Shared pain shared gain: the start of a monopoly in leading-edge logic chip manufacturing ?

[siliconbruh999]

Both TSMC (Taiwan, ROC) and ASML (the Netherlands) are located in long-time U.S. allied countries. The U.S. doesn't need to find a counterbalance to TSMC from the PRC/CCP.

Yes but ASML is kind of way safer bet vs Taiwan

 

[jim_22222]

A monopoly is better than fake competetion for the customer.

A good single supplier beats 3 or 4 less good suppliers every day of the week.

And like any "competition" one supplier is usually always best and the followers get crumbs

how can it be good. as long as it's capitalism.

 

[Barnsley]

how can it be good. as long as it's capitalism.

DId you miss something off the end of your sentence?

 

[lefty]

they are talking monopoly. you know what does that mean. that means a customer has no bargain power.
Imagine there is only one phone brand one car brand in market.
you might need to pay 2k$ for a piece of iphone 13.

They could do that, but they don't because the know they would make many enemies. TSMC and ASML charge reasonable amount of mark up, despite being monopolies. BTW, Nvidia is probably charging it's customers 10 times higher mark up for it's AI products.

 

[MKWVentures]

The recent discussions around Intel Foundry Services in relation to the performance of TSMC reminds me to the situation some 10-20 years ago around leading-edge lithography and the battle between Nikon, ASML and Canon (see e.g. https://siliconsemiconductor.net/ar...non_Ever_Catch_ASML_in_the_Lithography_Market).

I think this is a good analogy but a oversimplification, my opinion based on my experience:

1) Intel Used Nikon (I was there before Nikon.... It was not fun, we are talking stone age)
2) Intel brought in ASML as second source
3) 2003-2008 ASML was much better, One could easily predict yields and speed based on whether you used Nikon or ASML
4) ASML was so much better, it was no longer possible to use Nikon for Critical layers.
5) ASML Delivered on all future developments. Nikon did not.
6) You can use Canon, or Nikon if you want, there is no monopoly.

Similarly:
1) TSMC was a non-leading edge supplier
2) TSMC delivered well on non-leading edge
3) TSMC delivered on leading edge
4) Others are still there, TSMC is just better... a lot better

anyone can buy from Samsung, UMC, GF, SMIC (if the government lets you). There is no monopoly

The age old argument is whether one company doing much better that competitors (despite starting behind them) is a monopoly. IMO ASML and TSMC are just better companies, with better employees, with better management, who deliver better. Same with Nvidia.

If you don't like TSMC or ASML or Nvidia being dominant, you can buy from others.

If Nvidia, Intel, AMD, Apple want to buy a stake in a new foundry to compete. Go fo it. Just realize that TSMC will still be better (even in Arizona)

Now, the leading company gets leading support from equipment vendors so that adds a tailwind. But Intel, IBM, Samsung, others were still able to lose leadership despite this.

Just an opinion

 

[user nl]

I think this is a good analogy but a oversimplification, my opinion based on my experience:

1) Intel Used Nikon (I was there before Nikon.... It was not fun, we are talking stone age)
2) Intel brought in ASML as second source
3) 2003-2008 ASML was much better, One could easily predict yields and speed based on whether you used Nikon or ASML
4) ASML was so much better, it was no longer possible to use Nikon for Critical layers.
5) ASML Delivered on all future developments. Nikon did not.
6) You can use Canon, or Nikon if you want, there is no monopoly.

Similarly:
1) TSMC was a non-leading edge supplier
2) TSMC delivered well on non-leading edge
3) TSMC delivered on leading edge
4) Others are still there, TSMC is just better... a lot better

anyone can buy from Samsung, UMC, GF, SMIC (if the government lets you). There is no monopoly

The age old argument is whether one company doing much better that competitors (despite starting behind them) is a monopoly. IMO ASML and TSMC are just better companies, with better employees, with better management, who deliver better. Same with Nvidia.

If you don't like TSMC or ASML or Nvidia being dominant, you can buy from others.

If Nvidia, Intel, AMD, Apple want to buy a stake in a new foundry to compete. Go fo it. Just realize that TSMC will still be better (even in Arizona)

Now, the leading company gets leading support from equipment vendors so that adds a tailwind. But Intel, IBM, Samsung, others were still able to lose leadership despite this.

Just an opinion

Thanks for the update on the history of the litho-battle@Intel!

To the best of my knowledge (I'm not working in semi), for leading-edge logic (so 7 nm and below) you really need EUV in combination with iArF, and that makes ASML effectively a full-monopolist in litho. No-one seems to really care anymore in the western world.

Now how about the leading-edge logic fabless customers like Apple, Nvidia, AMD, Mediatek, Qualcomm etc.
(https://www.tomshardware.com/tech-i...customer-accounting-for-11-of-revenue-in-2023)

When they are reasonably happy with the leading foundry performance of a geographically spread TSMC, why would they use Intel Foundry when IFS cannot compete in price, performance, ecosystem, innovation and trusted delivery?

What is the leading-edge logic foundry market share that IFS will need to keep up the coming 10-20 years with the needed investments and manufacturing innovation that (backed by multiple-governments in Taiwan, Japan, USA and EU) TSMC will provide?

Probably Korean backed Samsung will always provide foundry for logic, so "desperate second tier" fabless customers will be able to go there. The memory business of Samsung may help to financially support the foundry business, in combination with the national pride.

So, why is IFS needed, how will they be able to keep up with TSMC in the coming 10-15 years? Why would Taiwan, Japan, EU sink their government money/subsidies in IFS?

 

[jim_22222]

They could do that, but they don't because the know they would make many enemies. TSMC and ASML charge reasonable amount of mark up, despite being monopolies. BTW, Nvidia is probably charging it's customers 10 times higher mark up for it's AI products.

historical ly many companies have been subject ed to monopoly investigation. bell, Intel, ms etc.
there is a certain legal process to prove it.
take Intel as an example, with AMD, customer has opportunity to switch if they wanted to.
I am not a n expert on this, but it's definitely not what you think

 

[Jert]

Lots of interesting insights. Thoroughly enjoy reading the comments and appreciate the perspectives.

Very much agree with Mr. Nenni, we NEED Intel AND TSMC AND Samsung. IMHO, I do not think it's necessary for Intel to "beat" TSMC (an almost impossible task anyways regardless how much cash to burn), roughly on par or even 1 or 1.5 gen behind TSMC should be fine. Old timers may remember UMC did very well for many years as a strong second fiddle to TSMC (until later TSMC suddenly pulls away with FinFet).

"Everybody has a plan until they get punched in the face".

Again IMHO, I think it's paramount and to everyone's interest (particularly those of Apple, AMD, Qualcomm and the like) for TSMC to diversify and spread out geographically.

That punch could be triggered and come very suddenly (relative to the many years timescale needed to build advanced fabs and transfer process to outside of Taiwan), in the form of an "air, land, sea" blockade of Taiwan, by Winnie the Pooh who is sitting across the pond and everyday staring at the island. Conquering Taiwan is the only legacy worthwhile to "etch" his name into the Chinese history, because everything else is just petty and rubbish. He is 71, not too many years to sit idle...

I believe US Gov must figure out ways to keep Intel funded, at least to stay not too far behind TSMC, and prepare for conflict.

 

[hist78]

I think this is a good analogy but a oversimplification, my opinion based on my experience:

1) Intel Used Nikon (I was there before Nikon.... It was not fun, we are talking stone age)
2) Intel brought in ASML as second source
3) 2003-2008 ASML was much better, One could easily predict yields and speed based on whether you used Nikon or ASML
4) ASML was so much better, it was no longer possible to use Nikon for Critical layers.
5) ASML Delivered on all future developments. Nikon did not.
6) You can use Canon, or Nikon if you want, there is no monopoly.

Similarly:
1) TSMC was a non-leading edge supplier
2) TSMC delivered well on non-leading edge
3) TSMC delivered on leading edge
4) Others are still there, TSMC is just better... a lot better

anyone can buy from Samsung, UMC, GF, SMIC (if the government lets you). There is no monopoly

The age old argument is whether one company doing much better that competitors (despite starting behind them) is a monopoly. IMO ASML and TSMC are just better companies, with better employees, with better management, who deliver better. Same with Nvidia.

If you don't like TSMC or ASML or Nvidia being dominant, you can buy from others.

If Nvidia, Intel, AMD, Apple want to buy a stake in a new foundry to compete. Go fo it. Just realize that TSMC will still be better (even in Arizona)

Now, the leading company gets leading support from equipment vendors so that adds a tailwind. But Intel, IBM, Samsung, others were still able to lose leadership despite this.

Just an opinion

From a political point of view, I don't think TSMC will need to worry about this potential "monopoly" issue for the next 5 to 6 years, unless it behaves very poorly.

The reason is simple: concerns about national security, national interest, supply chain resilience, economic development, AI, and geopolitics all have much higher priority than the monopoly issue. The U.S., EU, Germany, Japan, and India have spent or will spend billions of dollars to attract TSMC and other semiconductor companies to build fabs in their territories for these reasons.

These same governments, which gave huge amounts of money to TSMC and similar companies, can't just turn around and sue TSMC without undermining their own policy objectives mentioned above.

 

[jorgequinonez]

Both TSMC (Taiwan, ROC) and ASML (the Netherlands) are located in long-time U.S. allied countries. The U.S. doesn't need to find a counterbalance to TSMC from the PRC/CCP.

That's not what I was talking about. I meant competition at the leading edge: where a company like Apple could go to TSMC vs. another one and actually have an option. A decade ago or so, companies could do something like that. You could go to Samsung for 14 nm vs. TSMC for 16 nm. Now, there is no option, only TSMC.

 

[hist78]

That's not what I was talking about. I meant competition at the leading edge: where a company like Apple could go to TSMC vs. another one and actually have an option. A decade ago or so, companies could do something like that. You could go to Samsung for 14 nm vs. TSMC for 16 nm. Now, there is no option, only TSMC.

How can I say? Should we blame TSMC or Intel or IBM or Samsung or Globalfoundries for lacking competition on the leading edge nodes?

Intel did have the money, technologies, and opportunities to be the one but it killed it by itself. It's a free market after all. The US government has limited ability to force any companies not to make mistakes.

 

[jorgequinonez]

How can I say? Should we blame TSMC or Intel or IBM or Samsung or Globalfoundries for lacking competition on the leading edge nodes?

Intel did have the money, technologies, and opportunities to be the one but it killed it by itself. It's a free market after all. The US government has limited ability to force any companies not to make mistakes.

True, we can't blame other companies' incompetence/bad luck for TSMC's success. I'm obviously out of my area of expertise. I had a thought that it maybe a future case of antitrust, but, after reading a little on the topic. There is no case for it.

 

[hist78]

True, we can't blame other companies' incompetence/bad luck for TSMC's success. I'm obviously out of my area of expertise. But, it remains to be seen if TSMC breaks any antitrust laws. When you get that big, like Standard Oil and Microsoft and others, you attract the attention of governments. Maybe TSMC and ASML are special cases where antitrust doesn't apply: an apple to orange comparison.

We can speculate about the antitrust legal issues TSMC may face, although, in my opinion, they are almost zero for the next 5 to 6 years.

The EU, US, and Japan have no appetite to address such issues in the coming 5 to 6 years. See my earlier post:

Shared pain shared gain: the start of a monopoly in leading-edge logic chip manufacturing ?

Both TSMC (Taiwan, ROC) and ASML (the Netherlands) are located in long-time U.S. allied countries. The U.S. doesn't need to find a counterbalance to TSMC from the PRC/CCP. Yes but ASML is kind of way safer bet vs Taiwan

semiwiki.com

 

[jim_22222]

True, we can't blame other companies' incompetence/bad luck for TSMC's success. I'm obviously out of my area of expertise. But, it remains to be seen if TSMC breaks any antitrust laws. When you get that big, like Standard Oil and Microsoft and others, you attract the attention of governments. Maybe TSMC and ASML are special cases where antitrust doesn't apply: an apple to orange comparison.

tsmc apparently can utilize its size and dominant position to create an unfair competition. intc tried to match its scale and half dead now.
also concern ing is its using some proxy companies to cover up its real size and capability. such as vanguard etc.

its investment in Singapore is a move to solidify its dominance in apac region. and monopoly of the engineer resources.

the few capable Chinese engineer s are like oil and Uranium to this industry and tsmc is taking them all and left competitors nothing to grab

 

[jorgequinonez]

We can speculate about the antitrust legal issues TSMC may face, although, in my opinion, they are almost zero for the next 5 to 6 years.

The EU, US, and Japan have no appetite to address such issues in the coming 5 to 6 years. See my earlier post:

Shared pain shared gain: the start of a monopoly in leading-edge logic chip manufacturing ?

Both TSMC (Taiwan, ROC) and ASML (the Netherlands) are located in long-time U.S. allied countries. The U.S. doesn't need to find a counterbalance to TSMC from the PRC/CCP. Yes but ASML is kind of way safer bet vs Taiwan

semiwiki.com

Thanks for the link. I ChatGPT'd the question “is TSMC's dominance of the fab market a possible antitrust case?” and the response I got was so convincing, I felt like there was zero chance at the moment of any antitrust, as you commented.

 

[jim_22222]

Thanks for the link. I ChatGPT'd the question “is TSMC's dominance of the fab market a possible antitrust case?” and the response I got was so convincing, I felt like there was zero chance at the moment of any antitrust, as you commented.

i doubt the version of gpt you asked.
it might still consider Intel a viable competitor which is apparently not the case with past two weeks fall out

hard to imagine Intel still has money to pay for more euvs with current financial situation. and compete

 

[user nl]

Thanks to all contributors with their comments/ideas. The main point of my original story "shared pain shared gain" was:

1) the transition in HVM to EUV litho around 2018-2019 for leading edge logic chip manufacturing has resulted in major changes and appears to still have quite important consequences.

2) the capital intensity and EUV-tooling experience / R&D that is needed to economically employ this EUV-technology has resulted in almost unsurmountable market barriers for new users of EUV-litho in logic chip manufacturing.

3) there are 3 leading edge logic manufacturers that now use EUV: TSMC, Samsung and Intel. Two are (original) IDMs (Samsung and Intel), one is Foundry only (TSMC).

4) it appears that EUV-litho is only long-term economically feasible when used in large scale and flexible Foundry operation/applications. The 2 IDMs have realized that and have been developing Foundry operation of EUV logic chip manufacturing.

5) the question now is: what is needed and optimal in the future for the leading edge logic chip manufacturing community in the western world: a monopoly (TSMC), a duopoly (TSMC + Samsung) or a triopoly (TSMC + Samsung + Intel-IFS).

6) a monopoly sounds scary, but looking at the history of the litho-market shake-out (ASML versus Nikon and Canon) has learned that the open-innovation systems-approach of ASML was the winning strategy. And the semi world continues to function more or less just fine after getting used to the trusted litho-monopoly of ASML. The base of ASML is in NL and its many other locations for subsystem manufacturing in Germany, USA guarantees a trusted supply line operated by this monopolist ASML. And with the US calling the shots regarding allowed customers of ASML's EUV tooling (no China).

7)Taiwan (23 Million people) is a relatively small country, similar like NL (18 Million people), so it has (quite) limited internal resources of people and finances. Taiwan and NL have quite a similar GDP of around 1 Trillion US$.

8) it appears to me that TSMC is using the same open-innovation systems approach in large-scale (now EUV-based) leading edge logic chip manufacturing. Since 2008 TSMC has developed this open-innovation approach: see https://www.tsmc.com/english/dedicatedFoundry/oip
At the same time TSMC has developed some hugely beneficial internal proprietary EUV tooling expertise (see eg https://pcge.eu/2024/05/tsmcs-euv-triumph-amplifying-tools-boosting-wafers-and-unmatched-pellicles/)

9) TSMC is rapidly diversifying it's geographical manufacturing base expanding to USA, EU and Japan. I think this approach is key in keeping the support of the western world related to supply chain and geopolitical worries around chips.

10) TSMC appears to aim for ~30% (relative to total revenue)) annual capital investments to keep up with demand. This rapid growth with healthy gross margin of around 50-60% appears to be economically sustainable, especially now that for the first time other governments like USA, Japan, EU are supporting TSMC financially with these huge capital investments.

11) the late entry of Intel-FS into the present duopoly market of TSMC and Samsung (IFS is estimated to receive some meaningful revenue only around 2027 (https://www.reuters.com/technology/...-meaningful-revenue-2027-cfo-says-2024-09-04/) makes one wonder how IFS will keep up with the relentless expansion/investments of TSMC during the coming 10 years.

12) if IFS is strategically/financially controlled by Intel and if the fabless Intel Product segment is really free to choose its optimum leading edge logic manufacturing foundry (as it now seems to do), what is the need to keep these 2 Intel units in a single company in the long-term? Looking at the success of the fabless leading edge chip developers (Apple, NVIDIA etc) there seems to be no need for tight single-house IDM's.

13) if TSMC can serve the leading-edge logic chip manufacturing well in the coming 5-10 years, why would private investors invest in IFS. There are many interesting investing options available for private investors?

14) I do not worry about regulatory issues for TSMC's effective monopoly in a EUV-based logic chip manufacturing market. The major western countries in this ecosystem (USA, EU, Japan) have already decided to invest tax-payers (!) money in TSMC.

15) trust takes a long time to develop and it seems TSMC is now the Trusted Foundry for the western leading edge logic manufacturing.

Will IFS get the time and funding that is needed to develop a similar trust to have a meaningful role to become a happy player for all its stakeholders as a (relatively) small player in a triopoly market governed/ruled by geopolitical supply considerations?
Or will the "shared pain shared gain" model that resulted in the current litho market (ASML, versus very small players Nikon and Canon) be also the final winning model in EUV-based leading edge logic chip manufacturing (TSMC, versus small players Samsung and IFS)?

Any further thoughts/comments are very much appreciated.

 

[jim_22222]

Thanks to all contributors with their comments/ideas. The main point of my original story "shared pain shared gain" was:

1) the transition in HVM to EUV litho around 2018-2019 for leading edge logic chip manufacturing has resulted in major changes and appears to still have quite important consequences.

2) the capital intensity and EUV-tooling experience / R&D that is needed to economically employ this EUV-technology has resulted in almost unsurmountable market barriers for new users of EUV-litho in logic chip manufacturing.

3) there are 3 leading edge logic manufacturers that now use EUV: TSMC, Samsung and Intel. Two are (original) IDMs (Samsung and Intel), one is Foundry only (TSMC).

4) it appears that EUV-litho is only long-term economically feasible when used in large scale and flexible Foundry operation/applications. The 2 IDMs have realized that and have been developing Foundry operation of EUV logic chip manufacturing.

5) the question now is: what is needed and optimal in the future for the leading edge logic chip manufacturing community in the western world: a monopoly (TSMC), a duopoly (TSMC + Samsung) or a triopoly (TSMC + Samsung + Intel-IFS).

6) a monopoly sounds scary, but looking at the history of the litho-market shake-out (ASML versus Nikon and Canon) has learned that the open-innovation systems-approach of ASML was the winning strategy. And the semi world continues to function more or less just fine after getting used to the trusted litho-monopoly of ASML. The base of ASML is in NL and its many other locations for subsystem manufacturing in Germany, USA guarantees a trusted supply line operated by this monopolist ASML. And with the US calling the shots regarding allowed customers of ASML's EUV tooling (no China).

7)Taiwan (23 Million people) is a relatively small country, similar like NL (18 Million people), so it has (quite) limited internal resources of people and finances. Taiwan and NL have quite a similar GDP of around 1 Trillion US$.

8) it appears to me that TSMC is using the same open-innovation systems approach in large-scale (now EUV-based) leading edge logic chip manufacturing. Since 2008 TSMC has developed this open-innovation approach: see https://www.tsmc.com/english/dedicatedFoundry/oip
At the same time TSMC has developed some hugely beneficial internal proprietary EUV tooling expertise (see eg https://pcge.eu/2024/05/tsmcs-euv-triumph-amplifying-tools-boosting-wafers-and-unmatched-pellicles/)

9) TSMC is rapidly diversifying it's geographical manufacturing base expanding to USA, EU and Japan. I think this approach is key in keeping the support of the western world related to supply chain and geopolitical worries around chips.

10) TSMC appears to aim for ~30% (relative to total revenue)) annual capital investments to keep up with demand. This rapid growth with healthy gross margin of around 50-60% appears to be economically sustainable, especially now that for the first time other governments like USA, Japan, EU are supporting TSMC financially with these huge capital investments.

11) the late entry of Intel-FS into the present duopoly market of TSMC and Samsung (IFS is estimated to receive some meaningful revenue only around 2027 (https://www.reuters.com/technology/...-meaningful-revenue-2027-cfo-says-2024-09-04/) makes one wonder how IFS will keep up with the relentless expansion/investments of TSMC during the coming 10 years.

12) if IFS is strategically/financially controlled by Intel and if the fabless Intel Product segment is really free to choose its optimum leading edge logic manufacturing foundry (as it now seems to do), what is the need to keep these 2 Intel units in a single company in the long-term? Looking at the success of the fabless leading edge chip developers (Apple, NVIDIA etc) there seems to be no need for tight single-house IDM's.

13) if TSMC can serve the leading-edge logic chip manufacturing well in the coming 5-10 years, why would private investors invest in IFS. There are many interesting investing options available for private investors?

14) I do not worry about regulatory issues for TSMC's effective monopoly in a EUV-based logic chip manufacturing market. The major western countries in this ecosystem (USA, EU, Japan) have already decided to invest tax-payers (!) money in TSMC.

15) trust takes a long time to develop and it seems TSMC is now the Trusted Foundry for the western leading edge logic manufacturing.

Will IFS get the time and funding that is needed to develop a similar trust to have a meaningful role to become a happy player for all its stakeholders as a (relatively) small player in a triopoly market governed/ruled by geopolitical supply considerations?
Or will the "shared pain shared gain" model that resulted in the current litho market (ASML, versus very small players Nikon and Canon) be also the final winning model in EUV-based leading edge logic chip manufacturing (TSMC, versus small players Samsung and IFS)?

Any further thoughts/comments are very much appreciated.

one point is tsmc is (or maybe)sqeezing ifs purposely, with its investment in az.
with deeper pocket it could poach many critical talent from ifs.
if ifs want to put a fab in taichung, guess how tw gvot would react.

the recent AMD design center in tw was asked to hire around fifty percent of all employee from out side.

 

[user nl]

one point is tsmc is (or maybe)sqeezing ifs purposely, with its investment in az.
with deeper pocket it could poach many critical talent from ifs.
if ifs want to put a fab in taichung, guess how tw gvot would react.

the recent AMD design center in tw was asked to hire around fifty percent of all employee from out side.

That is good both for AMD and for Taiwan! ASML in the Netherlands has a big problem of attracting enough skilled employees, NL is not large enough, the NL (and EU) does not provide enough skilled workers. The competition in high-tech is about people (not finances). A huge "little India" has developed in Eindhoven/Veldhoven where ASML is based . Quite interesting, now that India wants to start developing its chip industry.....Geopolitically of great importance to keep them connected to the western world.....

 

[jim_22222]

That is good both for AMD and for Taiwan! ASML in the Netherlands has a big problem of attracting enough skilled employees, NL is not large enough, the NL (and EU) does not provide enough skilled workers. The competition in high-tech is about people (not finances). A huge "little India" has developed in Eindhoven/Veldhoven where ASML is based . Quite interesting, now that India wants to start developing its chip industry.....Geopolitically of great importance to keep them connected to the western world.....

so you think Arizona is large enough to have multiple fabs in same neighbor hood. a lot of work ers like it for job opportunity.
intc maybe is just dumb.
or smart, they know tsmc can't compete with them.
what you think