I’ve done a handful of videos on semiconductors and there’s a very good reason for that. The production of semiconductors and the companies involved will be under the spotlight for the next few years as the entire industry gets shaken up.
TSMC has set up chip production in Arizona, despite initially resisting relocating to the US. This facility isn’t doing the cutting-edge stuff, but it’s still producing chips on the higher end of the spectrum. TSMC has also managed to achieve a high recovery rate on these chips in Phoenix, not quite a major breakthrough but at least it reduces production costs.
Most of the chip manufacturing is automated, so the higher labor costs in the US and skill gaps relative to Taiwan aren’t playing as big of a role as expected. However, to expand the reaches of these facilities and begin development of cutting-edge chips, some major investments will need to be made. Let’s look at what Intel is doing on this front next.
Here at Zeihan on Geopolitics, our chosen charity partner is MedShare. They provide emergency medical services to communities in need, with a very heavy emphasis on locations facing acute crises. Medshare operates right in the thick of it, so we can be sure that every cent of our donation is not simply going directly to where help is needed most, but our donations serve as a force multiplier for a system already in existence.
For those who would like to donate directly to MedShare or to learn more about their efforts, you can click this link.
Hey, everybody. Peter Zeihan here, coming to you from the Boston Logan Airport. It is not quite five in the morning. Anyway, it’s still a decent backdrop. So we’re going to take an entry from the Ask Peter forum—specifically, could I give an update on the status of TSMC’s efforts to establish chipmaking here in the United States?
I’m happy to report that it’s actually going a little better than I thought it was going to. The very short version is that Donald Trump almost forced TSMC to relocate some of its production capacity to the United States and made it very clear that he wanted the very, very, very top end to be made here.
TSMC said, “Sure, of course, whatever,” and then proceeded to drag its feet in every possible way. Remember that the leading edge of chips these days is less than three nanometers, getting into two nanometers, and probably within the next couple of years, getting to 1.5 and maybe even one nanometer. The facilities that are under construction in Arizona have been dragged out, dragged out, dragged out, and dragged out, with, in many cases, TSMC not even providing proper architectural blueprints so far.
So there’s been construction, then they tear things down, and then they rebuild something and tear it down. They’re basically just buying time. But the first facility actually is operational. It’s just not the cutting edge—it’s like four nanometers, which is still pretty good, but it’s not the kind of stuff you’re going to probably put into an AI server farm or anything anyway.
Part of the big news that came out in late October was the idea that they’re getting a higher recovery rate from the new facilities in Phoenix than they’re getting anywhere else. While this is an important development, you shouldn’t get too excited.
The process for making the chips: you take a little bitty seed crystal, you put it into a pool of liquid silicon, and then you steadily pull it up over the course of several days to grow a crystal. That crystal ends up weighing more than a Volkswagen. It tends to be over a foot or two across and about nine feet long. I mean, it’s a little different at every facility. You get this giant ingot, and then you slice it laterally into thin discs.
You then use a combination of lithography, baking, and doping to etch those chips. You bake them to make sure that everything sticks, and then you do it again and again and again—something like 90 times. It takes a few months to make each individual sheet.
The waste is one of two things.
Number one, you have a section of the semiconductor sheet that just doesn’t work. So that would be waste. Or maybe it’s just the shape because, usually, your chips are squares or rectangles, and the disk is round. So you can have waste at the edges.
TSMC is famous for having the highest recovery rates in the industry. With its four-nanometer nodes, it’s something like 90% coherent and only 10% waste. The TSMC facility is now 94% coherent. So it is an important technological jump. It does drop the overall cost of the items you can produce, and since U.S. labor is more expensive than Taiwanese labor, you know, that’s great. But don’t get too excited about it.
Something else to keep in mind about these facilities is the labor that is necessary.
Very highly skilled? Yes. Is there a lot of labor? Not really. Most of this is automated because you’re using a lithography facility that is being produced by ASML, the Dutch company. You know, it’s automated. The whole point of extreme ultraviolet is it doesn’t require a lot of manual adjustments.
The old technology, deep ultraviolet? No, that did. When you are doing DUV, you’re constantly making changes to every individual machine for every individual run. You get much higher wastage because the chips aren’t all exactly the same. With EUV, it’s all automated. You have to do it once, and you can apply it across the entire system for every lithography machine in your facility. The chips come out much more regular. It’s kind of like an analog versus digital sort of thing.
One of the constraints we have faced with moving this stuff from Taiwan to the United States is that the labor costs more and isn’t quite trained right. But with EUV, that doesn’t matter as much as it would have with the older technologies.
Anyway, it’s moving ahead. Facilities two through five? God knows when those are going to be operational because those are supposed to be the higher-end ones. But this low-end, high-end chip of four nanometers seems to be moving along just fine. Just keep in mind that the real breakthroughs are going to be coming from TSMC this year.
If the United States is really going to get in the game of high-end semiconductors, it’s going to be using a new lithography technology called High Numerical Aperture, which is like the next generation of extreme ultraviolet.
TSMC isn’t bothering to work with that. That’s an Intel project. The Dutch company ASML has provided the technology to both companies, and only Intel has bit. That is the technology that is going to be used at the Columbus facility, which hopes to begin operations in 2026.
We’ll see.
