Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/will-tsm-crush-all-others.14182/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Will TSM Crush all others?

Arthur Hanson

Well-known member
With TSM having the clear leader in most of the semi technologies and expanding it, it gives TSM the advantage of being able to use its knowledge and equipment base for a longer period of time, lowering its cost compared to others substantially. On top of this is its economies of scale in just about everything, lowering its costs even further. Also being the clear leader in technology and the fact that it doesn't compete with its customers, lowers its marketing costs, which I suspect are also the lowest in the industry, the advantages TSM has been overwhelming, and the leads in these areas will only increase. I haven't even covered all the advantages TSM has or the fact that TSM knows how to use compounding of resources on a scale unmatched perhaps in manufacturing history and the only way for the rest to catch up or for TSM to make some major mistakes. TSM has become a compounding and utilization machine unmatched in manufacturing history, much as Henry Ford revolutionized manufacturing in transportation. Any thoughts, comments, or additions solicited and welcome.
 
Last edited:
I want to see more actions from Intel and Samsung's about how they develop their foundry ecoystems. Odds are stacked against both though. Firstly, it will take years. TSMC started back in 2008. Secondly, Intel, with its "only the paranoid survive", may be able to overcome the issues at 7nm and close the gap a bit. But on the foundry side, the paranoid mentality won't help build a functioning ecoystem/market-place for its partners and clients. They dont need to be reminded what happened to the suppliers in the wintel days. Samsung has a similar culture to Intel on this regard.
 
Anybody concerned that Bejing takes over in Hong Kong and Tiawan? Biden is right to invest in semi manufacturing as a strategic resource.
It's certainly a concern, but even there, TSMC seems to be making good strategic decisions. They're continuing to invest heavily in Taiwan-based manufacturing while hedging somewhat with manufacturing outside Taiwan, including in the US. The Taiwan concentration makes it tough for China to invade and potentially disrupt worldwide chip supplies including chips China needs as much as anyone. This may be the best implementation of "too big to fail" we've seen yet!
 
Rule of investing in tech is when there is a clear best of breed company in a segment, that's the only one you should bother investing in. Just look how much better Google did vs Yahoo, Amazon vs Ebay, or Facebook vs Twitter/Snap, or even Intel vs AMD in the 1990s. Unfortunately for Intel TSMC is now the clear best of breed in semis.
 
The West needs a strategic second source at all levels of thetechnology. IT does not necessarily have to be the best - just good enough. The problem is how to get there, and whom to chose as the lead partner

Whatever Intel says, its success has historically been due to "seizing the moment" as much as technical strategy*. Indeed, they have been quite explicit about how this worked in the early days. If IBM corporate had been serious about PCs and willing to put in the spadework neither Intel's processors nor Microsoft's operating system would have become dominant as they have.
I know nothing of Intel's current generation of semiconductor engineers, but it seems that Intel appear to have lost the technical edge that they once had; I'm not equipped to judge whether this is a matter of financial focus or the competition's technical superiority. One thing is clear: Intel has blown lukewarm and cold on foundry services over the years: is this really an activity Intel management could actually agree to be part of its long-term strategy? (yes, I know money talks, but...)

Many years ago, IBM made the "buy" decision for leading-edge geometries. However, they maintained their physics and device development teams to develop special-purpose devices. Recent announcements suggest that these teams are still fully competitive with the best in the field.

Global Foundries have somewhat floundered over the years. Part of this is clearly financial, but I believe that the way that they were formed may also have cut off the process development engineers from the device engineers - and I'm much too distanced from the reality to assess how deeply entrenched the consequent practises have become.

Both TSMC and Samsung both have highly competent teams. TSMC's business strategy has generated more investment money for semiconductors than Samsung's. Although in principle this need not be a problem going forward, TQC's comment strongly suggests that I am far from the only person who has been bitten by their historic inconstancy on foundry.
However, the bigger problem is that both companies operate in a geopolitically difficult region. If we are looking for insurance against the absorption of Taiwan, I suspect that working with Korea would only provide a modest breathing space if-and-when Taiwan is once absorbed.
I also suspect that both companies will feel slightly more secure with "the West" dependent on their home operations (I know I would in their positions).

So, in my view we need to "keep it local", and look at the structures that are most likely to lead to success. Clearly we need to learn from the Taiwanese, just as in the past they learnt from us.
Then. if we look at the semiconductor companies that were spun out of the old Eastern Germany, it appears that the ones that managed to "go it alone" are those that were formed around the advanced development facilities, rather than those set up for the volume production. Similarly, it seems Intel was strong when the first fabs at each node came from a dominant development base. At the moment the strongest development base appears to be at IBM, and the "natural" spin-off company for fabs developed by extended teams looks to be GF. All that is missing to make this happen is the strongly committed management and the finance.
It will probably take 14 years of government financial commitment and autonomous commited management commitment to catch up. The process could become self-funding after a few years, but the financier needs to understand the timescale - and fund consistently (!). However, there is at least an existence proof
 
The West needs a strategic second source at all levels of thetechnology. IT does not necessarily have to be the best - just good enough. The problem is how to get there, and whom to chose as the lead partner

Whatever Intel says, its success has historically been due to "seizing the moment" as much as technical strategy*. Indeed, they have been quite explicit about how this worked in the early days. If IBM corporate had been serious about PCs and willing to put in the spadework neither Intel's processors nor Microsoft's operating system would have become dominant as they have.
I know nothing of Intel's current generation of semiconductor engineers, but it seems that Intel appear to have lost the technical edge that they once had; I'm not equipped to judge whether this is a matter of financial focus or the competition's technical superiority. One thing is clear: Intel has blown lukewarm and cold on foundry services over the years: is this really an activity Intel management could actually agree to be part of its long-term strategy? (yes, I know money talks, but...)

Many years ago, IBM made the "buy" decision for leading-edge geometries. However, they maintained their physics and device development teams to develop special-purpose devices. Recent announcements suggest that these teams are still fully competitive with the best in the field.

Global Foundries have somewhat floundered over the years. Part of this is clearly financial, but I believe that the way that they were formed may also have cut off the process development engineers from the device engineers - and I'm much too distanced from the reality to assess how deeply entrenched the consequent practises have become.

Both TSMC and Samsung both have highly competent teams. TSMC's business strategy has generated more investment money for semiconductors than Samsung's. Although in principle this need not be a problem going forward, TQC's comment strongly suggests that I am far from the only person who has been bitten by their historic inconstancy on foundry.
However, the bigger problem is that both companies operate in a geopolitically difficult region. If we are looking for insurance against the absorption of Taiwan, I suspect that working with Korea would only provide a modest breathing space if-and-when Taiwan is once absorbed.
I also suspect that both companies will feel slightly more secure with "the West" dependent on their home operations (I know I would in their positions).

So, in my view we need to "keep it local", and look at the structures that are most likely to lead to success. Clearly we need to learn from the Taiwanese, just as in the past they learnt from us.
Then. if we look at the semiconductor companies that were spun out of the old Eastern Germany, it appears that the ones that managed to "go it alone" are those that were formed around the advanced development facilities, rather than those set up for the volume production. Similarly, it seems Intel was strong when the first fabs at each node came from a dominant development base. At the moment the strongest development base appears to be at IBM, and the "natural" spin-off company for fabs developed by extended teams looks to be GF. All that is missing to make this happen is the strongly committed management and the finance.
It will probably take 14 years of government financial commitment and autonomous commited management commitment to catch up. The process could become self-funding after a few years, but the financier needs to understand the timescale - and fund consistently (!). However, there is at least an existence proof
This is well stated and captures much of the multi-variable matters that are involved here. Technology is only 1 of them - very key to be sure - but still just one. Economics will always be the primary driver, for investing and incentivising the industry; but again, many factors at play. Your last one about "keep it local" seems to be missed in so many comments about whose tech is better - while key & headline grabbing, security of supply (& IP) from foreign intrusion and takeover is very real and much more important that the capacity of each node made in a fab. Very few products & even less defense systems rely only on leading edge tech (always), because by the time of marketing or deployment that edge has moved. This has been the case since the mid-80s and likely going forward will be similar. This factor will push decisions and locations of fabs much more than the auto industry's lazy mis-management of supply chain chips. Lots of media play but very little concern to IDMs and device makers. They're too busy to pay much attention to that noise.
The IBM-GF connection and comment is intriguing - maybe Intel "helps" with their planned IPO and makes that a division of Intel. Would resolve quite a few problem areas on both sides (cash & tech access for GF; capacity & foundry capability for Intel). Not to mention a US-based expansion on several levels.
 
TSMC seems to be "crushing" Samsung at this point. Scotten mentioned that Samsung was badly lagging TSMC yesterday. Today's article on Nikkei Asia just provides further evidence to reinforce his view:
Nikkei News said Samsungsemiconductor department profit drops YoY.
With DRAM doing well last year that means Samsung foundry was not doing well. 8nm and 5nm yield must be bad.
 
Nikkei News said Samsungsemiconductor department profit drops YoY.
With DRAM doing well last year that means Samsung foundry was not doing well. 8nm and 5nm yield must be bad.

Thanks for the excellent link. This points directly to the fact that TSM is able to spread costs over a larger and larger base that increases its marketing power at an ever-increasing rate. TSM has created a very unique position economically and structurally in one of the most key industrial endeavors in the world. Rarely if ever has a single company come to worldwide dominance in a key industry.
 
  • Like
Reactions: VCT
Thanks for the excellent link. This points directly to the fact that TSM is able to spread costs over a larger and larger base that increases its marketing power at an ever-increasing rate. TSM has created a very unique position economically and structurally in one of the most key industrial endeavors in the world. Rarely if ever has a single company come to worldwide dominance in a key industry.
It's not TSMC marketing power. Customers have more confident that TSMC can deliver 3nm on time with a very large qty. If you bet on the wrong foundry you could turn out losing market shares. That's what happened to Qualcomm right now.

Japanese security indicates that Samsung foundry lose (I don't think it's lose. Could be YoY profit drop instead of lose.) Came from CPU and communication IC. Communication must be Qualcomm.
It could be nVidia GPU instead of CPU.

However, it looks like nVidia 8nm GPU supply is getting better now.
 
Back
Top