Where does TSMC's *primary* "value" as a semiconductor supplier lie:
- equipment
- technology
- personnel
- "volume"
I.e., in the event China went after Taiwan, would you bemoan the loss of the equipment, know-how, staff or "capacity"?
Said another way, how *replaceable* is TSMC (given some practical timeframe)? I.e., do all of its capabilities already lie in the West -- but at a reduced scale/volume? Would airlifting personnel do the trick? etc.
The other aspect of that question is: how likely would China be able to exploit TSMC's abilities if:
- the equipment was sabotaged (or relied on parts from The West)
It's the continue development of personnel/people. China brought some of the key people of TSMC before, but without the right environment, they stop developing.
On a sunny day (Fri, 08 Apr 2022 16:07:30 +0200) it happened Jeroen Belleman snipped-for-privacy@nospam.please wrote in <t2pfj2$m1r$ snipped-for-privacy@gioia.aioe.org>:
This is an odd question as it would only apply to those who are actually customers of TSMC. To the rest of us, what is important is obtaining chips from the companies who *are* customers of TSMC. We don't have the insight as to what the issues of production might be or how to mitigate them. We could only speculate with little real insight. In other words, a perfect s.e.d question.
I expect TSMC's most important aspect is the institutional knowledge of making chips at the newest process nodes. I don't think China would be able to take over the plants and make anything important without full cooperation of the key employees. I don't think sabotage is required. The process is delicate/fragile enough to be broken easily by simply not knowing what you are doing.
I believe that an invasion of Taiwan by China would have *much* more severe impact on the world than anything going on in the Ukraine, neon gas or not. Taiwan's semiconductor industry is much harder to replace.
That is wrong question. AFAICS TSMC "value" is a a whole: all ingredients are there and it works _now_.
Of course TSMC is replaceable, this is just matter of time. IIUC current estimate is that China needs about 10 years to catch up to TMSC and this time is mostly due to western embargo. Building advanced fab seem to be about 3 years (but possibly another 2 before initial troubles are worked out and fab operates smootly at full capacity).
IIUC at most advanced process Samsung has similar capability. Much of production is on older processes, and those are available in several fabs outside Taiwan.
IIUC China has know-how and people capable of operating such a fab. ATM China is not capable of making some of needed equipment. If crucial equipment was destroyed, then fab probably would be useless to China. If equipment was merely faulty, there is some chance that China would be able to repair it. There are consumables (notably photoresist and pure chemicals) needed to operate such a plant. I am not sure if China _now_ can make all of them.
Without smooth transition almost surely there would be some dip in production.
OTOH if China is smart and can learn from history they will not attack Taiwan in next few years. ATM China is probably first industrial power in the world, but they are depenedent on trade. And currently China is too weak to protect its trade from hostile US actions (like naval blockade). So if they are smart they will wait till they have stronger position.
So, by that reasoning, ANYTHING China might do would irreparably damage TSMC.
Any staff who were demotivated (or executed for not performing to the level required by The People), any equipment damaged or rendered ineffective due to parts/supplies, etc.
I can't believe that TSMC has a monopoly on fab.
Rather, I suspect that they represent *capacity* at *lower cost*. And, that any loss of capacity there would be compensated for, over time, elsewhere.
The question becomes one of what limits the rate at which that capacity can be restored: the lack of "spare" equipment? the lack of staff to operate same? cost of that labor/materials?
E.g., the US has now started to reinvest in semiconductor fab. No doubt, the cost per item will be higher than at TSMC simply due to differences in labor/operating rates. But, there's no "secret recipe" that is at risk of being lost if TSMC (effectively) closed its doors.
That suggests there is value in the *equipment*. But, ignores the expertise/experience provided by the existing staff.
OTOH, if China were to undertake such an "expansion adventure" and be met by the same sort of sanctions that Putin is suffering, there may not be much *need* for that large of a fab (if it is only supplying domestic demand)
They rely on an inflow of currency to pay for the upgrades they've not undertaken in their "ancient" society. They likewise also rely on western tech (and education) to further those goals. What cost to their plans if, for example, all chinese students were expelled from western schools (in the event china becomes a pariah)?
That's not to say that Draconian measures couldn't "entice" them to be more productive.
NKorea seems to have no problem getting advanced munitions technology. Possibly some of that is "carrot-based" (better living conditions for those individuals who "produce") and I imagine a fair bit of "stick-based" (no "rule of law" to fall back on if you feel yourself "unduly pressured").
They don't. But they are better at it than anybody else at the moment.
They represent a capacity that would cost a lot - in money and time - to equal and surpass. Representing the state of the art is always a transient honour, but it isn't easy to get into that position.
But not for quite a while.
Don Y imagines that being a state of the art setup is a "capacity that can be restored". For all practical purposes it is a capacity that has to be re-invented. It can be done but it take time and money.
Re-inventing a state of the art production capacity requires experts who are skilled in the art. There aren't many of them, and it takes a while to recruit candidate experts and work out which of them will end up being able to do better than the current generation of experts.
But has it started to invest in the kind of managers who won't cut corners to try to get to the desired performance faster than the experts who work for them imagine to be possible?
But there's a lot of in-house expertise that is really difficult to transplant.
Probably not. You'd have to airlift a great many of them, and the information resources they rely on for their day-to-day work. Semiconductor fabrication is a very complicate process with about 120 sequential operations on each wafer - probably more now - and getting any one of them wrong wrecks the wafer.
Developing the expertise/experience of the operating staff is where the two years years needed to get the fab operating smoothly at full capacity gets used up.
But lack the experience of doing it.
ASML is just the tip of an iceberg of supporting technologies.
To nothing, for years.
And if they are even smarter, they will recognise that they will be even stronger as collaborators in a cooperative world than they can ever be in world that was motivated to collaborate against them.
The first round of sanctions against Russia haven't done enough damage to Russia to make Putin's position obviously untenable. What his colleagues actually think won't be obvious until some of them have work out a scheme to minimise the damage.
Even if there is a process at another fab that has the same approximate capabilities as a particular process at TSMC, the exact characteristics of the devices will differ so that for anything with analog parts (adc, pll, high speed serdes etc.), the circuit design will need to be altered, by hand, and re-simulated. The design rules will be at least subtly different, so that even if an analog circuit still works in simulation, its layout will generally need to be redone. This would take many months per part (and the people who designed those products may not work there any more, figuring out how the old product worked is not trivial), and a several semiconductor companies would have many hundreds of parts made by TSMC. Some customers (automotive, medical etc.) would also not accept the new design as being equivalent to the old one, so there would be another delay to qualify the new parts at the customers.
But all that is true of *any* foundry. It's not like they have the only 5nm process or are the only fab that can handle mixed-mode or...
Their strength appears to lie in capacity (though less than Samsung in terms of wafers produced). And, possibly *price* (?).
I.e., if they suddenly closed up shop (or WERE closed up), the industry would recover, after a delay (as it will re: covid).
The announcement of new fabs in the US (including TSMC) suggests that economics was the previous driving force; entice manufacturers with subsidies and they "suddenly" see the merit of building domestic installations. The bottleneck of qualified personnel and kit just inserts a delay in that schedule (but it takes time to start ANY business operation).
[I believe Intel is adding two plants to its AZ location -- along with TSMC increasing its investment, here. The question (for the industry as a whole) then boils down to how "general purpose" those fabs may end up being. E.g., Micron is likely not going to be producing bleeding edge *processors* -- for ANY of its clients!]
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.