to support three 300mm semiconductor wafer fabs under...

news.ti.com

 

[Dayman1225]

Intel and Amazon ink a multi billion dollar deal for 18A “Ai fabric” chip and custom Xeon6 on Intel 3 process. Looking into deeper collaboration for future chips for 18A, 18AP and 14A

Intel and AWS Expand Strategic Collaboration, Helping Advance U.S.-Based Chip Manufacturing

www.intc.com

Intel also announced they are setting IFS up as an independent subsidiary within Intel and that Germany and Poland fab buildouts will be paused for 2 years

Pat Gelsinger on Foundry Momentum, Progress on Plan

A message from Intel CEO Pat Gelsinger to employees regarding the next phase of Intel's transformation.

www.intel.com

 

[OriAr]

Intel and Amazon ink a multi billion dollar deal for 18A “Ai fabric” chip and custom Xeon6 on Intel 3 process. Looking into deeper collaboration for future chips for 18A, 18AP and 14A

Intel and AWS Expand Strategic Collaboration, Helping Advance U.S.-Based Chip Manufacturing

www.intc.com

Intel also announced they are setting IFS up as an independent subsidiary within Intel

Pat Gelsinger on Foundry Momentum, Progress on Plan

A message from Intel CEO Pat Gelsinger to employees regarding the next phase of Intel's transformation.

www.intel.com

That's pretty big for IFS, a public vote of confidence from Amazon will do IFS a world of good regarding confidence from other potential customers.

 

[Dayman1225]

Pat also spoke with Patrick Moorhead, saying that the custom chip goes beyond what they’ve done in the past

https://twitter.com/x/status/1835772073060553147

 

[Hitman928]

Pat also spoke with Patrick Moorhead, saying that the custom chip goes beyond what they’ve done in the past

https://twitter.com/x/status/1835772073060553147

Thanks. The additional details are good news. Seems like Intel is finally offering actual semi-custom designs for big players and scored a nice win with AWS because of it, though that is still speculation at this point. The AI fabric is a decent consolation prize I guess to not getting your actual AI chips sold. From the foundry perspective, this may help a little, but it's still all Intel made products from what I can tell, so I'm not sure it actually helps IFS outside of more revenue in general.

The plan to make IFS a subsidy may help get some of their competitors to use them in the future, but I don't think that's what was holding others (e.g., Broadcom) back, so they'll still need to fix the main issues first.

 

[Doug S]

Maybe Intel's recent troubles will be a good thing in the long run. I have been saying since day one that Intel had to spin off the foundry for it to be successful, but figured their x86 teams enjoy their special treatment so much that it would take until the foundry did more external business than internal (which I guessed would happen around 2030 or so) before the board would be willing to pull the trigger.

Now they're forced to do it earlier, which will really help the foundry since their x86 guys won't have that special access any longer which I think would be a blocker for some of the bigger potential customers, plus companies will be able to make strategic investments without worrying they are propping up a competitor (i.e. AMD and maybe Qualcomm if ARM PCs ever make fetch happen)

If A18 really is a separate die, or even if not their plan is for the Pro and non-Pro SoCs to be seperate dies in the future, that would open up the door for Apple to hand Intel a lot of volume. Less risk making the non-Pro version since it would be going into the lower end model, and Intel would probably give them good pricing in exchange for a fat prepayment that let them build out the space they'd need for such a big commitment of wafer volume. Be a good "beta test" for the relationship, and if Intel can match/beat TSMC then Apple would eventually send most of their volume Intel's way.

 

[Hitman928]

Maybe Intel's recent troubles will be a good thing in the long run. I have been saying since day one that Intel had to spin off the foundry for it to be successful, but figured their x86 teams enjoy their special treatment so much that it would take until the foundry did more external business than internal (which I guessed would happen around 2030 or so) before the board would be willing to pull the trigger.

Now they're forced to do it earlier, which will really help the foundry since their x86 guys won't have that special access any longer which I think would be a blocker for some of the bigger potential customers, plus companies will be able to make strategic investments without worrying they are propping up a competitor (i.e. AMD and maybe Qualcomm if ARM PCs ever make fetch happen)

If A18 really is a separate die, or even if not their plan is for the Pro and non-Pro SoCs to be seperate dies in the future, that would open up the door for Apple to hand Intel a lot of volume. Less risk making the non-Pro version since it would be going into the lower end model, and Intel would probably give them good pricing in exchange for a fat prepayment that let them build out the space they'd need for such a big commitment of wafer volume. Be a good "beta test" for the relationship, and if Intel can match/beat TSMC then Apple would eventually send most of their volume Intel's way.

Just for clarity's sake, Intel isn't spinning off the foundry, at least not yet. IFS will still be owned by Intel, it will just operate as a subsidiary. If Intel does spin it off completely, I expect it will be a bit at a time as they sells shares to outside investors to raise funds.

 

[Doug S]

Just for clarity's sake, Intel isn't spinning off the foundry, at least not yet. IFS will still be owned by Intel, it will just operate as a subsidiary. If Intel does spin it off completely, I expect it will be a bit at a time as they sells shares to outside investors to raise funds.

Yes but it will be a separate corporation with its own separate board so it is "separate enough" in my mind, even if it won't have its own stock symbol yet. It would still be able to solicit outside investments in the form of prepayments years in advance like Apple made in TSMC that got them to where they are today. That's probably what AWS is talking about, even if Intel is behind TSMC they might get an equal or better process simply because they aren't getting early access to TSMC nodes like Apple. Plus Intel is probably willing to cut them a deal on price for the early commitment.

 

[controlflow]

Yes but it will be a separate corporation with its own separate board so it is "separate enough" in my mind, even if it won't have its own stock symbol yet. It would still be able to solicit outside investments in the form of prepayments years in advance like Apple made in TSMC that got them to where they are today. That's probably what AWS is talking about, even if Intel is behind TSMC they might get an equal or better process simply because they aren't getting early access to TSMC nodes like Apple. Plus Intel is probably willing to cut them a deal on price for the early commitment.

In addition to getting outside investments as prepayments, Intel could also IPO the subsidiary and have its own stock symbol and still retain majority ownership similar to what they did with Mobileye. I suspect something like that won't happen in the near term though since Intel Foundry on its own today has horrific financials, but doing this could make sense down the road if Foundry can start to pull some moderate income.

Overall, this seems like a pretty logical move from Intel to make foundry more separate and satisfy customers and create a setup to potentially allow more external investments all while still mostly owning the fabs.

I think it is important to have a viable alternative to TSMC so hopefully this plan is successful.

 

[poke01]

Apple Mobile Processors Are Now Made in the USA. By TSMC

[Exclusive] The iPhone maker is ready to be the first client of TSMC's new Arizona fab. Finally, the iPhone will boast a processor that's Made in America

timculpan.substack.com

Apple A16 so N4P being made in USA

 

[Doug S]

Apple Mobile Processors Are Now Made in the USA. By TSMC

[Exclusive] The iPhone maker is ready to be the first client of TSMC's new Arizona fab. Finally, the iPhone will boast a processor that's Made in America

timculpan.substack.com

Apple A16 so N4P being made in USA

It was always expected that the Arizona fab would make N5 family stuff when N3 was state of the art. They'll probably start making N3 class stuff in Arizona around the time Apple's leading edge is using A16.

Arizona is never gonna be high volume for Apple unless TSMC stops restricting the leading edge stuff to Taiwan. Obviously for multiple reasons Apple would like it if TSMC had leading edge production in both Taiwan and the US, but Taiwan's govt may not see that as being in their interest as far as US willingness to intervene in a conflict with China.

 

[jdubs03]

Nice technical write up. I’m not a paid subscriber so there’s more detail about 2nm costs, but still good stuff above that.

Clash of the Foundries: Gate All Around + Backside Power at 2nm

Fab Cost, SRAM Scaling, WFE Implications, Backside Power Details, TSMC, Samsung, Intel, Rapidus TSMC won FinFET. All noteworthy leading edge logic designs, even Intel’s, are manufactured on TS…

www.semianalysis.com

 

[DavidC1]

Nice technical write up. I’m not a paid subscriber so there’s more detail about 2nm costs, but still good stuff above that.

Clash of the Foundries: Gate All Around + Backside Power at 2nm

Fab Cost, SRAM Scaling, WFE Implications, Backside Power Details, TSMC, Samsung, Intel, Rapidus TSMC won FinFET. All noteworthy leading edge logic designs, even Intel’s, are manufactured on TS…

www.semianalysis.com

They made a mistake.

18A offers 30% density improvement over Intel 3, not greater than 50% as it's listed. Intel themselves revealed it, so how can they get that wrong? Whole lot of words and gets the publicly revealed facts wrong.

 

[SiliconFly]

.

 

[Hitman928]

Wrong. Intel never said anything about density numbers (those are just rumors). Daniel Nenni himself stated that density improvement is around 1.6X, which is perfectly in line with the semianalysis article's number of >50%.

Intel 3 to 18A is a big jump.

N3E to N2 is expected to be minor jump only. This puts them on equal footing in density with 18A probably a bit behind in density and a bit ahead in performance due to BSPDN.

 

[SiliconFly]

.

 

[DavidC1]

Thats lower than expected. This should put 18A behind N2 in density somewhere between 10% to 15%. But should still put it ahead in performance.

Have you taken into account 18A is 15% when previous Intel claims were 15% and 10%?

 

[RTX]

That's just chip density without utilization.

 

[MadRat]

Imagine that, as you shrink distances between parts you can focus on signal attenuation and frequency. Who'd thunk.

 

[jpiniero]

TSMC's 2nm process will reportedly get another price hike — $30,000 per wafer for latest cutting-edge tech

2nm to cost 2X more than 4nm and 5nm

www.tomshardware.com

TSMC N2 might be over 30k a wafer. Which sounds about right.

 

[gdansk]

Really seems like a market opportunity for Intel or Samsung if either of them can beat N3 at a lower cost.