Friday, August 20, 2021

About Taiwan's microchip production...

 

Yesterday I mentioned that, in the light of Chinese belligerence towards Taiwan:


If I were a US manufacturer reliant on computer chips from Taiwan, which makes about half the world's supply, I'd be begging, borrowing and stealing any and all supplies I could get out of there before the invasion.


Reader Andrew D. sent me the following comment via e-mail. He's allowed me to share it with you.


I thought you might be interested in an assessment of the likely future acquisition of Taiwan by China from a different perspective.

I am an electrical engineer, and work in the particular niche area of designing integrated circuits.  Every part I've worked on in the last 25 years has wound up being fabricated at TSMC.  They aren't 'just a chip company'.  They are the 800-lb gorilla in the IC industry.  They make a huge fraction of the world's chips, and have gotten to that position because they are very, very good at what they do.  We haven't used the TSMC factory because we like them, but because they are so good at what they do that they are the compelling option.  They consistently deliver product on time and to specification.

When people talk about Taiwan, I occasionally hear them mention TSMC as a minor concern.  From my perspective in the electronics field, China acquiring Taiwan is a minor concern, but China acquiring TSMC would be world-shattering.

Chip foundries are incredibly expensive, and it takes a bunch of highly skilled people to get one working.  Not only are they expensive, but the equipment is very fickle.  Fickle as in 'the three secret factors that affect this ion implanter are the barometric pressure, the phase of the moon, and the mood of your mother-in-law.'  Without both motivated and experienced workers and equipment in good condition, there is no working fab.

If China invades Taiwan, then you can be fairly certain that some fraction of the TSMC foundries will be shut down, either by accidents on the part of the Chinese military, or action on the part of Taiwanese partisans.  Drop a bomb on a fab and you might as well just build a new one from scratch.

Pretty much everything electronic that you own that has one or more chips in it fabricated at TSMC.  Even if most of the ICs come out of Samsung or Hynix, there will almost always be at least one part from TSMC.  If you have 99 of the 100 chips needed to make an iPad, then you can't make an iPad.  If China takes over Taiwan, then they will basically control the entire world's supply of electronics.

Particularly for the newer fine-line processes, building a fab line involves many billions of dollars and 3-5 years of construction.  Getting one running well requires hundreds of workers highly skilled in very specialized areas.  Even if we had the committment and available experience, it would be years before we could have any alternatives to TSMC.

In short, my assessment is as follows.  If China invades Taiwan, then:

1) In the immediate future the supply of basically everything dependent on electronics (TV's, microwaves, automobiles, air conditioners, computers, etc.) will dry up.

2) If we are lucky and not much of TSMC is damaged, then we might start getting new electronic-dependent products in 6-12 months

3) If we are not lucky and the TSMC fabs are destroyed, then we might start getting new elecronic-dependent products in 5-8 years

4) If China has control of a functional TSMC, then:

a) they will be able to control who gets what when it comes to electronic-dependent products (which is basically everything)

b) every company that manufactures parts through TSMC will have to decide whether they want to provide the Chinese government with their intellectual properties

5) Alternative fabs will get constructed in other companies, and in 5-10 years we will start having other options than TSMC.


Sobering stuff, isn't it?  Thanks, Andrew, for sharing that with us.

Peter


21 comments:

  1. There are downsides to allowing comparative advantage to determine where things are made, and the possibility of an adversary gaining control over critical supplies is a biggie.
    Once upon a time, we had top-of-the-line fab capability in Silicon Valley, Texas, and Arizona (and probably other domestic locations that slip my mind at the moment). But, well, construction and operating costs are higher in the U.S. than in countries that value industry, so the last few generations of expansions and upgrades happened overseas. That, and it became so much more efficient for chip companies to outsource their manufacturing to the one big service bureau that had all the equipment and skills.
    We're seeing another round of side effects from the efficiency-at-all-costs school of management. Just-In-Time leads to supply crunches when factory operations are disrupted, and outsourcing leads to an industry-wide single point of failure.

    ReplyDelete
  2. Actually the US still has plenty of fabs and fab expertise. In fact TSMC is building fabs in Arizona - https://arstechnica.com/gadgets/2021/05/tsmc-is-considering-a-3-nm-foundry-in-arizona/ - and Intel has a number of fabs there

    If the PRC delay their invasion until those fabs can be up and running there's a chance global disruption will be limited. But even a foundry like TSMC can't simply transfer production from one plant to another. Some production sure, but no two fabs are the same and there will be chips that one can make and the other can't

    Of course given TSMC's ubiquity in the chip business the PRC itself would end up being hurt by TSMC failing to be in good shape. All those assembly lines in Shenzhen would end up idled and everything else that China makes that contains TSMC chips. I doubt the PRC leadership have thought of that though.

    ReplyDelete
  3. But hey, didn't all those companies 'save money' by shutting down chip production in the US? Nothing more important than 'saving money' at the cost of national security and possibly wrecking infrastructure for want of some electronics that monitor refinery output, every electric motor in the world like for you know, maybe a water pumping station, and maybe the fleet of trucks to support all of that.

    ReplyDelete
  4. Of the various kinds of chips, where TSMC is the 800 lb gorilla is in the smallest geometries, which are the processors. These are kind of the glamor sector of the industry while the bulk of what's produced are larger part geometries and much of what's produced are fairly old designs.

    In Analog and Power semiconductors, the US is still either leading or in the running. The thing that inhibited US semiconductor fabs was not just short term business focus and "saving money" - rather the incentives to go offshore to save money were created by government regulations both federal and local.

    This is not to criticize Andrew's comments. While I'm a retired electrical engineer, I never worked in IC design. In your home entertainment world, you need all of the above: processor and audio chips; power supply and video processor. Without the processors, the things don't do anything.


    ReplyDelete
  5. China is playing the long game. And they're winning.

    ReplyDelete
  6. BRM, I pondered this somewhere yesterday as well. In point of fact they could quietly starve out the West in terms of technology, while working to build up Allies of the Silk Road.

    Could the West rebuild? Perhaps as Francis Turner suggests, yes. But would it be in time?

    ReplyDelete
  7. One big upside to having the equipment and expertise concentrated in a fab company is that it enables the existence of "fabless chip companies" - there are many of these, and going the fabless route can cut startup costs for a chip venture by three orders of magnitude or more, thus bringing many innovative products to market (if only briefly in most cases).
    But, well: it creates a single point of failure for a multitude of enterprises, both the fabless chip companies and those who depend on their products.

    ReplyDelete
  8. I'm just left wondering if Taiwan has 20+ agencies promulgating various "safety" rules (that have the force of law) for the factories that produce the chips. It's not just the cost of skilled labor in the U.S. that keeps the manufacturing costs here so high.

    ReplyDelete
  9. Years ago (Reagan was president), friends worked for a company called GCA making chip production equipment. One of the issues was mechanical accuracy and another was optical accuracy. Zeiss was the leader then in chip making optics. I believe the Japanese makers like Nikon and Canon put GCA out of business. Are these factors, optical and mechanical accuracy still the reason chip making is dauntingly expensive to get into? Do they still need the best lenses, if so, wouldn't that source be a factor as well?

    ReplyDelete
    Replies
    1. The fine lines connecting the circuits are now narrower tha the wavelength of visible light. They have moved up (smaller) the wavelength spectrum past x-rays.
      It is very hard to make a flashlight that shines only 3-4 nanometers wide.

      Delete
  10. libertyman:

    In that era, chip making machines used visible wavelength through the mask to expose the chip layout on a wafer. Production line width geometry was ~1.25 micron. That optical system would only be viable down to ~0.7 micron width, and then the actual wavelength would be too long to fit into the spaces. Shorter wavelength systems would have to be developed to go smaller in sizes for more compact chips. I'm not sure optics are even used for this purpose anymore. They were talking about x-ray wavelengths might be utilized.

    Chipmaking crashed worldwide in '85, and lots of fabs were shuttered, and some were never restarted here in the US/Silicon Valley. Sad to hear that Intel basically tossed dustbags over our machines (Ultratech Stepper) and locked the doors of most of their fabs. Quite a sight to see 50 steppers in one room, and it be a 4? story building. Company went from building/selling ~30 Steppers a month worldwide, to none in a months time, and sold virtually nothing for the next year or so. Those steppers sold for $750k for 1.25micron, with the optional 1.0 micron lens adding $100k (mostly used for R+D).

    ReplyDelete
  11. Regarding the optics: for Ultratech, it took near a year to get good lens elements into stock to build the lens system of 5 elements (it was unique, it was a one to one replication system). The other players (Nikon, etc) used image reduction systems 10 to 1, or 20 to 1, so quality wasn't as critical for them.

    ReplyDelete
  12. Why is everybody assuming that the PRC can roll over Taiwan the way they did over Hong Kong? The Taiwan Strait is about 90 miles wide, and an amphibious invasion is no joke to put on.

    The ROC on Taiwan knows that the PRC would love to take them over, and they've been preparing obsessively since 1949. They know where the invasion will come. And the last time the PLA did anything big-scale, in 1979 in Vietnam, they got their heads handed to them.

    ReplyDelete
  13. Technomad: Happens I just overheard a discussion of the prospect that the PRC might take over Taiwan.
    Given a couple of decades of ever-strengthening economic ties, the takeover might not take the form of military action. The Taiwanese government might be persuaded to go along with reunification, for some value of "persuaded".

    ReplyDelete
  14. Of course, even if TSMC survived an invasion, it would then be exclusively under PRC control. Even if you could get the chips from them, would you really want them? Could you be certain there were no, er, 'Imperial entanglements' included free with every order?

    Convergence of crises, indeed. And the beat goes on...

    Mike in Canada

    ReplyDelete
  15. Given a couple of decades of ever-strengthening economic ties, the takeover might not take the form of military action. The Taiwanese government might be persuaded to go along with reunification, for some value of "persuaded".

    If that were to occur I would expect significant resistance from Taiwanese who do not want to be re-unified. Best I can guess that's close to 50% of the population, possibly more. Economic sabotage would be one very effective way for the opposition / resistance to act. And a ton of it could be "accidental" and/or "incompetently following orders" and it would be very hard to figure out what was a real accident/mistake and what was sabotage. Actual sabotage of key PRC assets (on the mainland and on the island) would be another and it would be very very hard for the PRC to protect against because there's no easy way to tell the difference between a Taiwanese and a mainlander, let alone between a loyal islander vs a disloyal one. The UK had that problem with the IRA to a degree, Taiwan would be a lot worse.

    ReplyDelete
  16. Yeah I've been trying to tell people about this for a long while now, going back to discussions with enthusiastic Trump supporting friends who thought his plan of imposing steep tariffs on Chinese imports was somehow going to bring manufacturing back to the US.

    Contrary to a couple other comments here, although we do have some fabs in the US, they are not sufficient to replace east asian fabs. For one thing we lack the expert manpower to scale up to meet demands of the entire electronics market, or even its most critical parts.

    We need orders of magnitude more technicians and engineers to accomplish that, and they can't be produced with "learn to code" style "bootcamps" that your wordpress "developer" friend attended. They will need *years* of training and on the job experience to even be basically competent.

    For another thing not all fabs are created equal, and we cannot produce high-end components in our fabs (such as high end LCD & LED screens, GPUs, CPUs, RAM, and nvram used in solid state storage - SSD hard drives, firmware, USB devices, etc).

    Forget about iPads. There are industrial microcontrollers running critical infrastructure and what little remains of our domestic manufacturing that cannot be produced, or not in sufficient quantities, in the US. Let alone vehicles, all of which depend on foreign manufactured ICs (and other parts, some of which could be done here more easily). Not just your sedan - also commercial and industrial vehicles used in agriculture, logistics, etc. If these can't be built, and/or replacement parts can't be produced in sufficient quantities to maintain existing fleets, you have bigger problems than whether your kid can watch Youtube on his smartphone.

    The solution to all of this is not war with China to protect Taiwan. This will probably end up with all of us dead, but even if it doesn't, such a war would cause supply chain disruptions similar to the worst-case PRC outcome.

    The solution is to hang (out to dry) all the grifters in our finance and manufacturing industries and get moving on scaling up our manufacturing capability ASAP. This is a decade-long project at best, *if* we don't waste resources and people on a cold/hot war with China. There is going to be a lot of pain and suffering in the meanwhile either way, and it's better if the long-term outcome is more self-sufficiency instead of a mountain of corpses and another multi-trillion dollar invoice for a temporary and uncertain can-kicking session.

    ReplyDelete
  17. "expect significant resistance from Taiwanese who do not want to be re-unified"
    Very likely, yes.
    But... just how out-of-touch is the PRC leadership?
    How are things going in Hong Kong, and what are those at the top being told about how things are going?
    The transnational ruling class has developed a serious case of hubris, and seems entirely clueless about popular sentiment, and about how far the masses can be pushed before they start resisting. (See, e.g., France and Australia just lately.)

    ReplyDelete
  18. One vlogger I would recommend to BRM readers is for a dose of optimism is laowhy86 (on Youtube, but also Odysee). He was the first (to my knowledge) last year to bring credible evidence, based on Chinese-language sources, of the lab-leak origin for the virus. He is a lot less pessimistic than many commentators on the free world's chances, and a lot less impressed with the CCP's long-term prospects for dominating the world.

    ReplyDelete
  19. Has anyone mentioned the pre-installed spyware that will be on the chips once China controls TMSC?

    ReplyDelete

ALL COMMENTS ARE MODERATED. THEY WILL APPEAR AFTER OWNER APPROVAL, WHICH MAY BE DELAYED.