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.

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


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

 

[Doug S]

You don't understand. Apple could've boosted up the smaller players like GloFlo by investing more in them. They could've prevented TSMC from becoming a monopoly. I wouldn't be surprised if bribes went from TSMC to certain decision makers at Apple to prevent them from making future looking decisions that could put TSMC at a disadvantage.

Wow is there any subject in which you can't find a way to inject your personal beliefs that Apple is corrupt in every way? What is it like to live in such a conspiracy theory filled head? Was Apple behind the Trump assassination attempt too?

GloFo was already behind the other foundries, and didn't have close to the capacity Apple required. If Apple has to give them a bunch of money to build that capacity and even more money to help them catch up process wise, they might as well have just broken ground on their own fab at that point. Even Intel didn't have enough capacity for Apple. The only two that did were Samsung and TSMC. Apple wanted to get away from Samsung, for obvious reasons. So they had only one choice, or two if you include the fanciful option of Apple building their own fab.

 

[adamge]

You don't understand. Apple could've boosted up the smaller players like GloFlo by investing more in them. They could've prevented TSMC from becoming a monopoly. I wouldn't be surprised if bribes went from TSMC to certain decision makers at Apple to prevent them from making future looking decisions that could put TSMC at a disadvantage.

GloFo's stunted ability to provide cutting edge transistors has been plain for all to see for 10 years now. It would have been a terrible business decision to get in bed with such a partner. Instead they chose the partner that has consistently been at or ahead of the space.

 

[Tuna-Fish]

Anyone seriously expects a significant change in Chip volumes considering the lead times to produce, install & ramp needed fabs. Also, the tiny matter of trained AND experienced staff.

Note that chip share and revenue share are two very different things. A few years back, someone tallied up that over it's existence TSMC had made more net profit than the entire contract foundry sector, TSMC included, did in the same time. That is, everyone else put together made a loss. TSMC had only produced ~20% or so of the total silicon volume over that time.

The massive revenue share is driven by the fact that there are two kinds of nodes: The ones where the customer can get competitive bids from multiple foundries for similarly performing silicon, and the ones where you can only go to one place to get the performance. And for logic silicon, TSMC pretty much has a monopoly on the second kind. There is a very significant price difference between those two types of processes.

 

[oak8292]

You don't understand. Apple could've boosted up the smaller players like GloFlo by investing more in them. They could've prevented TSMC from becoming a monopoly. I wouldn't be surprised if bribes went from TSMC to certain decision makers at Apple to prevent them from making future looking decisions that could put TSMC at a disadvantage.

Gloflo never got their planar 20 nm to market and AMD had to write off all the development work they did on that node. Gloflo went to Samsung to get a working 14 nm FinFET process. Gloflo was part of a consortium between IBM, Gloflo and Samsung. Gloflo was not a serios option, they had even less wafer volume than Samsung.

When Apple moved to TSMC they needed about 40K wspm and only TSMC was really big enough to do that on a new node at ‘Moore’s Law’. Even Samsung was not running that fast on logic. The Austin Tx facility that Apple was using was only about 40K wafer starts per month.

When Apple went to TSMC they traded places with Qualcomm who went over to Samsung. Qualcomm was TSMC largest customer at over 20% of revenue prior to Apple. Qualcomm just wanted the lowest cost wafers to compete with Mediatek. Over three years Apple and Qualcomm switched wafer suppliers. Where is Qualcomm getting wafers today?

 

[jpiniero]

Note that chip share and revenue share are two very different things. A few years back, someone tallied up that over it's existence TSMC had made more net profit than the entire contract foundry sector, TSMC included, did in the same time. That is, everyone else put together made a loss. TSMC had only produced ~20% or so of the total silicon volume over that time.

The massive revenue share is driven by the fact that there are two kinds of nodes: The ones where the customer can get competitive bids from multiple foundries for similarly performing silicon, and the ones where you can only go to one place to get the performance. And for logic silicon, TSMC pretty much has a monopoly on the second kind. There is a very significant price difference between those two types of processes.

TSMC still makes a third of it's revenue on obsolete nodes.

 

[Doug S]

Gloflo never got their planar 20 nm to market and AMD had to write off all the development work they did on that node. Gloflo went to Samsung to get a working 14 nm FinFET process. Gloflo was part of a consortium between IBM, Gloflo and Samsung. Gloflo was not a serios option, they had even less wafer volume than Samsung.

When Apple moved to TSMC they needed about 40K wspm and only TSMC was really big enough to do that on a new node at ‘Moore’s Law’. Even Samsung was not running that fast on logic. The Austin Tx facility that Apple was using was only about 40K wafer starts per month.

When Apple went to TSMC they traded places with Qualcomm who went over to Samsung. Qualcomm was TSMC largest customer at over 20% of revenue prior to Apple. Qualcomm just wanted the lowest cost wafers to compete with Mediatek. Over three years Apple and Qualcomm switched wafer suppliers. Where is Qualcomm getting wafers today?

So actually it was Qualcomm who could have / should have "boosted GloFo" according to Igor's "logic".

 

[Aapje]

TSMC still makes a third of it's revenue on obsolete nodes.

It's part of their business model to keep older nodes running, to be able to amortize over a longer period. Intel is moving to the same model, instead of only producing the newest nodes.

 

[igor_kavinski]

So actually it was Qualcomm who could have / should have "boosted GloFo" according to Igor's "logic".

They are a puny player. They never had enough to be tripping over cash on hand: https://www.cnbc.com/2020/01/28/apple-q1-2019-cash-hoard-heres-how-much-cash-apple-has-on-hand.html

Apple had the means to ensure they didn't have to depend on just one foundry but they just didn't care that much. Even now, if their investors ask them about the Taiwan situation, what are they going to tell them? What's their backup plan?

 

[moinmoin]

Yes, TSMC was competitive. And competitive is a good thing. But putting all eggs into that basket is a disaster waiting to happen. Who could have made chips? With billions floating around from governments and if only companies were willing to use multiple foundries, it could have been Samsung or GlobalFoundries. But no, most of tech dumped all money into TSMC--basically giving them a monopoly. And the more money going to them the worse the problem gets.

This is what capitalism is essentially all about: Optimize for the best case to eek out even the tiniest margin increases. Multi-sourcing is so old school, who does that anymore in times where everybody is eager to do perfect JIT manufacturing. Optimizing for worst cases is leaving money on the table, so better left to countries to worry about. The countries will shoulder all the costs anyway in case of disasters of any kind, so why should any company bother about worst cases? (semi /s)

 

[poke01]

They are a puny player. They never had enough to be tripping over cash on hand: https://www.cnbc.com/2020/01/28/apple-q1-2019-cash-hoard-heres-how-much-cash-apple-has-on-hand.html

Apple had the means to ensure they didn't have to depend on just one foundry but they just didn't care that much. Even now, if their investors ask them about the Taiwan situation, what are they going to tell them? What's their backup plan?

Doesn’t work like that… Samsung has all the cash in world and yet cannot make decent volume nodes and Intel is a failure too.

Apple used to use Samsung before TSMC and ever since the A10 they used TSMC because only could deliver on Apples volume and only TSMC has leading edge nodes that could be used for the iPhone.

Global Foundries was an aging relic much like the forgotten Xerox. Times change and people move on.

The better question to ask is why didn’t Intel nor Samsung do a better job than TSMC?
All Intel and Samsung do is hype and hype
but it’s all for show.

Apple’s not a charity to fix other companies problems due the inability to innovate and bad management led them to stagnation.

AMD, Apple, Qualcomm. mediatek, Nvidia and soon Google and even Intel use the TSMC foundry if TSMC/Taiwan falls the PC/tech industry falls and trust me it won’t just be Apple investors who would be worried but the whole world sectors.

 

[jdubs03]

I think with 18A Intel will get back on the right track. From what has been put out there so far, its been pretty positive.
They’ll be on the top tier node that should at least be able to fill their own supply. Then it’s a matter of expansion and bringing on additional customers. We should be rooting for them; it is a much needed diversification off a geopolitical choke point. If they can prove their competence with yields I see no reason why they cannot be successful. It wouldn’t be absurd if then Apple gave them a look.

 

[oak8292]

This is what capitalism is essentially all about: Optimize for the best case to eek out even the tiniest margin increases. Multi-sourcing is so old school, who does that anymore in times where everybody is eager to do perfect JIT manufacturing. Optimizing for worst cases is leaving money on the table, so better left to countries to worry about. The countries will shoulder all the costs anyway in case of disasters of any kind, so why should any company bother about worst cases? (semi /s)

Countries like China and the U.S. that have a sense of ‘controlling’ their own semi industry are fooling themselves. Do they have a replacement for ASML, which is essentially a monopoly?

Both supply and demand are global in semi’s. Taiwan is a very large player in the latest growth spurt primarily because consumer demand for smartphones drove the industry. Military and corporate users are not driving development of new nodes. It is small die for consumer goods. The internet is consumer focused and smartphones are consumer focused. These are global.

Countries need to break up the internet to make it more resilient. Countries need to develop their own OS to provide diversity. Every country needs to develop their own telecom systems to give more diversty. etc. etc. sarcasm

 

[igor_kavinski]

Countries need to break up the internet to make it more resilient. Countries need to develop their own OS to provide diversity. Every country needs to develop their own telecom systems to give more diversty. etc. etc. sarcasm

I can just imagine the fun of browsing the "international" internet.

Visiting Google US, pop up: "Click Yes to enter the Internet zone of USA."

Visiting Google China, pop up: "Please enter your GPS co-ordinates and your Chinese National ID number and wait while we verify your access".

Visiting Google North Korea, pop up: "There is no such Internet Zone. Click Yes to close tab or Click No to visit some other Internet Zone".

 

[oak8292]

I can just imagine the fun of browsing the "international" internet.

Visiting Google US, pop up: "Click Yes to enter the Internet zone of USA."

Visiting Google China, pop up: "Please enter your GPS co-ordinates and your Chinese National ID number and wait while we verify your access".

Visiting Google North Korea, pop up: "There is no such Internet Zone. Click Yes to close tab or Click No to visit some other Internet Zone".

Just as absurd is saying that Apple needed to use Gloflo to prevent TSMC becoming a monopoly.

“Apple could've boosted up the smaller players like GloFlo by investing more in them. They could've prevented TSMC from becoming a monopoly”

Why didn’t Intel use Gloflo to prevent TSMC from becoming a monopoly. Intel had a lot of free cash flow that they used to buy back shares. Intel could have diversified their suppliers and taken wafers from AMD. Intel wouldn’t have been dependent on their own failing nodes like 10 nm and they would have strengthened the countries supply chain. You know for insurance. sarcasm

Seriously, business decisions are made for business reasons. There are times that diversity of supply, diversity in work force and diversity in products is the right thing to do. It can cost more but it can pay back. However, like most insurance there are also times it just costs money, sunk cost.

Intel currently has a wafer supply agreement with TSMC which was set up to finish Aurora. I think Pat would like to get out of it because he would like to spend that money on Intel wafers. However, this is a business decision that was made when Intel 7 nm EUV or now 4 nm was not yielding for Ponte Vecchio.

Also, Intel has also bought a number of companies that use TSMC for wafers and many of these never move over to Intel manufacturing. Gaudi is still on TSMC. I believe that even a lot of Altera die are still from TSMC. Should these be off TSMC to prevent TSMC from becoming a ‘monopoly’.

 

[DrMrLordX]

Apple had the means to ensure they didn't have to depend on just one foundry but they just didn't care that much. Even now, if their investors ask them about the Taiwan situation, what are they going to tell them? What's their backup plan?

You overestimate their importance. Apple would require state backing (US + EU) to achieve this goal. Either to create a competitor from whole cloth, or to bribe TSMC into licensing some of their foundry tech to a second source.

 

[igor_kavinski]

...to bribe TSMC into licensing some of their foundry tech to a second source.

Did the Chinese bribe TSMC to get to 7nm with DUV? Apple could've invested in some expensive EUV machines, given them to GloFlo to experiment with and surely something would've come out of that gamble. $400 million isn't a lot for Apple, in order to ensure the future of their chips.

 

[SiliconFly]

... Apple could've invested in some expensive EUV machines, given them to GloFlo to experiment with and surely something would've come out of that gamble. $400 million isn't a lot for Apple, in order to ensure the future of their chips.

I don't think it's just the money alone. It's the capability. They're just etching machines. Building the libraries and getting the yield is more of an art than science. Requires significant R&D, IP ecosystem and customer confidence which requires track record, none of which they possessed.

 

[igor_kavinski]

Building the libraries and getting the yield is more of an art than science. Requires significant R&D, IP ecosystem and customer confidence which requires track record, none of which they possessed.

Show of faith can act as a miracle and transform ordinary people into geniuses.

 

[maddie]

Show of faith can act as a miracle and transform ordinary people into geniuses.

Nope.

 

[oak8292]

Deleted

 

[oak8292]

Did the Chinese bribe TSMC to get to 7nm with DUV? Apple could've invested in some expensive EUV machines, given them to GloFlo to experiment with and surely something would've come out of that gamble. $400 million isn't a lot for Apple, in order to ensure the future of their chips.

Here is the man you are looking for Liang Mong-song. Liang was very high up in R&D at TSMC. He went to work first at Samsung and then he became co-CEO at SMIC. TSMC sued him and won in Taiwan but he had already transferred a lot to Samsung on FinFET technology.

https://www.eetimes.com/leaker-of-tsmc-secrets-joins-smic-as-co-ceo/

Gloflo had EUV equipment in Fab 8 when Anandtech visited in 2018. A video and article are still here on the site. Gloflo did not have the wafer volume to go to EUV. Apple wasn’t going to bail them out. You also have to remember that Gloflo was privately held by Mubadala, a private equity firm based in Abu Dhabi’s. Global Foundries was only spun out as an IPO in 2021.

 

[NostaSeronx]

Gloflo did not have the wafer volume to go to EUV. ... Global Foundries was only spun out as an IPO in 2021.

GlobalFoundries did not have the customer base to support an EUV roadmap. Most of the big customers switched to TSMC (AMD) or Samsung (IBM) before 7LP even finalized. No customers = no money, as GloFo to use long-term agreements as a loan would need $15B+ in LTAs to support Fab 8.2 as a EUV FAB. Which with the reduced order size from AMD and IBM wouldn't cover anywhere close enough. As Mubadala's $10B investment into the Samsung 14FF plan hasn't actually made a full return yet. As Malta was ran for a loss (extended cost on top of investment) from 2015 to 2022 for 14LPP/12LP/12LP+. Malta only became profitable when the 45nm node was brought over.

"We're also diversifying our manufacturing footprint by accelerating the transfer of technologies such as 22 FDX, 28 nanometer high voltage and 40 nanometer ESF 3 into our FAB eight facility in Malta, New York." - Q1 2024 Earnings call. The big one is that 22FDX finally officially getting the port over to Malta. As it is the highest design win(long-term agreement) revenue with highest profit/operating margin out of all nodes currently. US fab -> US customers, majority of the revenue. Where EU Fab -> EMEA customers, minority of the revenue.

Even with the IPO, Mubadala still owns >80% of Globalfoundries. The only other group that can have a vote is Fidelity Investments.

 

[igor_kavinski]

Even with the IPO, Mubadala still owns >80% of Globalfoundries. The only other group that can have a vote is Fidelity Investments.

Apple needs to hire world class belly dancers to woo them into submission.

 

[SiliconFly]

This just in:

https://twitter.com/x/status/1814935845960581429

Apparently, there are no TSMC N2 products next year! Q4 2025 is way too late in the cycle.

On the other hand, we have 20A coming out late this year. And 18A products shipping mid next year.

Process leadership in 2025 is happening after all.

 

[igor_kavinski]

Process leadership in 2025 is happening after all.

Yeah, process leadership and performance minimization leadership through downclocking.