Is TSMC just following directions?

[soAsian]

I'm skeptical about Peter Zeihan's claim that TSMC's role is limited to assembly, with Americans or Japanese handling management. If TSMC merely follows designs and instructions, why do Intel and Samsung continue to struggle to match TSMC's performance?

 

[samwilde]

Peter Zeihan is just a pretender. He knows nothing about semiconductors.

 

[VCT]

Peter Zeihan believes that Intel acquired high NA EUV technology before TSMC, which he thinks will make Intel surpass TSMC.

 

[samwilde]

Peter Zeihan believes that Intel acquired high NA EUV technology before TSMC, which he thinks will make Intel surpass TSMC.

That is not in the linked video. Anyway unless there is some kind of catastrophe at Taiwan that is not so likely.

As any new system it is likely that High NA EUV will have some kinks initially and Intel will have to figure those out at the same time that they are hemorrhaging cash. I still think Intel is doing the right decision but it is going to be difficult to get there. TSMC is just way more conservative, they will milk the regular EUV machines they have some more, before going into High NA EUV.

Intel has tried being in the foundry business before, the thing is, they provide such shoddy support to 3rd party customers that people just leave. Until Intel Foundry learns to be customer focused that won't change.

 

[Paul2]

That is not in the linked video. Anyway unless there is some kind of catastrophe at Taiwan that is not so likely.

As any new system it is likely that High NA EUV will have some kinks initially and Intel will have to figure those out at the same time that they are hemorrhaging cash. I still think Intel is doing the right decision but it is going to be difficult to get there. TSMC is just way more conservative, they will milk the regular EUV machines they have some more, before going into High NA EUV.

Intel has tried being in the foundry business before, the thing is, they provide such shoddy support to 3rd party customers that people just leave. Until Intel Foundry learns to be customer focused that won't change.

TSMC success is not entirely hinging on being "leading", they only started to strongly lead in the node race at around the start of finfet age. There were far more riskier, and aggressive contenders before who wanted to "skip the line", while TSMC only pursued mass market nodes.

Becoming the number two again would not be the end of the world for TSMC at all, it would not even impact its bottom line that much.

 

[Daniel Nenni]

TSMC success is not entirely hinging on being "leading", they only started to strongly lead in the node race at around the start of finfet age. There were far more riskier, and aggressive contenders before who wanted to "skip the line", while TSMC only pursued mass market nodes.

Becoming the number two again would not be the end of the world for TSMC at all, it would not even impact its bottom line that much.

Prior to FinFETs TSMC was still the leader but other foundries copied TSMC IP with harmful intent. Customers would first go to TSMC then 2nd and third source at UMC, SMIC, and Chartered Semiconductor with relative ease. TSMC did all of the heavy lifting but did not get the full reward. That is not possible with FinFETs thus the TSMC strong lead today.

 

[Paul2]

I easily remember when, besides Intel, IBM, a few Japanese foundries were ahead in the process technology, and GloFo was still in the front runners bunch. Early 200X.

I am surprised nobody has figured this business model yet:

Buy an outdated speciality process fab with tons, and tons of captive clients still making 200X era chips, then rise prices tenfold.

People were surprised how many nineties era chips were still in use by very hard to redesign stuff: modern cars, modern appliances, power distribution equipment, PLCs, aerospace, defence, civilian RF equipment, test equipment.

Now imagine how much more 200X chips are there around.

 

[Artificer60]

Intel has tried being in the foundry business before, the thing is, they provide such shoddy support to 3rd party customers that people just leave. Until Intel Foundry learns to be customer focused that won't change.

Intel's previous foundry "efforts" (using that word very loosely) required their customer's to navigate their non-standard design process and basically do things the Intel way. Very much the historic this is what we did and it is what you get mentality. Given the fact Intel is adopting industry standards metrics and design methodologies in addition to changing how they engage with customers I would say that are committed to becoming customer focused. They may still have a long way to go, but they are at least headed in the right direction this time.

 

[soAsian]

Peter Zeihan is just a pretender. He knows nothing about semiconductors.

the more videos he put out beyond his comfort zone (demographic), the more he looked like he didn't know anything. His only talking point that I can agree with now is population and demographics

 

[BruceA]

Intel's previous foundry "efforts" (using that word very loosely) required their customer's to navigate their non-standard design process and basically do things the Intel way. Very much the historic this is what we did and it is what you get mentality. Given the fact Intel is adopting industry standards metrics and design methodologies in addition to changing how they engage with customers I would say that are committed to becoming customer focused. They may still have a long way to go, but they are at least headed in the right direction this time.

Customer focused, they still have a long way to go. If TSMC represents the home run goal, Intel is rounding first base at best

 

[cheesehead]

I am surprised nobody has figured this business model yet:

Buy an outdated speciality process fab with tons, and tons of captive clients still making 200X era chips, then rise prices tenfold.

Probably because it is a rather short lived success. Yes, you can make a whole lot of money short term. But raising prices in such a way is a commercial end-of-life. As you also stated logical consequence will be re-design.

 

[horace]

Probably because it is a rather short lived success. Yes, you can make a whole lot of money short term. But raising prices in such a way is a commercial end-of-life. As you also stated logical consequence will be re-design.

Broadcom has perfected the art of raising the price. Milk every last cents from the customers right up to the threshold before redesign.

 

[chipsntexas]

Prior to FinFETs TSMC was still the leader but other foundries copied TSMC IP with harmful intent. Customers would first go to TSMC then 2nd and third source at UMC, SMIC, and Chartered Semiconductor with relative ease. TSMC did all of the heavy lifting but did not get the full reward. That is not possible with FinFETs thus the TSMC strong lead today.

TSMC may have been the foundry leader prior to FinFETs, but they were not a clear cut technology leader until Intel's massive stumble & stagnation at 10nm. Up to that time they were back and forth with Samsung for the #2 slot behind Intel's lead. I don't think that this is in much dispute, but I could be wrong on the timing.
But to think that Intel will take over #1 position in foundry is laughable and not within Intel's grasp - or strategy. They can and may "catch up" in technology, but that will be ever so slight and very much product/market specific. If they accomplish that within this decade it will have been a remarkable recovery and turnaround.
But that is a very large "IF" - and will cost many billions of dollars in R&D and capacity, both of which require way more customers than their current client list.