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.

TSMC Radically Boosts CapEx to Expand Production Capacities, To Reach $14B For 2019

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.

Samsung’s Aggressive EUV Plans: 6nm Production in H2, 5nm & 4nm On Track

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.

 

[lightisgood]

This isn’t engineers this is skilled craftsman. Here is a quote from ASML on installation of EUV.

Peter Wennink

Good question. We need to realize it. If you look at the reasons, predominantly the push-out had to do with fab readiness and that was basically driven by construction skills. And you think, "Well, how can that be?" You just hire a couple of construction workers and you just build the fab. Well, just building a $20 billion fab that's going to do a 5- or a 3- or a 2-nanometer product is a skill and people don't seem to realize that when we start building, those fabs across the globe now and are everywhere, that skill has been refined over the last couple of decades in only a few places on the planet, and predominantly in Taiwan and in Korea and a bit in China.

Now having to do that now and accelerate this will lead to all kinds of issues, because we are still building those fabs in Korea and in Taiwan, but also in other places on the planet, also in the U.S., for instance. And so, getting access to the requisite skills and skilled workers to keep the construction plan on time is a challenge, as at least what customers tell us, and this is the main reason. So, you can easily look at a delay of a couple of months or a quarter.

ASML Holding N.V. (ASML) Q2 2023 Earnings Call Transcript

ASML Holding N.V. (NASDAQ:NASDAQ:ASML) Q2 2023 Earnings Conference Call July 19, 2023 9:00 AM ETCompany ParticipantsSkip Miller - Vice President, Investor...

seekingalpha.com

AZ Union Fires Back at TSMC's 'Offensive' Claims of Unskilled U.S. Workers

'These workers have been building fabs for Intel for decades,' the head of union claims

www.tomshardware.com

> 'These workers have been building fabs for Intel for decades,' the head of union claims

Frankly, TSMC is cheat, is'nt it?
In addition, I find TSMC plotting for fleeing from the environmental impact assessment in Kumamoto, Japan.

熊本県はTSMC(JASM)の環境アセスメントをしない方針｜Green Sustainability Kumamoto

熊本県の菊陽町に台湾の大手半導体製造企業TSMCの熊本工場JASMが建設中である。 規模も大きく、排ガス・排水量も莫大であるこの工場。排出される有害物質も甚大な量になることが予想される。 しかし、なんと熊本県は、環境アセスメントも健康リスク評価もしないというのだから、開いた口が塞がらない。 台湾でさえ環境アセスメントや健康リスク評価している ただし、内容は杜撰である・・・ 台湾においては、形式上、環境アセスメントや健康リスク評価はなされている。ただし、適正に行われず、周辺環境は住民に多大な被害を出している。 『新新聞』の調査取材によって、半導体業界において、使用する危

note.com

 

[oak8292]

AZ Union Fires Back at TSMC's 'Offensive' Claims of Unskilled U.S. Workers

'These workers have been building fabs for Intel for decades,' the head of union claims

www.tomshardware.com

> 'These workers have been building fabs for Intel for decades,' the head of union claims

Have they installed a EUV machine that yields?

 

[lightisgood]

Arizona workers accuse TSMC of fatalities at 'dangerous' construction site

Construction workers building TSMC's Arizona plant allege that it's the "most unsafe site," and accuse the chip maker of injuries, fatalities — and reneging on pay deals.

appleinsider.com

Clearly, the trigger of this conflict between TSMC and Arizona union had arisen from slavery in Fab 21 construction site.

TSMC says “Ah… we demand skills about installing EUV tools…”, NOW ?
It's too late now, isn't it?
This is an empty logic.

 

[sdifox]

Arizona workers accuse TSMC of fatalities at 'dangerous' construction site

Construction workers building TSMC's Arizona plant allege that it's the "most unsafe site," and accuse the chip maker of injuries, fatalities — and reneging on pay deals.

appleinsider.com

Clearly, the trigger of this conflict between TSMC and Arizona union had arisen from slavery in Fab 21 construction site.

TSMC says “Ah… we demand skills about installing EUV tools…”, NOW ?
It's too late now, isn't it?
This is an empty logic.

TSMC : we don't want a US fab cuz it is too expensive there.

AZ: We'll give you a sweet deal.

TSMC : Fine

Union: why are we cut out?

TSMC: sigh

 

[A///]

construction site is dangerous? oh no, what's next the sun in the summer in the desert is too hot?

 

[Ajay]

construction site is dangerous? oh no, what's next the sun in the summer in the desert is too hot?

Actually, yes. The sun (temperature) in the desert is getting too hot. Also, construction sites do not need to be 'too' dangerous. Sounds to me more like International Semiconductor Giant doesn't like the fact that American workers have agency. Wait till they get to Germany - TSMC is going to faint.

 

[Hitman928]

Actually, yes. The sun (temperature) in the desert is getting too hot. Also, construction sites do not need to be 'too' dangerous. Sounds to me more like International Semiconductor Giant doesn't like the fact that American workers have agency. Wait till they get to Germany - TSMC is going to faint.

From my reading of the situation, it's not that the union/workers are complaining that the construction site is inherently dangerous (all construction sites are) but rather they are claiming that TSMC is not following U.S. regulations/standards on safety at the construction site, making the site more dangerous than it should be.

 

[Ajay]

From my reading of the situation, it's not that the union/workers are complaining that the construction site is inherently dangerous (all construction sites are) but rather they are claiming that TSMC is not following U.S. regulations/standards on safety at the construction site, making the site more dangerous than it should be.

Yes.

 

[lightisgood]

Arizona union destroys the cheat of TSMC.

Econ Commentators Join TSMC to Declare U.S. Workers’ Premature Defeat

The econ blogosphere hasn’t really checked, but knows in its heart American workers aren’t up to the job.

prospect.org

> Smith’s one specific claim is that American workers can’t install lithography machines.
> In fact, union pipefitters are already installing lithography machines at TSMC, as well as nearby Intel.
> (The equipment is not specific to TSMC. Both use equipment from the same Dutch vendor: ASML.) Workers have their own thoughts on technical details of the job—on which, more below.
>
> The broader point is that, with all the concern about how Americans have lost their dynamism and mojo, there’s a risk of mystifying the task at hand.
>
> “What pipefitters, electricians, and others do is hook up equipment,” Aaron Butler, president of the Pipefitters’ local and head of the Arizona Building Trades Council, told me.
> “The pipe does not care what it’s going to. My guys don’t make the chips, they’re not trying to run the lithography equipment.
> We connect the piping to the mechanical connection, pressure-test the line, make sure it’s clean and quality, and walk away.”

> Ironically, TSMC may be crushing the opportunity for learning-by-doing to emerge in the construction stage of the fab.
> TSMC has outsourced most of the job site—like its janitorial work—to temp contractors with little training and high churn.
> Even where it has brought in union workers with a career in the building trades, it has done so grudgingly and kept them on precariously, disincentivizing investment in training and growth.
>
> The company, in other words, has opted to work wherever possible with the least-skilled segment of the American workforce.
> Now, as it turns around to bring in its own skilled workers, would-be supporters of industrial policy are declaring that this was inevitable.

> IN THE CASES WHERE TSMC HAS USED UNION WORKERS, workers in the building trades have plenty of notes on how they’re doing the job.
>
> Luke Kasper of the sheet metal workers union, for example, worries about the quality of the Teflon-coated stainless steel ductwork.
> Where non-union contractors have done the work, he says, gaskets are missing in areas, and not all of the ductwork achieves an airtight seal.
> “Even though it’s negative-pressure exhaust ductwork, it’s not supposed to be done like that,” he said. Corrosive exhaust passing through the ductwork could leak.
>
> Butler, with the Pipefitters, described ongoing disagreements over welding techniques. Polyvinylidene fluoride (PVDF) pipes are the arteries of a fab, carrying pure, ultraviolet-treated water to the clean room.
> Arizona Local 469 pipefitters have been installing mechanical piping in clean rooms since Motorola production sites sprung up in the 1980s, and many members learned to weld the PVDF by hand, using socket welding, which requires a paddle.
>
> Butler says socket welding can be leaky, and he favors the Bead and Crevice Free (BCF) process, which is automated and produces data on quality.
> But TSMC favors socket welding, so they have continued to use the older technique.

 

[Saylick]

Looks like the Intel/Tower Semi deal is dead.

Intel scraps $5.4 bln Tower deal after China review delay

Intel Corp scrapped its $5.4 billion deal to buy Israeli contract chipmaker Tower Semiconductor Ltd after their merger agreement expired without regulatory approval from China.

www.reuters.com

 

[Panino Manino]

Looks like the Intel/Tower Semi deal is dead.

Intel scraps $5.4 bln Tower deal after China review delay

Intel Corp scrapped its $5.4 billion deal to buy Israeli contract chipmaker Tower Semiconductor Ltd after their merger agreement expired without regulatory approval from China.

www.reuters.com

How bad is this news for Intel?
They needed Tower to "survive"?

 

[jpiniero]

How bad is this news for Intel?
They needed Tower to "survive"?

Intel wanted Tower to run IFS. That the deal wasn't going to happen because of China isn't new news, Intel's already gone in a different direction.

 

[lightisgood]

Looks like the Intel/Tower Semi deal is dead.

Intel scraps $5.4 bln Tower deal after China review delay

Intel Corp scrapped its $5.4 billion deal to buy Israeli contract chipmaker Tower Semiconductor Ltd after their merger agreement expired without regulatory approval from China.

www.reuters.com

Intel was forced to modify their strategy...
What does Intel do next ?
Is it tried similar and smaller M&A which doesn’t need regulatory approval ?

 

[Ajay]

How bad is this news for Intel?
They needed Tower to "survive"?

Well, the pitch to investors is always - we absolutely need to spend $5B to assure IFS succeeds. When it falls through, the pitch becomes - we anticipated this and are prepared to move on without this investment. Hopefully, Intel has been poaching top foundry talent in various categories as ‘preparation'.

 

[Thunder 57]

Intel was forced to modify their strategy...
What does Intel do next ?
Is it tried similar and smaller M&A which doesn’t need regulatory approval ?

More layoffs.

Intel To Make Further Workforce Cuts in U.S., Report Says

Intel cuts research and development workforce to reduce costs.

www.tomshardware.com

 

[Thibsie]

More layoffs.

Intel To Make Further Workforce Cuts in U.S., Report Says

Intel cuts research and development workforce to reduce costs.

www.tomshardware.com

Ouch.

 

[A///]

cross your fingers it's low performers or people who've been coasting.

 

[Hitman928]

More layoffs.

Intel To Make Further Workforce Cuts in U.S., Report Says

Intel cuts research and development workforce to reduce costs.

www.tomshardware.com

Gelsinger promised the employees to restore salaries and bonuses by the end of September. He also promised investors to continue to cut costs. Unfortunately, this is part of the logical conclusion to be able to fulfill both of those promises.

 

[Doug S]

Well, the pitch to investors is always - we absolutely need to spend $5B to assure IFS succeeds. When it falls through, the pitch becomes - we anticipated this and are prepared to move on without this investment. Hopefully, Intel has been poaching top foundry talent in various categories as ‘preparation'.

Intel doesn't need "top foundry talent", they need fabs that handle older processes. A foundry can't offer only leading edge and recent processes, at least not if they want to compete with TSMC. That's what they were trying to buy with Tower, not so much the talent.

 

[Ajay]

Intel doesn't need "top foundry talent", they need fabs that handle older processes. A foundry can't offer only leading edge and recent processes, at least not if they want to compete with TSMC. That's what they were trying to buy with Tower, not so much the talent.

Obviously, I have a pretty bad memory at time. So, how do they plan to work around this?

 

[Joe NYC]

Obviously, I have a pretty bad memory at time. So, how do they plan to work around this?

Buying Global Foundries, in theory, but China will sink that one as well.

 

[Doug S]

Obviously, I have a pretty bad memory at time. So, how do they plan to work around this?

If they can't buy an existing foundry, they will have to buy trailing edge fabs and assemble their portfolio over time. I don't know how easy that would be to do, but they do come on the market now and then. Apple bought a fab in Silicon Valley from Maxim like 7 years ago that was doing like .18u to .50u production for MEMS devices, that supposedly they intended for research rather than actually manufacturing stuff to be used in their products. Haven't heard anything about it for a long time, but if a famously secretive company like Apple is using it for internal research it is never going to make the news lol

 

[NostaSeronx]

Buying Global Foundries, in theory, but China will sink that one as well.

GlobalFoundries is looking to gut its mature nodes (all nodes not at 45nm). So, there might be fabs for the taking that GloFo isn't willing to upgrade.

The nodes being kept for the 2025+ timetable.
45nm // 4th
28nm // 2nd Biggest Revenue *
22nm // Biggest Revenue *
12nm // 3rd *
*These three nodes account for ~55% of the revenue at GloFo; 22FDX#1, 28BLK#2, 12LP#3.

The funnel is 180nm/130nm/90nm -> 45nm/28nm/22nm, so at some point 180nm/130nm/90nm is expected to be <1% of revenue @ GF.
22FDX-BIC replaces SiGe nodes.
22FDX-BCD replaces BCD/BCDLite nodes.
22FDX-HV/ISP/DDI replaces other nodes. And other (45nm/28nm) conversions which accounts for the other ~45%.

The general leap:
2015 Imsys IM3000 ~367 MHz on 65nm (3.5 mm2)
2023 Imsys IM4000 ~1 GHz on 22FDX <== This one already has a ~2 GHz target (w/ 4096 core die ~2xx mm2) on 12FDX.

Everything is pretty much getting crushed onto 22FDX. DoD is dropping most mature nodes for Malta's 22FDX.

Anyone asking for when "12FDX" the European chips act via European Commission states: "GlobalFoundries (GF) is developing 12nm FD-SOI at Dresden; it will be production ready in 2023-24." Where as the agreement for GlobalFoundries includes "10nm/7nm/5nm" FDSOI and "3nm" Next-Gen FDSOI(complementary stacked FDSOI) nodes as well. Of which, the US chip act/DoD side, GlobalFoundries was told to focus on a FDSOI roadmap. GF is getting double teamed, but at least it should offer cash.

Edit: Qualcomm is likely to the biggest user of 12FDX.

Speculation: Snapdragon 2/2c;
1x P670 cores
3x P470 cores
Adreno iGPU

With KaiOS 4.0(Firefox OS (Firefox version 94+ supports RISC-V => FirefoxOS supports RISC-V)) touchscreen-based phones returning from the pits of hell:

 

[deasd]

Qualcomm may have stopped designing chips for the Intel 20A node, meaning that Intel 18A R&D and…

My latest survey indicates that Qualcomm has stopped developing Intel 20A chips. The lack of cooperation with a first-tier IC design vendor…

medium.com

Regarding news about Qualcomm cancelled 20A partnership to Intel, here's also an old news being unearthed which support the claim that Qualcomm do have a plan to cooperate with Intel on 20A node, which people forgot:

Intel’s First High-Profile IFS Fab Customer: Qualcomm Jumps on Board For 20A Process

www.anandtech.com

 

[qmech]

Intel doesn't need "top foundry talent", they need fabs that handle older processes. A foundry can't offer only leading edge and recent processes, at least not if they want to compete with TSMC. That's what they were trying to buy with Tower, not so much the talent.

Intel doesn't *need* fabs that handle older processes. They need to increase the volume of the current processes and increase the life of those processes. The latter seems to be conflated with "Intel needs older processes". It might be *nice* in the short term if Intel could offer mature foundry options, but only insofar as it helps with the above.

Look at the economics. Intel is spending billions on equipment and R&D. They need to increase the revenue they get from those expenses. The foundry way of doing that is to keep old fabs producing chips for customers that don't require the latest and greatest. Buying an existing foundry like Tower, or buying old Fabs from someone, doesn't do anything to increase the revenue from the money Intel has already spent on Intel7, Intel4 etc.

There might be some value in buying access to a wide variety of older processes if it helps bring in customers to Intel's own existing processes - either now or in the future. It is an indirect value, though, and there are other ways of accomplishing the same.

What Intel primarily wanted from the Tower deal appears to have been a foundry management team with contacts throughout the industry and experience in working with those customers. This is something Intel sorely needs and will need to slowly develop instead of simply buying it. There was also some interesting IP, but anything that wasn't related to Tower's specific processes could almost certainly be licensed by Intel, making it a non-issue.

There is a tendency in the popular press to underestimate how intimately foundries work with their customers, particularly large customers. Intel has almost zero experience in this area.