@HPC Podcast Archives - OrionX.net - @HPCpodcast-105: Dan Nystedt on Chips, China, AI Tech

Episode Date: November 14, 2025

Dan Nystedt, vice president of research at Tri-Orient Investments, joins us to discuss the Asian tech scene: geopolitical tensions and Taiwan's role, rivalries in chip manufacturing and AI, complexities of the global tech supply chain, and China's rare earth dominance and geopolitical impact. [audio mp3="https://orionx.net/wp-content/uploads/2025/11/105@HPCpodcast_ID_Dan-Nystedt_Chips-AI-China_20251113.mp3"][/audio] The post @HPCpodcast-105: Dan Nystedt on Chips, China, AI Tech appeared first on OrionX.net.

Those are the decisions, those are multi-billion dollar decisions you have to make as a chip designer when you're figuring out which foundry to use. You can look at numbers of data centers that have been built in the area, but from my perspective, it looked like a very concrete strategy of China's to use Singapore in particular. You know, you build the shell and then you can let that sit for a while because you want the clean room, you want it to be ready, and then moving in the equipment later because it can't be said enough that in rare earths, rare earths are not that rare. From Orion X in association with InsideHPC, this is the AtHPC podcast. Join Shaheen Kahn and Doug Black as they discuss supercomputing technologies and the applications, markets, and policies that shape them. Thank you for being with us. Hi, everyone. I'm Doug Black at InsideHPC. With me is Shaheen Khan of OrionX.net, and this is

the at-HPC podcast. We have a special guest with us. Today, Dan Naistet, he is vice president of research at Tri-Ori-Orient Investments. Dan is on the ground in Taiwan. Tri-Ori-Ori-Rient is a private institutional investor that has been active in Asian markets since 2010. Dan, is that a fair introduction? That's, yeah, that's perfect. And thanks for saying my name right. Okay, good. So if you would, fill us in a little bit about your background, how you arrived at, I don't know if you're in Taipei or near Taipei or just on the island, but just a quick bio on how you ended up there. Sure. I am in Taipei, and I have spent most of my time in Taipei, although I studied Chinese in the south for a while in a city

called Tainan, which I love. It's a wonderful town, and there's a university there called Chengong University, which is funny because Chengong in Chinese means success. And I was having a really hard time studying Chinese. And so I thought, look, if I can't make it at Success University, I'm not going to make it. So I went down there. And then it actually turned out, because I remember I said that to somebody, to one of my teachers down there. And she said, it's not named after success. It's named after there was a Chinese general, Koshinga, if that name rings a bell. From the Ming dynasty, the Chinese were being taken over by the Manchus, and the Manchew's army were sweeping down. So this one general, Koshinga, retreated to Taiwan. And that was a

was the first time there was a very big Chinese presence on Taiwan. And so that university is named after him. And I just didn't even know that. It's just his Chinese name was anyway. So to make a long story short, I arrived in Taiwan in the late 90s. I came here to study Chinese. And then when I did start to get a grasp of the language, I didn't want to leave because it had taken me so much longer than I thought it would to kind of start to get it. So I became a journalist at first for first for one of the local papers in Taiwan, and then for Dow Jones Wall Street Journal, and then I went to IDG publications, which I really enjoyed working for them on the pure tech side because Taiwan's a great place for that. And then, yeah, around 2011, a friend approached me

and said that they knew of an investment company that was looking for somebody to work on the tech side and that they didn't mind that I didn't have prior experience and the financial side. So I joined them and things have been pretty well actually that was very lucky a very lucky way to get into the financial side of of this industry so in a nutshell that's it that's awesome okay how is your chinese by the way are you fluent do you dream in chinese i'm pretty good i'm pretty good yeah i've been here for a while and i mean i'm very conversational and and and i attend meetings in chinese and uh yeah Amazing. Amazing. Okay, so I think a lot of our discussion tonight will be the geopolitical aspect of technology and chips. From a high-level perspective, give us your sense of the U.S.-China geopolitical rivalry and Taiwan's position within it, you know, especially from an advanced chip perspective. Well, and I was afraid you were to start off with a hard question. That's a big one. That's the biggest one.

Okay. Let's just go right to it. Go right into it. Yeah. So, yeah, that's a big one to chew on. So I'm sorry if I'm going back a little bit too far here, but I think when communism was falling apart in the, you know, in the late 80s and the 90s when the Berlin wall fell and Gorbachev, you know, gave up the kind of for the Russian communist side.

I think we all hoped that the new future would be that we could build, you know, an entire world without communism and that everybody would join, like China and Russia would join the World Trade Organization and the World Trade Systems that were already in place that had done very well for many countries in the world. And I think what instead happened was that China joined with the idea that, you know, we're just doing this until we get on our feet and then we're going to steal as much as we can and we're going to do as much as we can to build our military. And we want to be number one, period. But I think for a long time, we all hope that they would join the world in a very peaceful way. And it's very sad to see that that's not happened.

So what we find ourselves at today is a situation where for a long time, I think most people in America and even the government believed that China was our friend, so to speak, which from where I've been sitting the whole time I've been here, I've never thought that. And I don't think people in Taiwan view, whatever view it is that way, because of all the missiles aimed at Taiwan from China and the constant threats of invasion. So, but you find yourself in this position. and you had this trust for a long time, and that trust meant that you didn't think about things like that China was kind of doing a lot in the rare earth side, and that China had, actually China bought a lot of U.S. military technology, a lot of military technology was sent to China

as literally outsourcing, and some things were rerouted to China from other places, which is just the reality of what happened. One example that I know for sure about that is the, you know, the night vision goggles that were made a long time ago. That was a very nice U.S. military advantage for a long time. There was supposed to be a factory built in Singapore. And then that was rerouted and that was in fact built in China. And that is exactly how China got its night vision capabilities. And that's just one example of many examples of how that's happened over the years.

So today the U.S. finds itself in a situation where China has made a very aggressive stance in the South China Sea, so much so that even Vietnam, who was also literally a communist country and supposedly a friend of China, doesn't like China right now because China's basically said, we own your ocean. Like your coastline is now ours. Get your fishermen out of here. You know, they made this nine dash line down there, which, you know, a lot of people don't even know what that understand what that means. It just literally means instead of taking over land, they took over ocean. And they've sent a huge fleet down to defend that land or that ocean. sorry. And in the Philippines that you see there are daily videos. If you look on even Twitter or

YouTube, Filipinos are putting up videos almost every day about things happening and fishermen being harassed and blocked from traditional fishing areas. And the Chinese military, they're radical. And what the Navy is doing down there will probably start a war faster than what's happening in Taiwan. But you find yourself in that situation, the U.S. suddenly had to say, we can't keep doing this. We're basically, we've given China everything and the Communist Party is in absolute control of China and is absolutely hell-bent on, you know, they're in border conflicts with almost every country that borders them. And they have demands for more land on almost every country that borders them. You know, it's all partly the idea that we now have the power and now we should

exert that power. But, you know, you get to the point where you're looking at global politics now And on the chip side, the chip industry has always been very big and very free to travel wherever it wants. The controls for some items had remained in place even after the Cold War ended simply over the fear of, do we want people to be able to make these same weapons that we can make with these electronics? That's been a saving grace for us right now. But, you know, Taiwan is a place, the TSM built by Morris Jong, who was a former executive at Texas Instruments. and also was a CEO of a couple of companies in the U.S. before he came over to Taiwan, and the Taiwanese government at the time had been really trying to promote semiconductors,

and they asked him to see what he could do. And he had had this idea before about doing the foundry, and then obviously he did it. It's been a massive success to the point where now the most advanced chips in the world, the most manufacturing process in the world, is in Taiwan at TSMC, with probably the best competitors being Samsung Electronics. And of course, Intel's has done a great work catching up too. But so geopolitically, if China wants to take over Taiwan, you know, the U.S. is in a position to say, well, really, you can't let that happen if you're the United States. And so that theory from a long time ago called the Silicon Shield, that Taiwan has this resource that's just as important as oil is to the U.S., which is the silicon. And that's what's keeping Taiwan out of China's hands right now. The only thing I would say to that is you hear all the time, the only thing that U.S. cares about is semiconductors. And, you know, I would say the U.S. has been

defending Taiwan from far longer than there were ever semiconductors here. So it was, you know, it was part of the Cold War when China was taken over by the communists. The U.S. did defend Taiwan and some of the islands that are, there are islands that belong to Taiwan that are within sight of China. I mean, they're so close to China. If you're in Shaman, China, you can get up and look out the window on a clear day and you can see Kinman or Jinman. You can actually actually see it and you can take a ferry ride there that's like a five mile ferry ride or something like that just to go and hang out is that where they have the loudspeakers that that's right that's right there they're actually a couple there there are more than one island that that they

aimed those loudspeakers from on both directions yeah but that's your you're absolutely amazing and and then for all the hostility you're describing and aggression and we do hear people saying say Hoover Institute analysts saying, you know, Taiwan and Philippines, Vietnam is just the start. They actually have designs on Japan and then going further east. But I've also heard you say that it's either you or most people in Taiwan are not overly worried about a Chinese invasion or Chinese action against Taiwan. So is that just denial or fatalism or is it rational? No, I think it's denial because I'm sure you've heard of Sunza

He wrote a book called The Art of War It's the most famous Chinese military treaty It's You know, he's one of the things he wrote That I always fall back on is Basically the sentence translates something like You can't expect the enemy not to come You must make plans that the enemy is going to come

And he hits on that a couple times and now being in a Chinese society, because, you know, it's without question. This Mandarin is the language people speak here. Even Taiwanese, the local dialect, is a Chinese dialect. There's no question that this is a Chinese society. And they are in denial. And they just, oh, they're not going to attack. And you hear these things all the time, even behind closed doors and in small groups.

There are a lot of people who think China's not going to attack. But I guess I would love to have heard what were the English saying before Hitler launched. You know, did they really think he was going to attack them or not? Because I think to a certain degree, everybody lives in denial because you don't want to think about things like that. Yeah. That's a bit of a problem. But I think enough people now, they've looked at it and they're trying to figure out, how can we make ourselves at least very difficult to take over? You know, it would cost you a lot to take us over because it's still the case that taking over a country, you know, an island is much more difficult than land.

If you're Russia invading Ukraine, that's an easier proposition than going across miles of ocean to take over a place in terms of logistics in terms of everything. So they're wielding power in the region and wherever the reach is. How do you assess just exactly how much power there is? Because there's also a bunch of global head games that take place where countries try to come across as having something that they don't. And I think it makes it really difficult to know whether you're underestimating someone or you're overestimating someone. That's a great question. Well, you know, in Shunza Art of War, he would say, hide your strengths. And that's actually one of the most interesting things a teacher told me before I came here was you should reach Shunza art of war because that's how everybody treats each other.

He was a Chinese professor. he was American, but he was Chinese American. And he said, when my Chinese students come over, I tell them to read the Bible, and then they'll understand American culture much better because so many, so many things come out of the Bible and so many ethics come out of the Bible. And when I send, like you, over to China, I say you should read the art of war because you'll understand how people are a lot better. And it's fascinating because number one, DeShunzu, was information.

Probably number two is, I don't even know if these are different, but it's information in the sense. You want to know as much as you can about your, enemy, and you want your enemy to, A, know as little as possible about you, but also send signals about what might or might not be true. And that's one way of controlling the information about you. And you actually do see that in society. People are very quiet. People, it's not, it's not like in the U.S. if you, if you're in New York City in the morning going to work or even probably before the work time, you're just walking down the street about to buy a cup

of coffee. You might pass several, you know, several people and they'll all say good morning to you. And that's just a very normal thing in the states. Because I lived in New York City, so I know that. But I'm from a small town in Ohio where everybody says hi to everybody all the time. And it's kind of being, in American culture, being friendly is almost something you must do. Right. Yeah, exactly. But out here, it's not like that.

And the Chinese culture is much more, it's very family-oriented. But there are a lot of friendly people. Don't get me wrong when I say that. I just trying to explain a point. It's a family-oriented society in that sense. that it's like, this is your group, and that's who protects you, and that's who you can trust. So, yeah, that's kind of how it plays out in society. Do we know?

What do we do and what do we not know? You know, the Chinese are very good at propaganda. They're number one in the world of propaganda without question. But, but yeah, that's the case of what's happened. So you're spot on about knowing what's what. And so that's a danger is to try to figure out between what's real and what's not. I'm not sure where to go. I mean, part of me wants to ask more about geopolitical and a possibility of war, but what if we talked about a comparison?

We've heard you say that Chinese AI is overrated, yet on the foundry side, we wonder, you know, will Intel be able to catch up or compete with TSM? They've been doing some impressive things with their FAB 52, the two nanometer chips that have come out. But, of course, most of the action now, the real excitement is around GPUs. But then again, the U.S. seems ahead of China on AI, in part because we're denying China access to NVIDIA chips. So looking at, you know, on the foundry side versus the AI side, how would you compare the capabilities of the two sides? Okay, you just went through a lot of things right there. Do you want me to just go into the comparison of the foundries? Let's start there, actually, yeah.

Okay. Well, so I would not compare Intel's Foundry to any other Foundry right now because it's not really even in the, it's not even really a contender. Maybe it's a contender. I don't know, but I think what people have become way too optimistic about Intel. And I'm not saying there's anything wrong with the company. I'm just saying it's really hard to get into the Chip Foundry business. And that's what seems to be what people underestimate.

Because look, Global Foundries was spun off of, AMD, what, 2009? And it took them 13 years. Yeah, 2022, they finally became profitable. They got a full year profit. So it took them 13 years to reach profitability. And they had Mubadala that, you know, the Abu Dhabi sovereign wealth fund behind them. And that's the only reason they survived for 13 years. So this is not an easy industry to get into. It is not easy to do any of this stuff. And the biggest, the biggest challenge for Intel will always be, will not always be it at some point hopefully they'll get long-term customers but before you have any real long-term customers if i already use tsmc say i'm amazon or i'm google or i'm nvita or whoever

if i want to use intel's foundry i have to redesign my chip to match intel's process so the manufacturing process itself it has its own design rules and so you do actually have to re you know rejigger your design and you have to redo your design to go into that. And that's always going to be a challenge because the question would always be, why do I want to incur that cost? And of course, right now a good reason to incur that cost is because if China invaded Taiwan, you need chip manufacturing someplace. And TSM, of course, they're building in the U.S.,

but they don't have their most advanced process in the U.S. So Intel is a good alternative in that sense because you have a secure, you know, you have a secure supply. So I hope a lot of companies are thinking like that. I'm not sure they are. But and that again, that goes back to the foundry is very difficult business. You know, if you think about it, if you think about foundry from the perspective of the customer, the customer is Qualcomm or Apple or whoever's making a chip, that's the customer.

So the first concern is your intellectual property. So there was a rumor that AMD was going to use Intel as a foundry. And as soon as I saw that, I just thought, yeah, that's never going to happen. Because is AMD really going to send its best chips to Intel's foundry? They've been competitors for so long. I just don't see that happening. And they would be afraid that the foundry would give some secrets to the design team, the Intel design team.

And I would have that fear. I don't, you know, if somebody came in and said, there's no reason to fear that, we know, we're good. And I believe Intel is a great company and very honest company. I, you know, I think all those things are true. But if I'm AMD, I'm still never going to do that. So your second concern would be if I, if I use Intel Foundry, because I need a second source, you know, what, so first I, it's a big commitment of money. You know, my design, when you're talking about designs going for, you know, three nanometer, two nanometer, you're in the hundreds of millions of dollars for how much your design costs.

I think the last estimate I saw was $700 million for, my design cost is $700 million for a two nanometer chip. And of course, that's going to vary different. It's going to be different between different kinds of chips or different companies. That's a lot of money. So I'm thinking about, I've got my $700 million design. Who do I put it with? Do I put it with Samsung? Do I put it with TSM, who I've been using for the last 20 years?

You see what I'm saying. It's a big amount of money. But then on the other side, you've got your product. The product that comes out, if there's any problem with that product, especially if I'm like Qualcomm or Apple, say Apple, and I think Apple did run into this. at one point. The chips that come back to me, I put them in the iPhone, I sent them out, and suddenly a bunch of users are saying, my phone doesn't work, or my phone's overheating, or my battery is draining so fast. And that's actually, they did have some overheating problems

and some battery issues. And I think they were able to fix it with like a software fix. But if you remember that, that was a while back. And I hate saying the names of companies, because some of this was speculation. I'm not sure if it was ever confirmed. But it was said that that was a Samsung chip that that was happening on. But, but, you know, but that's your other concern. So the chips coming out of the foundry. If they are poorly made, my business this year is sunk. So, so those are the concerns you have if you're going into that. It's, it's, again, it's the intellectual property. It's the cost that you put into the developing your chip and designing your chip. And then third would be the end product that comes out. And whether it, you know, whether

it succeeds in the market or tanks in the market. So those are the decisions. Those are multi-billion dollar decisions you have to make as a chip designer when you're figuring out which foundry to use. So yeah, and I'm not saying that to hurt Intel's chances or anything like that. I'm just saying these are the realities that it seems like a lot of people are not even looking at. So I actually do think that the federal government, if they can push a few companies to try to work with Intel, I actually think that's a good idea because it's a difficult decision to make. And I hope Intel is able to succeed on those cases because if that happens, if they do get some government help pushing wise and then they fail on the product side that's

it's it would be very hard to rebuild any trusts if something like that happened so they have to deliver they have to deliver so yeah exactly but you know in the in the most recent uh intel direct connect where they talked about their foundry i i thought they the way they talked about it i really liked the way they talked about it because they they really understood roadmap because that you know another consideration is you i'm as a chip designer I'm not planning for, you know, my two nanometer chip. I'm planning for my 2 nanometer, my 1.6, my 1.4 nanometer, my 1.4 nanometer, my 1 nanometer or 10 angstrom, however you want to call it. You know, I'm planning for, it's generation after generation after generation because I'm trying to put out a new chip every year.

So, so that's roadmap. And they talked a lot about roadmap. They talked about making sure they had that they were listening to their clients and that they were putting, you know, they're putting out their, their PDK. to make sure that that was the way people wanted it done and that had the information that people wanted and that it was easily easy to use for the clients and et cetera, et cetera. Those things are all important and those things are good to hear. So, yeah, fingers crossed.

It's really hard to come from behind. It is, it is. It's hard to start a new business like this and it's hard to come from behind. Well, they were ahead and they lost the lead and now they're trying to recapture it. And it seems like once you've, fall behind, it becomes just exponentially, well, maybe that's not quite the right word,

but it becomes, there's compounded factors that get in the way. When you look at the supply chain from chips to systems, to rack scale, to AI, to data centers, and that across the entire spectrum, are there any parts of this that are getting overheated in your view? Everything's overheated. Prices are rising everywhere. No, it's pretty incredible right now because, you know, this the supply chain, basically you had this this massive surge in data center building because of COVID, you know,

when the, when the, but those were just cloud data centers, but that still provided some extra cash for companies. But then as soon as that was over, it was like in 2023 that the, the, like the PCs, the smartphones, the consumer side of things just collapsed. And so you had a situation. where the data centers and the servers were still kind of maintaining, they were the only game in town for such a long time. And so the companies that make all the supplies and the even chemical substrates, CCL, you know, printed circuit boards, motherboards, et cetera, all those

guys, they were okay for a while because they had no demand from one side of the spectrum, where they had the brand new demand from a new side of the spectrum. So that probably helped in 2023 and part of 2024, that probably helped keep the supply situation okay, because at that time, the bottleneck was really co-os, advanced packaging. And maybe a bit, there were tight supplies of other things. I think substrates were in tight supply because of co-os and because of, you needed a lot more of ABF substrate for a co-os package than you did for previous packaging, right? So you had, you kind of had a situation where it was kind of okay because, you know, only part of the

part of the tech industry was doing okay. And then here we are in 2025 where smartphones, they're not really picking up. They're a little bit better than they were last year. But they're definitely getting more silicon with, you know, with some of the new chips that are going in for AI. I mean, directly for AI, like NPUs or NPU cores that are going in. And that's, so they're beefing up a little bit, which does, that's more materials that are needed. But if, I guess if the PC industry were suddenly to just take off next. year because everybody decided they want an AI PC, you're not going to have memory. You know, you're not going to have DRAM for that.

You're not going to have a lot of different materials are going to be in. Disc drives. Yeah, yeah. It's one of those things where the things have taken off so much that at this point with, and this is still driven by the AI data centers and kind of the build on that has just taken off in a way now that it didn't. Previously, it was it kind of was getting started and there was an understanding that you needed a little bit of time for the supply chain to organize.

to build a bit more capacity, et cetera, et cetera. But yeah, you had some of these component sides that were starving for money because they didn't have their consumer demand for a while. And then they were just being kept afloat by the AI server demand. And they had always been very low price. And then very suddenly this demand shoots up. And they're kind of like, well, yeah, it's a bottleneck, but it's only because I need to build a new plant or two new plants or, you know,

and it's hard to forecast. And they don't have much trust on forecasting because, every time there's a downturn in the tech industry, it's they're left holding the bag. I built this plant for, you know, for these guys who said it was going to be great and now it's not great. So that's kind of what happens in the supply chain. And that's really right now, you know, it's, it's not that there's any, there are probably some materials that are simply short of supply because even the raw material itself is

short of supply. But it's mainly that they're, they don't have the plants prepared and they're building new plants for certain things that we really need to get up and running to provide the materials for the next, you know, to go up on up the supply chain. Dan, and then over on the AI side, we did see your comment that your sense is that Chinese AI is overrated. Could you elaborate on that? Sure. And don't take the comment wrong. I think there's, there's tremendous AI talent in China. And a lot of it is trained in the U.S. I mean, they, you know, they went to U.S. schools and

they worked for Google. They worked for different AI companies and whatnot. And then China's universities on their own are now becoming quite good at AI and software development in that area. So it's not a knock on that at all. It's more of kind of what Shaheen mentioned earlier of how do you assess the information that's coming out. And even, you know, Deepseek, when they put out their report saying that they had only used a very small number of GPUs for this, blah, blah, blah. And that turned out not to be true at all. You know, semi-analysis put out a great report later saying, yeah, we don't believe that number.

And here's what we know they had available to them. But we, you know, obviously, they can't make an assessment of exactly what was used, but they can say that they know that the company had a lot more available to them. And so, and even now, you know, with what's going on in AI, you know, I'm not sure people understand, okay, it's not been hard for China to get as many enviable. GPUs of the top quality that they can for a long time. And I thought the Biden administration did a good job trying to put out a policy where you broadly allow certain things and then you start to tighten up on the policy

and you tighten up on the supply and export controls. And there was a bit of a whack-a-mole problem with China's able to, you know, Chinese companies will just open a new company. If you, so Huawei is under a sanction, they'll open up another company and they do things like that all the time. But you had the administration working on that. And I thought that was not a bad way to go about things. I'm not saying the Trump administration is doing worse. They seem to be a little bit more of we're not going to allow anything on this side and we'll allow a little bit on this side. So it's more of a blanket approach. So there are a couple

difficulties with that. One is you have, there are plenty of data centers that have been built in Singapore, Malaysia, and other places near China where the Chinese use those. And there are loopholes and export controls. You know, if I'm a Chinese company, can I just use Amazon AWS, just rent GPUs from them? I don't have to buy GPUs. Can I just rent them from some other place? And so some of these loopholes have been filled and some of these loopholes have not been filled. And so specifically, when I say that their AI is overrated, I'm talking about what they say they're able to do and the way they say they're able to do it on no chips or low power chips or whatever, I don't believe that side of it because they have plenty of access. Yeah, they have plenty of access to

NVIDIA chips. Plenty. And, and, you know, even if you look at the numbers of, you can look at numbers of data centers that have been built in the area. But from my perspective, it looked like a very concrete strategy of China's to use Singapore in particular because you had Singapore and then, oh, there's a city. It's always called J.B. So I always forget the. name Jehor. I don't remember the name of the whole name of the city because it's always called J.B. And I got used to saying that. And anyway, but it's literally 45 minutes from downtown Singapore. And then for Indonesia, there's an island off the coast of Singapore. It's also, it's like an hour ferry ride from Singapore. But there you have three different nations. So if you're,

if you're trying to fight US export controls, if the US decided, oh, we're going to control Malaysia, we're going to sanction Malaysia for for these data centers that are giving, you know, some access to the Chinese. Well, then the Chinese companies that have already opened an office in Singapore, they'll just use a Singaporean data center. Oh, and if that goes wrong, then they'll use an Indonesian one. So there's obviously been a strategy for them to be able to use different nationalities. And of course, now there are data centers, you know, being built all over the place and that Chinese companies can have access to. And so that's mainly what I'm talking about with whether their AI is, is robust or not. It's not about whether they're the actual software

and what they're producing. The side of what they're producing, the part that's not robust is it is communist Chinese AI 100%. And they can't make anything else. Because if they make anything that questions the CCP or the history that the Chinese Communist Party spews out, the Nell will get shut down. So they must tow the line, unquestionably must tow the line. So for any company that is using the free, you know, the open source systems coming from China, you know, that you've got to filter that out. You've got to find a way to debug it or you've got essentially communist thought going through your AI all the time. There's some bias baked in. Yes. Wow. Yes. Now, all these data centers, are they going to get access to power and allocation of

chips and SSDs and disks and everything that we said was in either short supply or, you know, even if you have the money, you need to get the allocation and or are they all? just like lining up behind each other to eventually do it if demand is still there? Well, you know, the best strategy is probably to sign long-term contracts right now because the memory, especially in memory, that problem looks deep and long. Because you guys know how long it takes to build a fab. You know, at least Samsung S.K. Heinex and Micron had at least started on shell fabs, which is the strategy you do, you know, you build the shell and then you can let that sit for a while

because you want the clean room, you want it to be ready and then moving in the equipment later because the later you move in equipment, you get a lot of advantages from that. One, equipment is 80 to 90% of the cost of the overall fab, which is just the way it's always been. Two, if you get, if you buy equipment two years from now, it's two years more advanced than the equipment you buy today. And if demand shifts, you have an empty fab shell that you can fill. And there's not a whole lot of that. But they've been building some of it and some of it will come online later next year. But again, you're looking at, it's a problem where even just the AI data center demand has gotten so big that the guys that are making, you know, the DRAM companies are overwhelmed.

If the PC industry takes off as well, you know, well, we're just in a, we're just in a lot of trouble. Right. So, yeah, that's an issue. You know, another aspect of this big supply chain is the name of the companies, because sometimes I read your tweets and you're speaking of companies that I feel that maybe not everybody in the West has even heard of them. Now, there are those like Foxconn that are obviously global brands and everybody knows and they're huge.

And then you go kind of down the list and maybe Quanta, people have heard, maybe Gigabyte and MyTac people have heard and then you go to Wistron and We Win and Kytus and then these are not household names here but they're huge and you sort of like go Google and say how big is such and such and oh my God this is like you know billions and billions and billions and of course like in the West like the names you do here

are HPE Dell Lenovo super micro is even a relative recent entrant And what is it that causes that massive players to show up there? And where does the business come from? And then, you know, you look at chip design and you say, oh, you know, Microsoft signed up with this, you know, I never heard of their name, but it's like a $5 billion company that does this thing for chip design. Can you speak to just the breadth of the supply chain? Yeah, sure. You know, a lot of the companies you just mentioned, especially the ones that are,

that are AI server makers would be identified like Foxconn and YWIN and Quanta, Wistron, all those companies. Those are massive contract manufacturers. And they got their start a long time ago in the computer industry. And their customers have always been companies like Dell, HP, Apple, and whatnot. And so, you know, Quanto, its claim to fame years ago was it was the number one laptop maker in the world. And of course, they're making laptops for, They made laptops for Apple or assembled laptops for Apple and other companies. And so that's where all those companies come from is just years of being in the part of the computing industry.

And they're experts at what they do. And they're able to build factories quickly. They're able to ramp up quickly. They know the business very well. And that makes them the best partners out there for a lot of this stuff. And then Japan's always interesting because after Japan's, semiconductor industry kind of imploded you know they that was interesting because they basically they were number one in memory chips and then Samsung comes along and they had actually

to get around the US trade war they had started working with some of their companies that started working with Korean companies and one of those companies was Samsung so Japan helped build Samsung and and so because you could export from South Korea instead of exporting from Japan and then you get around the trade you know the the barrier the trade barrier. And that's what they were doing. So and then Samsung one day says, well, wait a minute, you know, the Japanese companies are all making essentially mainframe DRAM. And the U.S. wants PC DRAM and it doesn't need to be, it doesn't need to be that good. You know, they need something far less powerful, far less intricate, which is kind of what

happened with mobile DRAM as well. So in Japan, you have companies. There's a food company called Aji Nomoto that does the Aijitomoto film is what's used. an ABF for the ABF substrates that you're talking, you know, that we're talking about with so right for packaging, chip packaging. And Aginamoto did a, did work with Intel. Intel worked with them because they had found out that this Japanese company had this, this technology. I think it's a protein technology and they made this film and simply a more heat resistant, you know, it could handle a bit more heat, which sometimes your CPUs run hot. And of course, GPUs run very hot as well. And that's how, but yeah, that's, and then EBiden is the Japanese company that makes substrates.

That's probably not as familiar as in terminology and does a lot of packaging materials, actually. So, yeah, so Japan is just full of a number of these companies. And you kind of, they were all built during the time when Japan was number one in memory chips. And number one in chips overall for a brief moment. So is it practical to try to. fragment the global supply chain into kind of spheres of political alliance, is it possible to build a high-end 2-nometer, you know, 14-onkstrom chip entirely in one country? It's, you know, it's not been done that way in the past. I mean, for certain parts of the

industry, I would have to say no. I mean, in the very simple reason. It's like really hard. So in the U.S., it really hard because even like in the U.S. there's great technology, but they still don't have ASML's EUV lithography machines, right? And that's ASML, of course, is a, as a Netherlands company. So, and then even if they succeeded, then where do you get your masks from? Because the best photo masks are made in Japan. And of course, you know, there are other companies that make it. There are other countries that make it.

But when you're talking about two nanometer, you need the best chemist, You need the very best photo masks. You need the very best of everything. And that's, which thankfully, you know, in Taiwan, what's been great for Taiwan has been the growth of TSM and TSM doing so well because I think every company, you know, when TSM has a problem, like they're trying to build a co-os production line, they go to Taiwan equipment makers and say, can you guys make us this? Like, you know, because the PCB industry has been so big here for so long. So that was, there's a bit of a natural fit in terms of if you make PCB. equipment, that equipment can be, you can modify some of that. You can do things differently and and just build something for TSM and that. So that's done very well for a number of Taiwan

equipment makers, but also the chemical companies here have, have purified certain chip chemicals so much for the advanced processes that they're the best, they're the best in the world or they're at least competitive with the best in the world. And so, yeah, I would say in the chip industry, not only is it difficult to make a silo in each country, if you really are going to do that you have to get everybody on board, you have to get other industries involved, like your chemical industries and other industries, and just let them know, you need to make the best, most pure, you know, of this chemical in order for us to move forward with this. So it's not an easy thing to do. It's not an easy thing to do, especially when

you're going against companies that have been doing it for, you know, decades. Right. And they're executing. They're not falling behind. And executing. Yeah. And executing. Yeah. This is like the second law of competitive intelligence. That was one of my jobs in the old days was whoever takes it more seriously wins. Yeah, absolutely. I mean, it's only one of the laws, but it rang through and it still does. I've heard you speaking previously, you seem comfortably disposed or favorably disposed toward some of the so-called circular financing deals that we've seen.

I think you've said it's a creative response to the need for massive capex for building out AI infrastructure of the future. Is that kind of an accurate take on your feelings or thinking? Yeah, let me be very brief with that one. And then we can do another one if you want. I just think our house, like our investment firm's view is that we're still very early in the AI build out and that we've seen bottlenecks, not just supply chain bottlenecks, but we've seen bottlenecks. Now we see bottlenecks in energy. You know, there's not enough power for AI data centers. And there's also a bottleneck in financing and figuring out how do we pay for this or how do we get this done in a short amount of time is essentially the question it's not how do we get it done

it's in a shorter amount of time and then it's something to keep an eye on but it's certainly not something that we're worried about right now and and it's as I just as I said before you have to find creative ways to get around these bottlenecks or or else it's going to stop progress and I think the last thing companies want right now is to stop progress but but yeah at some point it may become ridiculous, but to right now, I don't think it's ridiculous. And you're talking about companies that have a lot of money and that have good cash flow. And at the end of the day, people forget about business sometimes that a lot of good businesses can go under if they don't have cash flow because they simply won't be able to make payments to their employees and to their suppliers and to

everybody else. On the other side of things, a lot of bad companies can can last a long time because they have cash flow. They're able to make those minimum payments. They're able to make payments to their employees, they're able to make payments to their suppliers. And I only, and I'm not saying these are bad companies, but the memory industry has been doing that for decades where they have a boom, you know, they go through that nasty boom bust cycle where right now the like the big memory chip guys, Micron and Samsung on Hynix, they get this big influx of money all at once. Believe me, once they build enough fabs, all that money is going to disappear when the, when the bust comes, which always does. And then they survive. They will, they will keep making

chips at a loss for a while until the market rebounds because they need that cash flow because they have loans to repay and they have people to feed and they have suppliers to feed, et cetera, et cetera. So yeah, it's a creative response to what's going on in my view. It's filling a bottleneck or getting rid of a bottleneck at this point. Are there other technologies besides semiconductor and computers and AI, things like quantum computing or biotech or, you know, obviously rare airs, not quite a technology, but like an industry, cryptocurrencies, or even cryptography for that matter, that you're observing that Asia in general and Taiwan in particular are building strengths in. I think the most interest is in quantum

computing, but certainly there's a lot of interest in bio. There's a lot of interest in other areas, but there's certainly a lot of interest in quantum computing. And that's probably led by China, but it's also true of Japan and Taiwan and South Korea for sure. I think people see that as going hand in hand with computing in general, and especially Taiwan looks at itself as a, like its first name, the Republic of China is its official political name. And for a long time, it was called the Republic of Computers or the Republic of Computing. And so, yeah, they want to be there on quantum computing.

And yeah, so those are some very big issues, very big technologies out here. Just one comment real quick on the rare earths. It can't be said enough that in rare earths, rare earths are not that rare. They're found everywhere. They're in the U.S., they're in Canada. They're all over North America, Australia, they're Siberia, you know, they're found everywhere. China's only the best at processing those rare earths. And that's only because it's a very difficult thing to do and a very time-consuming and expensive thing to do.

And so a while ago, the West just kind of said, oh, if you want to do, it at a loss or you you can find a cheap way to do it you go ahead and do it but it doesn't mean that nobody else can do it just takes time it just takes some time and it takes some yeah exactly so why they call rare well they are more rare than other things and they're found in they're found in smaller quantities and like you might have a copper mine that's huge but then within that copper mine you find a very small amount of certain rare earths and that's why they're called more rare I think they're not in big deposits and they're not quite as easily found. And that's probably what makes it expensive to process because you're trying to get that trace amount.

And the U.S. has some minerals and things that China needs that, you know, I would hope the U.S. remembers that with some of these negotiations. The U.S. can literally shut down China's entire semiconductor industry if it said, you know, we won't sell you. What is it? In North Carolina, there's a, there's a, there's a, there's a, I don't. material there that can only be found there that's used in silicon wafer yeah i remember reading about that when there was that hurricane that hit north carolina that's that's exactly right and that was

a big concern for the industry kind of like is this going to shut these shut these guys down for you know a long time is at the worst possible time when you have an a i boom going on but um but yeah it's so it's the u.s has power too the u.s can can it's more difficult in some ways because it's not really something the U.S. does, you can't withhold that material from silicon wafer makers because that would hurt everybody, you know, that kind of thing. You're right. Okay. Interesting world we live in, yes. Very interesting world. Hey, guys, this is... Quartz, some kind of a quartz, was it? Yeah, that's, that's right, that's right. The name escapes me as well, but it would be better, it would sound more intelligent of me to say the day.

Well, thank you very much for your time, Dan. It was a pleasure to connect and look forward to doing more of this. Yeah, I'd love to be on again. It was very fun. Thanks for having me on guys. Perfect. Thanks so much. All righty. Bye-bye. That's it for this episode of the At-HPC podcast. Every episode is featured on insidehpc.com and posted on Orionx.net. Use the comment section or tweet us with any questions or to propose topics of discussion. If you like the show, rate and review it on Apple Podcasts or wherever you listen.

podcast is a production of OrionX in association with InsideHPC. Thank you for listening.