Fault Lines
I'm not a seismologist. In fact, I'm not a scientist of any kind, unless you count political science which, unfortunately, a lot of people in my academic peer group did and surely still do.
Given my complete lack of professional expertise in the natural sciences (which is to say real sciences or "hard" sciences, so to speak), I'm not qualified to posit a link between earthquakes and climate change. But I bet there are links, even if we don't fully appreciate or understand them all just yet.
On
The news on TSMC’s fab plant in AZ has been discouraging in the past few months, troubled as it has been by a shortage of workers and various delays. However, two days ago there was a news release (in Chinese – I haven’t seen a link) that apparently stated that, at least for the 4nm fab things were actually going ahead of schedule and that they hope to see some level of commercial production begin before year’s end (the higher end 3nm fab is foreseen to begin production in 2027). TSMC has a financial conference call scheduled for April 18 where it is expected that we will learn more.
Excellent. As an investor in Micron and Western Digital, I’ve always found it a little curious, not to mention more than a little disingenuous, that whenever potential frictions develop with China, these companies (and most of the rest as well) tend to tout their Taiwan strengths, sort of as if it were some non-China China sanctuary — all the benefits abd diversification with none of the direct risks. At a minimum, as this post underscores, we are drifting away from the ideal pole we’ve enjoyed for a long time now and now heading straight for “less of the benefits with more of the risks.”
And if AI were truly worth its salt, it would be screaming at us to get its heart and soul off that island.
Mr H I need to correct you, TSM are already building the most advanced chip manufacturing plants in Arizona. Production of the 4nm chips will start at the first plant in ’25.
There’s no need to “correct” me. Until that plant starts turning out cutting-edge technology at a rate that’s sufficient to supply the US with the chips it needs in the event Xi invades Taiwan, or blockades it, or a natural disaster cripples production, then this remains a “promises, promises” type of thing.
And maybe I should’ve been a little clearer. What I mean is that the White House should say: “Listen, if you want us to risk American lives in a military-to-military confrontation with Xi’s army, then you’re going to move heaven and earth to get your most advanced processes in place and running as quickly as possible on American soil, and there aren’t going to be any more excuses. If that means you have to fly part of your Taiwan staff over here and we have to find housing for them in Arizona (or wherever), then that’s what it means. But there’s not going to be any more foot-dragging.”
I mean, seriously. Think about this: We’re guaranteeing, essentially, that the US would engage in a conflict that’d probably fire the starting gun on World War III to protect that little place, and there’s still a very real sense in which TSMC is determined to keep its most advanced capabilities domestically-located. That’s not tenable. These aren’t potato chips. Or sedans. We need those damn semis. And given what it is we’re (implicitly) promising to do for Taiwan in the event Xi pulls a Vlad, we should have them.
I hope people and the system listens to this sensible discussion. It makes no sense to increase WWIII risk by acquiescing to TSMC’s desire to monopolize semi production in Taiwan. Just putting one plant on U.S. soil dedicated to one company does not resolve any crisis, it may just whet an appetite for conflict.
Maybe we should push harder on TSMC’s customers since that has had some results. We could also push cooperation with the other members of G-20 to buy local for all chips. Would tariffs be enough or do we continue the carrot approach also?
Excellent recommendation Mr. H. A lot of that CHIPs Act $$$ money should be going to subsidize relocating and paying salaries of experienced Taiwanese engineers to start making the most advanced and strategically vital semiconductor chips here now, rather than subsidize Intel to build some giant boondoggle plant in Ohio and wonder if they will even be successful in manufacturing the most advanced chips by 2030. Also, the production capacity here should be reserved for the most strategically vital accelerated computing chips initially, and not whatever chips Apple needs for their latest MacBook or iPhone.
I think the Biden team needs to 1-issue the 500 visas for Taiwan staff to pick up the pace at the plant in AZ, 2-not listen to the local unions opposition to this, and 3-finish the contract for US funding already appropriated. Taiwan just replaced the guy heading the AZ project because of the delay.
It’s not a given that they meet that goal at that date.
Those 4nm chips are not the most advanced node.
That said, tech insiders suggested that the 4nm was NOT their original plan but was forced upon TSMC during a meeting between Tim Cook from Apple, TSMC’s largest customer, and the then chairman of TSMC. In return, Apple promised to take ALL of the output from the first AZ fab. (Sorry AMD & Nvidia.)
Why did Cook have to pressure TSMC? It’s because their government puts relentless pressure on the firm to NOT place their cutting-edge technologies offshore. That is to bolster “the Silicon Shield” which they believe will compel the US to protect Taiwan from the PRC. After all, Americans will certainly support going to war so their children and grand kids will not be the only one in their middle school class without the latest iWatch!
INTC to the rescue? It only lost $7B on its foundry business in 2023. Rock and a hard place, truly.
Seems like someone’s CHIPS act was some pretty forward thinking, eh??
Not knowing much, but just thinking: If TSMC meaningfully moves chip production to the U.S., wouldn’t it be much harder to count on the U.S. to risk WW III to save Taiwan? Looking through the Ukrainian lens at least, counting on the U.S. to “defend democracy” in the absence of self interest (is there a lot of other U.S. interest in Taiwan?) is tenuous at best. And how willing is China to let them go? SMIC is setting up 5 nanometer production, so how tolerant would they be about being behind? In another vein, anecdotally, I have a friend who escaped INTC manages an A to D chip design engineering group. He finds his best engineers in Mexico.
That is why Taipei pressures TSMC to keep most of their production on the island.
You bring up a good point that the monopoly creating a greater risk of WWIII is caused by market manipulation. Monopoly concentration can be reduced by way of G-20 applying pressure.
I thought about this post while looking over INTC’s recast 2021-2023 financials with Foundry and Products separated. The stock dived on the release, amid an analyst chorus about how disappointing and alarming Foundry’s losses are.
Really, that was a surprise? INTC has revenues 2-3X its nearest competitor, broadly comparable prices, and yet its margins stink. Everyone knew there’s some big money-losing going on.
The recast financials make it clear where that is. Intel 10 and Intel 7 processes are so expensive and uncompetitive that if Foundry sold those wafers at market prices – what a competitive foundry would charge – its gross margins would be negative. And so they now are revealed to be. Intel 3 is less uncompetitive, Intel 18A will be competitive, Intel 14A will be better – so INTC claims – but it won’t be until 2027 before competitive nodes are the bulk of Foundry’s wafers and Foundry is – again, claimed – solidly profitable. That’s a while away.
So investors have, I think, valued Foundry at near-zero or put another way, nearly 0X book, and repriced INTC to just the value of Product – if a decently profitable fabless semi were worth less than 15X P/E which implies no long term growth.
So I see a lot of potential revaluation over the coming years. In the meantime, I can’t think of a better hedge against a Chinese invasion of Taiwan. Well, other than HII and GD.