Can 3D Printing Save US Manufacturing?

A 3D printer

We’re entering an era where restructuring global manufacturing will be non-negotiable. As supply chains collapse and tariffs complicate this process, can technology like 3D printing take some of the pressure off?

Investments in US manufacturing have declined under Trump’s protectionist policies, since relocating abroad can help avoid the tariff rabbit hole. 3D printing offers a promising solution for reshoring some of that manufacturing, but it’s too inefficient for large-scale production as of now.

As 3D printing improves and finds niches that align, this could be a disruptive technology. However, we won’t be replacing mass manufacturing with these printers anytime soon.

Prefer to read the transcript? Click here

Transcript

Hey, all. Peter Zeihan here. Walking down Indian Creek on my way out. Dreaming of Mexican food. But you know that’s not going to get satiated. Because there’s no good Mexican food anywhere near Denver. Anywhere. Today we’re taking a question from the Patreon crowd. And specifically, it’s, building off of some of the concerns that I’ve had with manufacturing. 

The short version is that the more complex the manufacturing system is, the more countries are involved. So when you put tariffs on the import of manufactured goods, either the finished product or the parts, what you’re basically saying is I don’t want to participate in the supply chain because it’s cheaper for everyone to move their production base out of your country. 

And then just import the finished product at the end of the day. Otherwise they have to pay the tariffs two, three, four, ten times. It’s one of the reasons why, the Trump tariffs are actually reducing investment in physical plant in the United States and reducing the amount of manufacturing products that we’re actually producing anyway. The follow on question from that is, is there a technology out there that might help us to get around that? 

And there, there might there might be, something called 3D printing. Basically, you take a powdered substrate, whether it’s a plastic or a metal, and then you sinter it with a laser, and grow a product. It’s often called additive manufacturing as well, instead of subtractive manufacturing. So subtractive manufacturing was more like punch holes and things. And you start with a block of material and you whittle it down until you have what you need. 

Additive manufacturing or 3D manufacturing? 3D printing is the opposite as you build it up layer by layer. Now, there are plenty of things that this looks very promising for. But the key thing to remember is if it has moving parts, especially moving parts that are different materials. It’s not that this technology cannot be used, it’s just that there are some pretty sharp limits materials, printers that can handle more than one type of material are pretty new, really just in the last 510 years. 

And the speed at which you can do things like this is very slow. So it’s very popular in things like, prototyping where every prototype is unique and then it doesn’t matter if it takes you hours to days to print the product. It’s also very popular in things where, abnormal shapes rule. So especially if you need a lot of strength but not a lot of weight. 

So you’re going to leave holes or bubbles within the material. So for aerospace, there are actually examples of 3D printers already on production floors and to a lesser degree in automotive as well. But the big thing to keep in mind here, speed, in the time that it takes you to stamp 100 products, you’re probably only going to make one 3D printed product, and so while 3D printing is getting incrementally better day by day and that’s great. 

And while it will undoubtedly, as the cost of manufactured products go up, as the globalization kicks in, it will obviously find more and more niches, where it’s the applicable technology, but it will always be coming from behind when it comes to mass application because of that speed issue. So I like the technology. I like the way it’s going. 

We should hurry up and get there.

The AI Race to Regression

The Open AI chatGPT logo on a phone

The AI race has been all the rage, but what if we were racing ourselves straight into regression?

OpenAI’s ChatGPT-5 is extremely powerful; however, it’s less user-friendly than its predecessor and is optimized for institutional users. Industrial and research applications are where the real power of AI lies. So, what happens when those energy-intensive data centers begin to falter?

Well, as globalization breaks down, that faltering is going to become a very real concern. Without an ecosystem that produces and shares all of the necessary components to make these AI behemoths run…we could see a technological regression that threatens the future of AI as we know it.

Prefer to read the transcript? Click here

Transcript

Hey everybody, Peter Zeihan here coming to you from McCurdy peak. Well, the actual peak is there. Anyway, Peter Zane Company from Colorado. Today we’re taking another question from the Patreon page. Specifically, can you please explain to me this new space age that we’re in the race for artificial intelligence, and what we should look for, what we should worry about? 

Well, let’s start by saying that most of the things that people are talking about with AI are generally, not quite on the mark, for example, a lot of folks think that, OpenAI, that’s the premier artificial intelligence company in United States, that their new program chat, GPT or 5.0, which is supposedly an upgrade, is actually a significant downgrade. 

They find it not as user friendly, not as personable, not as complete. That’s for personal users. AI affects potentially thousands of different applications, and how most people interact with artificial intelligence is in some sort of first person single seat. interface. Like what you get on your phone or your laptop. 

I mean, I’ve got that way too. And the jump from chat GPT four to GPT five was not designed for your single user. It was designed for people who do code for people who designed drugs. It’s designed to bring a huge amount of processing power to things on the back end to basically recreate something. So the institutional users, the design users, they’re actually finding ChatGPT all kinds of fun. 

And some Altmann, who is the CEO of open AI, is going back and kind of taking some characteristics from ChatGPT for to put it in the chat, GPT five, in order to make everybody happy. So that’s all going to work out. Here’s the problem. Software versus hardware. If I’m going to really sum it up, it’s that 

Chat GPT for the algorithm that we all found so groundbreaking really only took up about ten terabytes. And you could easily carry that on thumb drives in your hand. Chat GPT five, more advanced, is at least twice that, probably three times. But OpenAI is not saying. So we don’t know that number for sure. 

The point is, in terms of the raw memory required to make the AI function, it’s really not that impressive. And so if, the corporate espionage or an act of benevolence, OpenAI were to lose control of the algorithm and it got out there in the wild, so to speak, it really could be used by almost anyone. What makes a AI function in the way that we think of it today? 

Not this Skynet future thing, but how it is now requires massive amounts of processing power at data centers. The largest data centers that the world has ever seen are needed in order to deal with the inflow of requests that come in, run the algorithm and spit out the results. Which means that the limiting factor, for the moment, in artificial intelligence isn’t the software, it’s the hardware. 

And this is where we have a really big problem, and it’s not that far away. The ability to make the high end processing chips that Taiwan is famous for, requires, 100,000 steps, 30,000 pieces, 9000 companies, and they’re scattered around the world. The single biggest concentration is then the United States, which is something Americans conveniently forget when they’re talking about sovereignty. 

Number two, concentration is on the Taiwan centric zone. The single most important company is in the Netherlands, but it has facilities in Germany and in Austria and in California, in Japan. But you’re never going to be able to do the chips at all without all of these steps. And a lot of them are single point failures. 

So if you have any degree of globalization, it doesn’t matter really what the countries. It falls out of work. We can’t make them at all. And for the chips that we already have, life span when they’re in a data center is typically in the 3 to 6 year range. So when we get to the point where we realize that we can’t make the chips, we’re going to have a bit of a scramble to see who can control what’s left. 

And then the ability to use AI will shrink from something that you can all have on your phone to simply the handful of entities, whether governments or corporations, that are capable of having their own data center so they can run by themselves and that will be it. Until we reinvent the entire ecosystem and what we have been seeing with most government efforts around the world, including the United States, to reassure the sort of manufacturing it only focuses on the fabrication facilities, which is what is in Taiwan. 

It ignores the design, it ignores the material inputs, it ignores the photo mask, it ignores the wiring, ignores everything else that goes into a successful chip, much less the downstream stuff like testing and packaging that ultimately makes the stuff that ends up in a data center. No one, to my knowledge, is putting any effort into actually bringing the entire ecosystem under one roof, and I honestly don’t even think it would be possible anyway. 

There are too many pieces. There are too many players. And and if you’re looking at the United States, there are not enough technicians that are capable of doing it because we already have record low unemployment levels. So we are in a moment right now where AI is possible with ChatGPT 5.0 and all the rest that will not last. 

And in the not too distant future, we are going to see a technological regression as we lose the ability to make the hardware. And since it took us 60 years to figure out how to do that in the first place, it’s not something that we’re going to do in a season is going to take a mastery. Industrialization process of different parts of the world to do different things, coming together in different ways. 

And that is something that I am not looking forward to. But we’re going to see at the beginning of that within this next decade.

The Fourth Shale Revolution: Supermajor Tech

ExxonMobil neon sign at a gas station

ExxonMobil has introduced a new type of proppant that might just spark the next US shale revolution.

Prefer to read the transcript? Click here

Transcript

Hey all, Peter Zeihan here. Coming to you from Los Angeles on the California coast. And today we’re to talk a little bit about oil. There have been a couple of technological breakthroughs that I think are worthy of mentioning in the shale era. So ExxonMobil, big company, one of the largest players in the world, produces just under 5 million barrels a day. 

Has basically started mucking around with something called prop. So dial back…Hydraulic fracturing or fracking was basically how the United States produces 80% or more of its crude oil these days, as well as the vast majority of its natural gas. It’s not a fringe technology. It’s the backbone. What you do is you drill down vertically, and then you make a horizontal split that goes two, three, four, maybe even five miles. 

And then you inject water that is laced with sand. The water hydraulics, does not compress under pressure. So it cracks the rock apart. And then the water goes in with the sand and accesses tiny, tiny, tiny, tiny little deposits of petroleum. Then you stop the pumping. And because those tiny pockets of petroleum have now been exposed, they produce a back pressure that pushes the water out, but the sand stays lodged in the cracks, keeping them open so the flow can continue. 

The sand is called it, and it’s one of the biggest expenses in a fracking operation. Well, what ExxonMobil has now done is change the prop and is triggering what is basically the fourth shale revolution back backstory for that first shale revolution is when we figured out how to do this and brought out natural gas. The second shale revolution is when we figured out how to do this to bring out liquid oil. 

The third is when we built the infrastructure. Things like LNG facilities or chemical facilities or refineries to metabolize all this raw product where we’re now producing all of that stuff, all of that’s in the past. Fourth Revolution is taking the capital and the technological skillsets of the majors, like Exxon, and applying them for a whole new generation of technology. 

So one of the weird things about the shale revolution is when it started most of the super majors and kind of written off the American oil patch, and we had seen oil output from the United States dropped to historical lows well over the last century and a half. 

What that meant is we had small mom and pops that were doing everything, and they were trying everything they could come up with in order to get incremental increases. 

And that’s what generated the first few million barrels a day. Well, as time went on, oil does what oil does. And it rises and it falls and it rises and falls. And so we got a series of busts, and ExxonMobil was able to come in with its better capital position and buy up a lot of the smaller companies, to the point that it and Chevron now dominate the space and collectively produce almost 9 million barrels a day. 

Now you apply what Exxon has across its entire value chain, and you get a very different proposition. So for profit, specifically what we’re talking about today, they went into their refineries and they found waste product, something called petroleum coke. And they were able to manufacture that into a kind of a synthetic sand, if you will. The profit is where a lot of experiment has been going on and a lot of subsectors for the last several years, and you got some pretty expensive stuff that’s called ceramics called ceramics. 

It is ceramics. Petroleum coke is cheaper than that, more expensive than sand. But the real advantages it has that it’s a lot less dense, maybe 40, 50% less dense than sand, which means you can suspend it in the water better, which means it pushes into the formation better, which means it holds open cracks deeper in the formation. 

And for a small increase in cost, using what used to be a waste product. Exxon has seen their numbers increase by 10 to 20 to maybe even 30% in some wells. And that alone changes the math of the shale revolution a ten to a 30% increase in output for only a slim investment in what was a waste product. 

That’s amazing. And so the shale revolution is nowhere near done. You’ll hear people saying that eventually the shale revolution going to run out. There’s only much oil, but that misses the point. In the pre shale era, we were able to access about 10%, 9 to 10% of global energy reserves. There’s a lot down there that we just don’t have the technology to get to. 

The shale revolution doubled the percentage of what was accessible within the US space. So we’re talking about 150 years of output. All of a sudden we have access to something like that again. And we keep making these incremental increases, like with profit, that pushes the horizon back even further. So the shale revolution continues to set new records for output, adding somewhere between a half a million and a million barrels a day per year, and has now been doing that since 2009. 

You get a lot of output when you do it for that long. So this year is not the last year the shale revolution. Neither is next year or the year after or the year after that. 

Because the numbers keep getting better, the technology keeps pushing further, and the break even cost for what it takes to get a chunk of oil out in an economically viable way keeps going down.

The Semiconductor Frontier

Semiconductor being made

We’ve discussed how essential semiconductors are in our increasingly technological world, so here’s an update on ASML’s new High-NA EUV lithography machines.

ASML already builds the most advanced chipmaking tools, but these new models could enable chip designs at 1 nanometer or even smaller. Intel is betting on this new technology and could have mass manufacturing by the end of the decade, which would change the competitive landscape with TSMC. SK Hynix is similarly prepping new tech for these memory chips, which could give it the leg up on Samsung.

If it works, that is.

Prefer to read the transcript? Click here

Transcript

Hey, all. Peter Zeihan here. Coming to you from Colorado, today we are taking a question from the Patreon crowd. Specifically, if I could give an update on what’s going on with the new EUV machines that are coming out of the Dutch from ASML. Background, for those of you who don’t follow this, ASML is the company that makes the lithography machines that make high end semiconductors possible. 

Very, very, very short version. It’s basically a machine about the size of a bus that makes a tiny little laser that operates at an accuracy that is smaller than a DNA strand and allows you to etch semiconductors at the nanometer level. EUV in its current form can go down to about three nanometers below that. It basically loses coherence. And so there’s a new machine called a high numerical aperture that, in theory, can take you down to one nanometer and even below. 

And the idea is that the smaller you can edge your transistors, the more processing power you can cram onto a piece of silicon. And the more powerful the semiconductor on the other side can be. So EUV is a technology that’s been around since 2012, 2014 somewhere in there. And it was grabbed by what was then the industry underdog, which was TSMC in Taiwan. 

And over the course of the next several years, they leaped ahead of who the old industry leader had been. And that was America’s Intel. Now we’ve got the flip side. TSMC is reserving the right to maybe buy one of these new machines, the high end machines. But it is Intel that’s now betting their future on the new technology, hoping that they can repeat the feat that TSMC did and once again become the world leader. 

They have two of those machines are at their Hillsboro facility. They’re currently cranking out about. Well, their goal is to crank out about 10,000 chips a month, which is very, very, very small scale. They’re very much still in the testing phase. And it is just simply too soon to know if a this technology will work and be what its effects will be, and the be what its effects will be is really the question here. 

When we went from duv. Deep ultraviolet, which was the old technology, this is what the Chinese have still to EUV, which is now the standard for premium chips. 

The nature of semiconductor has changed because it wasn’t just about cramming more into less space, it was making them more energy efficient, was doing things with the architecture eventually leading to stacked chips. 

And so it wasn’t just a linear jump. And there’s possibility that with high end A, we will have another linear jump that will leave all the chips that that we make today behind, but we don’t know that until we have our first mass manufacturing run, that at the soonest will be at the end of calendar year 2026, and that will just be with a couple machines. 

Then we will have to have the industrial build out to build more of these machines. And these machines cost significantly more than a commercial aircraft. And then you’ll have to put them into the facilities and you have to start designing chips with the new hardware in mind. 

So we probably wouldn’t have enough chips to matter in a way that would really move the needle technologically before 2029 or 2030. That’d be more than enough to revive Intel’s fortune. But, you know, if, if, if, if then, there is one other company in the world that is trying out this new technology. 

It’s out of South Korea. It’s, SK Hynix, which is the company that makes the best, Dram chips. Those are memory chips. So what Intel does what TSMC does, those are GPUs. Those are processors. And those are important. But you have to pair it with a memory chip. And the Koreans excel at that. So as to the Dram side, SK Hynix is an industry leader, along with Samsung. They recently overtook Samsung in terms of total output, but in terms of quality, they’re pretty much neck and neck. And so now one of them has DNA and one of them doesn’t, if Na fails, I don’t think it’s going to be a disaster for this case. 

SK Hynix are already a fantastic company. But Samsung does have more capital coming up behind them. But really, what’s going on in Korea is nothing compared to the drama between Intel and TSMC. 

So let me give you a worst case. Best case for Intel. Worst case, this doesn’t work, in which case Intel is merely the second best chip manufacturer in the world. Americans get really pouty when they’re not number one, but this is still a solid company. And honestly, there are a lot more pieces of a supply chain under the hood of Intel than there are, in TSMC. The TSMC folks are great at what they do, but they basically follow the instructions that the designers gave them, and then they do the construction and everything in order to make the fabs function. 

But the real high value added work is done somewhere else. Intel does more of these steps, more like a traditional conglomerate, which means that they’re probably not as efficient at any individual one of them. But of the 100,000 supply chain steps that go into making a high and semiconductor, they have a higher proportion of them under their roof, probably as many as a quarter. 

So if this doesn’t work, Intel is fine. If it does work, TSMC doesn’t slip. They’re still making the chips that make today’s, silicon revolution possible. But then we also get a new frontier. The only caution I have is that currently, it takes a 100,000 supply chain steps to make a high end semiconductor with the high end American Aperture technology. 

We really don’t know what that supply chain is going to look like, but it would be very strange if it was simpler. So this is already the industry that is the most overextended and dependent on globalization and threatened by Trump’s tariffs threatened by the fall of globalization and threatened by the demographic crash. The idea it’s going to be with us from ten years from now is a stretch. 

But there is a little sliver of hope that, based on what the supply chains for high end are, maybe, maybe unlikely, but maybe it’ll be a little bit simpler. The first that we will have a good grip on, what that will look like will be in about a year.

Nvidia Purchases $5 Billion of Intel Stock

Photo of an INtel microchip

Nvidia announced a $5 billion purchase of Intel stock, but it’s not the game-changer that the headlines are making it out to be.

While Intel is America’s biggest chipmaker, it lags behind TSMC’s cutting-edge nodes. Nvidia is just a design firm, so they don’t possess the necessary manufacturing know-how to improve Intel’s capabilities. So, Intel’s need for the right ecosystem and advanced lithography to create the upper echelon of chips remains.

This is just another case of political appeasement. Nvidia has been in hot water with Washington and Beijing, so they’ll do just about anything to cool things down a bit. But hey, $5 billion is $5 billion.

Prefer to read the transcript? Click here

Transcript

Hey, all Peter Zeihan here come from Colorado. And today we’re taking a look at the 18th of September, purchased by Nvidia of roughly $5 billion of stock in American semiconductor manufacturing firm Intel. Now, Intel is by far the largest of the American fab companies. But it gets a bad rap because it’s not TSMC. TSMC, of course, is a Taiwanese based company that is the world’s premier. That makes all the leading processing nodes, especially if it’s below four nanometers. 

Intel is trying to catch up with mixed results. And, in the market, it generally is discounted significantly because it’s not TSMC. And every time they fail to catch up, they get punished. That doesn’t mean it’s not a good company. That doesn’t mean it doesn’t produce a lot of product. 

But if your goal is to make the best of the best, Intel doesn’t do it. 

This is not going to change that. Now, a few weeks ago, the US government under Donald Trump took a 10% share. This will give Nvidia roughly a 4% share. But let’s talk about how semiconductors happen. And then you’ll see that this is not nearly as big of a deal as it might appear at first glance. 

What typically happens is a large consumer of microchips, a Google and Apple, something like that comes to a company like Intel or TSMC, and says that we want to make a new chip that does X, Y, and Z. Here are the parameters we want in terms of performance. And Intel slash TSMC says you’re at the wrong place. You need to go talk to a design firm. 

And so you find a design firm and you jointly put this thing together. All the strategic architecture and then you take that back to your TSMC or your Intel, and then you redesign it again, and you build an instruction booklet that is a few thousand pages of all the steps that are necessary to craft each and every tiny little bit of what goes into each and every aspect of a semiconductor that is then farmed out to an ecosystem that is around the semiconductor fabrication firm, all the companies that build all the individual pieces, all the companies that do all the testing in the incorporation of those pieces into larger chips, motherboards and products. Hundreds of companies involved. And you then get this very thick instruction book, probably several thousand pages. Now, which you hand to TSMC or you hand to Intel. And they use that to follow the instructions to the letter to make the chips. 

Which means a design company like Nvidia partnering with a fab company like Intel. It’s not that it’s a negative, but it kind of misses all the steps in between. Now, Nvidia has been beat around the heads and shoulders first by the American government and most recently by the Chinese government, primarily over its seeming inability and unwillingness to apply technological sanctions and limit their sales to China. 

Nvidia is willing to bend the rules. There’s no argument there, and it seems that in order to placate the Trump administration, they’re putting a what sounds like a big investment, $5 billion into Intel. But this really doesn’t move the needle for anyone. It doesn’t speed up the process. All it does is perhaps give Nvidia an inside track to communicating with Intel in the circumstances, when they decide to build chips that are not cutting edge. 

So it makes a lot of people smile. It makes a lot of people think that, ooh, Intel is going to get better. Nvidia doesn’t have what Intel needs to get better. That would be TSMC. That would be ASML, the company that makes the high end lithography systems. That would be this constellations of dozens, hundreds of mid-tier companies that contribute individual pieces, a lot of which don’t exist in Intel’s network because they’re in Taiwan. 

So it looks nice. And having a few extra billion dollars is never a bad idea if you’re trying to expand your output. But if you’re thinking that this partnership is what is necessary for Intel to turn the page and all of a sudden move up to, say, 2 or 1 nanometer. No, because Nvidia doesn’t have that technology. Nvidia does design, not manufacturing. 

Don’t get the two confused.

Should the UAE Invest in a Tech Sector?

Photo of interior of computer chip

The UAE is pouring money into building a tech sector, focusing on semiconductor fabrication plants and data centers.

While semiconductor fabs are central to chipmaking, they require immense technical expertise, specialized labor, and integration across thousands of precise steps; meaning this is a nothing sandwich.

Data centers are more achievable, but that doesn’t mean they’re a good idea either. The UAE would need to subsidize access to scarce high-end chips, figure out the high cooling costs (because desert climate, duh), and even then, the geographic limitations will prevent them from becoming a global hub.

Prefer to read the transcript? Click here

Transcript

Hey, all Peter Zeihan here. Coming to you from Colorado. Today we are taking a question from the Patreon page and specifically about tech and some of the things that are going on in the Persian Gulf, specifically, a number of the Persian Gulf Arab states, most notably the United Arab Emirates, are splashing around a lot of cash and trying to build a tech industry. 

There are two forms of taking. The first is they’re trying to get kind of like what happened in Arizona, a high end semiconductor fabrication facility. And second, they’re getting a data center. Two very different pieces of technology that have very different requirements. So let’s start with the semiconductor fab facility. 

Semiconductor fabs like the kind that we now have outside of Phoenix, the one that’s being built outside of Columbus, Ohio. 

The ones that are in Taiwan, are incredibly sophisticated. And what people tend to forget is that these are not just like assembly locations. They bring some of the most advanced machining in one place. They bring some of the most advanced materials into one place. They bring some of the most sophisticated designs in one place. And basically, you’ve got something in excess of 10,000 pieces that come together on the floor of the fab. 

It’s not simply an issue of making a semiconductor. You have a high end machine that’s called a extreme ultraviolet machine that does etching, and you have to dope. Well, let me let me back up. Just show you the whole thing. The whole process. Step one. You buy some really, really expensive sand silicon dioxide. That’s just pure, pure, purified. 

Usually only comes from the United States. You melt it down and know that you put in a seed crystal. And over the course of several days, sometimes weeks, you grow it into a crystal that weighs more than a car. You then slice it laterally into wafers, and then you take those wafers into your semiconductor fab facility, because these are all done in different places. 

And then you hit it with lasers that come out of the EV system. You dope it, you bake it again, you dope it, you bake it again, you do that, you know, ten times, 20 times, 90 times, and eventually you get a bunch of semiconductors on your disk. You then break those into pieces and test them and eventually incorporate them into actual hardware, like, say, a motherboard or a flash drive. 

And then goes into the intermediate products trade. So fabs are essential. Absolutely. But they are one step in an entire process that has thousands of steps. They just happen to be where a lot of these steps come together. They are not the high value added part of the process. That’s going to be almost everything else. Does that mean that they’re not important? 

No. Does it mean you can do it with unskilled labor? No. 

the United Arab Emirates have skilled labor? No. So if the UAE were to pay the $2,025 billion it takes to build a top rated facility, then they would have to do exactly the same thing that the Chinese have had to do import the labor to make it run the most exacting work that is done in a high end fab facility is the quality checks at every step. 

And that is something that if the chips are above, say, 20 to 30 nanometers, the Chinese can’t do it at all. And the idea that the UAE could do it is absolutely laughable. So if they did built this, no one would want to probably work in the unless the pay was absolutely immense and you would have basically a white elephant project generating very error prone, high cost items. 

That’s probably going to happen. What is less unlikely would be, say, a data farmer or data center. This doesn’t require nearly. The maintenance work is not nearly as, worker intensive. Basically, you get a bunch of GPUs. You build into something called a module with a bunch of Dram and Nand chips. Now, Dram, our memory chips and Nand are long term memory chips. Flash memory, short term needs power. Nand is, long term memory, not as quick, doesn’t need the power, and the GPU is all the processing. So you basically build a module and then you put a bunch of them in a server, and you put a bunch of server blades in a rack, and you put hundreds of racks in a room with really good cooling, and then you just let it run. 

Data cables coming in, data cables going out. Traffic comes and goes, you can house AI algorithms on it. You can. How’s your AOL account on it? Whatever you want. Two problems. Number one, all of the hardware is really, really expensive. And demand for the high end chips is very, very high. So most server farms do not have the sub seven nanometer chips that are, for example, necessary for most AI applications. 

Second problem latency. As a rule, you want your data center to be as close to your demand as possible. So the United States of various quality sets of about 10,000 data centers. And we try to put them either right outside of a population center or somewhere roughly in the middle of the country for trans coast traffic. So the idea that the UAE, with a couple first world cities is going to need a couple of data centers makes perfect sense. 

The idea that it’s going to be a global information hub, no, because there are no countries near it that generate the volume and quality of data that would want to go all the way to Dubai and Abu Dhabi before then moving on. So if the Emiratis decide to go down this path. This won’t be nearly as much of a white elephant project as, say, building a fab facility, but they would have to subsidize it in order to get the high end chips. 

And it appears that’s exactly what they’re doing. There is one of what’s supposed to be one of the world’s most advanced data centers, under construction in the United Arab Emirates right now, if everything goes to plan really does with these things, if everything goes to plan, it’ll become operational before the end of 2026. But it will be in a very expensive place to operate because the single largest expense for data centers is cooling its electricity. 

And I don’t know if you knew this, but the UAE is in a near equatorial desert. So the operational costs will be massive. And while labor is not a huge component of a data center when it comes to costs, they still don’t have the labor force to do even that. So if they do this, they seem to be doing it. 

It will be very expensive and it’ll just kind of be a feather in their cap. It won’t be actually something that a lot of people want to use.

George Jetson Would Be Disappointed with Autonomous Vehicles

A Waymo autonomous driving vehicle

I hate to be the bearer of bad news, but it doesn’t look like we’re all going to have personal Waymos anytime soon. There are four major hurdles.

The subsidies that have made a lot of the progress on autonomous vehicles possible are being phased out, so all that infrastructure that is needed won’t be made. Lithium is the backbone of EV’s, but most of the processing is in China; the US would need to build local capacity for this to work. Tesla, the leader of US EV innovation, hasn’t released a new model in years. And of course, the chip shortages that are coming soon will restrict growth.

So, mass adoption for electric and self-driving cars isn’t happening. There are niches where this technology will likely do well…think convoy trucking.

Prefer to read the transcript? Click here

Transcript

Hey all, Peter Zeihan here. Coming to you from Oregon. And today we’re taking a question from the Patreon crowd, specifically what I think the future of electric and automated vehicles are specifically self piloting sort of stuff. It doesn’t look great. A couple of problems here. First of all, there are very few electric models that make sense to your average consumer without a huge amount of subsidies. 

And under the Trump administration, those subsidies are basically going to zero. And if you don’t have the infrastructure in place to support charging, and if you don’t have the subsidy support, mass adoption, you’re never going to build the infrastructure that you need to support support charging. So outside of some very specific places where there just has the right concentration of infrastructure and support places like say, Oslo and Norway, this really doesn’t have much of a future in the rest of the world, most notably in the United States, where distances are large, about half the population doesn’t have a grudge to charge. 

That’s problem one. Problem two is manufacturing. Right now, the vast majority, something like 80% of global lithium is processed and turned into lithium metal in China. And that is an infrastructure we’re going to lose. So if these are technologies you really want, you have to build the processing locally. And the Trump administration is actually moving in the opposite direction and removing some of the grants and the subsidies that the Biden administration established to build up this sort of infrastructure. 

So that’s problem two, problem three is legacy infrastructure. It’s not just that the United States loves their cars, but in the United States, the sort of engineering that is necessary to do this thing is really held within one company. And that one company is Tesla. Tesla is arguably the most what, until recently, the most subsidized company in modern American history. 

And those subsidies are also going to zero. The bigger risk here is that Elon Musk’s entire corporate empire is going to dissolve over the next couple of years. And I just don’t see Tesla, which hasn’t issued a new model in three years, being really part of the American automotive future. As for other companies, you know, Ford, Chrysler and the rest, they’ve all tried to get into EVs, running them alongside their other vehicles. 

But they have proven to be not as popular because, again, that infrastructure doesn’t exist. And that means you have a huge upfront cost for people who want to do it. And so with the last data we have, which is about a year ago, over three quarters of the people who owned an electric vehicle in the United States, it wasn’t their first car, their second car, it was their third or their fourth car. 

It was a showpiece. It was a talking point. It was allowing them to beat their chest and say they were environmentalists. Although I would argue, that if you look at the full cycle for producing an EV, the amount of carbon and energy it takes to build it in the first place, it’s actually not a very smart environmental choice. 

And then fourth and finally, and the one that’s going to become just crushingly important in the not too distant future, are the chips. An electric vehicle that is capable of automated piloting requires about 2000 U.S. dollars of chips. About half of those are relatively cheap and easy to obtain. The other half are the high end ones. And in the world that we’re moving into, that is the globalizing the capacity of the world to make the high end chips is going to go to probably zero, which means anything that is more advanced than, say, the iPhone 12, roughly, is something that we’re simply not going to be able to produce in sufficient volume. 

If you’re going to do true auto pilot, you need a significant amount of processing power on your vehicle that is not dependent upon the weather, that is not dependent on an uplink. It has to be done locally, otherwise you’re driving yourself. So what that means is at the end of the day, your individ person vehicles are not probably going to be auto piloted, or even electric, for much longer. 

That doesn’t mean the technology is going to die. It just means it needs to find a niche where it’s more appropriate. And what we’re probably going to be seeing is convoy in for trucks. Basically, your first vehicle has a driver software engineer, mechanic kind of person, maybe 2 or 3 people in it, and then a line of two, three, 4 or 5, six, seven trucks follows that first vehicle nosed tail. 

The technology that is required for that does not require electric vehicles, and it requires a much lower quality of chip to make it happen. That is something we can do with today’s technology. The reason that hasn’t happened yet has to do with legal issues. Until Congress codifies where fault lies, when something goes wrong, it’s difficult for auto manufacturers to really get into the business of making automated vehicles function. 

So the question is, you know, who’s at fault when there’s an accident? The driver, the person who did the last software patch, the manufacturer, the person who designed the sensor until that is sorted out, it’s going to be very difficult for EVs, as opposed to EVs to get to deeper claws into the American transport system, because if we don’t know who we’re supposed to sue when something goes wrong, then the liability gets spread out everywhere and it becomes a real mess that nobody wants to touch. 

All right, that’s my $0.02.

The Revolution in Military Affairs: Artificial Intelligence

ChatGPT logo with a synthetic brain hovering above

AI is working its way into just about every aspect of modern life. I mean, who didn’t fall for that video of the bunnies jumping on the trampoline. But artificial intelligence might not be the game-changer in warfare that you think it is…at least not in the short term.

AI promises faster processing, targeting, and decision-making, which all sounds great, until you throw in the wrench of deglobalization. As the globalized world collapses, the semiconductor supply chain will fall apart. The most advanced chips will not be able to be created anymore. Between the bottlenecks of EUV lithography and the countless single points of failure, we’ll be stuck with what we currently have (or yesterday’s tech).

When you factor this into military applications, it means older systems like cruise missiles and smart bombs will be mainstays. Fully AI-enabled systems will be severely constrained and reserved for the really important stuff.

Prefer to read the transcript? Click here

Transcript

Hey, all. Peter Zeihan here come to you from Cassidy arch. And where am I? Capitol reef National park. Sorry, it’s been a busy week. Today we are going to close out the series on the revolution in military technology. As advances in automation and digitization in materials science and energy transfer come together to remake how we fight. 

And we’re going to close out with something that you probably don’t need to worry about. And that’s artificial intelligence in war. The whole idea of AI is it can process faster than we can’t make decisions faster than we can, and potentially target with lethality faster than we can. 

I don’t think it’s going to happen. The problem is that the semiconductor supply chain for the high end chips that are capable of doing AI, and as a rule here, the cutting edge is going to be three nanometers and smaller, simply isn’t going to be able to survive the globalization age. So any chips that are not made in the next relatively short period of time, no more than a single digit of years, are really all we’re going to have for a good long time. 

And that means that the machines that are going out and doing the fighting have to rely on something that is older, that is not capable of processing and has to be linked back to something back home, either via wire or telemetry or some sort of radio communication. And that makes for a very different sort of beast. 

There are roughly 30,000 manufacturing supply chain steps that go into semiconductors. The high end stuff. And there’s about 9000 companies involved, and about half of those companies only make one product for one end user. There’s literally thousands of single point failures, and it only takes a few of them to go offline for you to not be able to make the high end chips at all. 

But the place that I think it’s going to be most concentrated, the place where we’re all going to feel like the place where is going to be obvious is going to be with the lithography. Specifically, we are currently using something called extreme ultraviolet, which is done by a company called ASML out of the Netherlands. And they are the world leaders in all of this. 

There are other companies that do the fabs other than TSMC and Taiwan, but the lithography can really only be done by the Dutch. And it’s not like this is one company. This is a constellation of hundreds of companies, and every time one of them either has a generational change or goes public, ASML basically sweeps them under the rug, absorbs them completely, puts the staff in different areas and puts it all under referential lockdown so there is no way to duplicate what they have. 

And so if you take this gangly supply chain that wraps the whole world and any part of that breaks, we can’t do EUV at all. And that means functionally, no chips that are worse than or better than six or 7 or 8 nanometers based on where you draw the line, we can still do something called deep ultraviolet, but extreme ultraviolet. 

It just becomes impossible. And that means that the best chips that we will have ten years from now are going to be very similar to the best chips we had ten years ago. And that limits what we can do with any sort of technological innovation. For the purposes of the military, it becomes very, very truncated. Old weapons like smart bombs and cruise missiles actually don’t use very sophisticated chips. 

20 year old chips are just fine. It’s the high end, the thinking, the processing, anything that’s more than guidance and requires a degree of decision making, that is what’s going to be off the table. So while I applaud all of us for having these conversations about the implications of AI, what it means for the workforce, what it means for culture, what it means for morality and legality. 

These are great conversations. It’s very rare that we get ahead of the technology in discussing what it can and can’t do, and start thinking about the implications for us as people, but I think we have some extra time because once this breaks, it’s going to take us 15 to 20 years to rebuild it. And that was back before everything accelerated with the Chinese fall and the Trump administration. 

Now it’s probably going to take longer. So have these discussions. I think that’s great. But it’s really probably going to be a problem for the 2050s.

The Revolution in Military Affairs: What’s Ahead

Photo of a soldier pointing to a tech screen

Before we close out this series on military tech, let’s discuss what military advances are on the horizon (and our last episode will cover something we don’t need to worry about).

Many of the larger evolutions coming down the pike are related to drones. Whether it’s strikes, surveillance, detection, or deadly jobs…drones will likely be taking it on.

These technologies are just the beginning though. As battery science improves and more advances are made, the battlefields will be going through countless iterations.

Prefer to read the transcript? Click here

Transcript

Peter Zeihan here coming to you from a foggy Colorado today. We’re do another in our Military Revolution series how changes in Materials Science and Data Transfer and Energy storage are shifting, the way the military works, and some of the new things that will be seen in the not too distant future. Today we’re going to talk about some edge cases that are likely to move into the mainstream in just the next few years. 

And these kind of fall the two general categories. First, you’ve got the topics where humans just aren’t the best tool for the job. These are things where they’re either dangerous, or expensive, we have to train someone up to an extreme level to do a job that then has a high mortality rate. 

You know, things you don’t want people doing. And the first one of those is saving other people, search and rescue in a combat environment uses a huge amount of resources to cover a large amount of land to save 1 or 2 people. It doesn’t matter if it’s a fighter pilot, it’s been shot down or someone who’s been shot out of the field having drones do this not only builds up your combat awareness for the field in general, but also allows you to provide, say, targeted supplies and of course, guide the real force in to pull the person out of trouble. 

The general topic of recon, something that is starting to be called perch and stair. Basically, you have a recon drone, but rather than flying around at altitude, it finds a place at, say, a quarter of a building and just parks and stays there. Maybe it has solar panels on its back so it can extend its battery life and it just looks around. 

It’s a mobile sensor that, for the most part, isn’t mobile. You know, you might call this a spotter or a spy in another condition, but if you can automate that, and instead of having one guy in one place that might be able to move around, you can have hundreds if not thousands of mobile sensors that can extend their life to span by just not flying the whole time. 

And third is something called an underwater swarm. Submarines are among the most expensive things that most modern navies can float. And if you can throw a few dozen things into the water, not only do you get some excellent acoustic collection for purposes of locating them, you know, you put like a one kilogram charge on each one. It doesn’t take a lot of those to take a multimillion dollar sub completely out of action forever. 

So these are technologies that you apply them to. What we know we need. And all of a sudden they really are game changers in terms of efficiency. Now the second category are things that we used to do and maybe even used to do well, but we haven’t done it for a long time. Keep in mind that the US military has not been preparing to deal with another peer adversary until just a few years ago. 

And the immediate post-Soviet era. We thought of the Russians no longer as an enemy. And so we stopped preparing to fight a global conflict with them. We then spent 20 years in the war on terror, focusing on counterinsurgency. That means going against the Taliban. And that means you don’t really need air power. You really don’t need air defense. 

And so certain aspects of our military were allowed to atrophy just from lack of use. And the two biggest ones are air defense suppression and hunting mines, whether land mines or sea mines. The general idea was, you know, if there is no big force out there fielding things you need to shoot through, then why would you maintain an entire arm of your force doing things that are just going to sit around? 

So, for example, we really only have a couple of minesweepers left, but naval minesweepers that are drones are a great idea. In essence, you have a drone that’s hooked up to your ship as it’s puttering around at a relatively low speed doing a sonar capture. You locate the drones and you send out a suicide drone to take it out. 

They’re already doing this in, say, Romania in the western part of the Black Sea. You can use aerial drones with radar to triangulate metal signatures in the soil and locate landmines before anyone can step on them. 

And for air suppression. Back in Vietnam, we had this thing called a wild weasel. Basically, it was a bunch of suicidal maniacs on a plane who would fly into Vietnam ahead of the bombers to activate air defense. Well, do that with drones. Don’t do that with a manned plane. In fact, do that with drones backed up by other drones so that by the time the real bombers get in the air, defenses are already gone. 

Same basic concept holds for coastal patrol. Now, the United States has never really been good at coastal patrol because we have oceans between us and everybody else. But this is one of those technologies that everyone else is going to find really useful. Again, in the post-Cold War world, everyone slimmed down their military spending, with navies seen as just something you would never need again. 

History was over. It was a world of commerce. Why would anyone shoot at anyone’s commerce? Well, that’s gone, but the time it would take to build up a coastal fleet and coastal patrol capability is going to be measured not in years, but in decades. Or you can just have a fleet of drones. It basically flies patrols out on your coast and then, if necessary, a more robust naval vessel can go out to take care of whatever the issue happens to be.  

So these are all things we’re going to see in probably just the next five years, certainly the next ten. And this is just the leading edge. These are things that me as a nonmilitary guy can kind of just think of based on the gaps in the system right now as the military technologies continue to evolve. We’re going to see radical applications of all of these. 

And keep in mind that drones are really just the leading tip of this. We don’t know what our material science is going to be in the next five years. Maybe we’ll get a new battery chemistry that allows for longer loitering, that generates an entirely new field of military tech. We’re just at the beginning here.

Will Trump Pump the Brakes on Greentech?

Both in the US and globally, the green energy transition has been all the rage for the past few years. With President Trump’s second term kicking off, how will it impact the green transition domestically and beyond?

For the green folks outside of the US, the impact should be minimal. Since the US doesn’t manufacture most Greentech components or provide much financial support, Trump’s influence is (mostly) contained to the US. But the story isn’t so pretty for those in the US.

The main challenges for the green transition in the US are transmission infrastructure and financing. Federal support is crucial for developing the infrastructure to get the energy from where it is generated to where it will be used. Trump could make this development and coordination process much harder. Wind and solar projects require more robust financing than a traditional fossil fuel plant, so cuts to federal incentives or subsidies could make these projects unviable.

Without federal backing, many of these green projects would stall. Private investors might try to step in, but they can’t match federal funding levels. Trump has the ability to significantly slow down the green transition, but at least that doesn’t extend beyond the US.

Prefer to read the transcript? Click here

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.

Learn about medshare here

Transcript

Hey, all Peter Zeihan here coming to you from Colorado today. We’re taking a question from the Patreon page. And that’s specifically what sort of impact can Donald Trump have on the green transition, both in the United States and wider abroad? Abroad, very, very little. The solar panels aren’t made here. The wind turbines are not made here. And U.S. financial support for anyone else’s transition is well below $1 billion a year. 

So, you know, you know, whatever. It’s all about what would happen here and here. The the federal government has a lot of, means for changing the way the green transition works. A couple things to keep in mind. Number one, green technologies, as a rule, require a great deal more transmission infrastructure because where most people live is where it rains. 

And so you can grow your own food. We don’t have a lot of desert cities. So in most cases we generate power with coal, nuclear, and natural gas relatively close to where we live. And so transmission for most power plants is well under 50 miles. But most of the places that are very sunny or very windy are not within 50 miles of where we live. 

It’s in the Great Plains, it’s in the desert southwest. And so you have to build these pieces of infrastructure to generate power well away from where people are. And then you have to wire that power to somewhere else. And that often means crossing jurisdictions. And if you cross a economic or political jurisdiction, the regulatory burden becomes more robust. 

And it’s up to the federal government to try to smooth that out. So if all Donald Trump does is not smooth things out, becomes a little bit more onerous to build green tech anywhere because you can’t hook it up to a source of demand, then that’s problem one. Problem two is much bigger. You see, if you’re doing a conventional, facility, whether it’s coal, natural gas or a nuke, only about one quarter of the cost of the facility is in the upfront construction. 

And then linking that up to the grid, most of the rest is fuel, especially for coal and natural gas. So as a rule, it varies based on where you are and how close you are to the fuel source. As a rule, about 80% of the cost of the lifetime cost of a coal or natural gas facility is the fuel. You basically buy it and burn it as you go. And so with that sort of model, you only have to finance the initial 20% that it’s required for the construction of the facility and looking it up to the grid and everything else. 

You have an income stream to defray and ultimately overpower the cost of the fuel moving forward. It’s not how green tech works. The whole point of solar and wind is that you don’t have fuel. The fuel is free. Well, that means that most of the costs, almost all the costs are upfront. Over two thirds go to the construction and linking it up to the grid. 

So the degree of financing you need megawatt for megawatt is more than triple what you need for a more conventional fuel system. Now, one of the things to keep in mind in the United States is that capital costs have roughly increased by a factor of four since 2019, as the baby boomers have retired, and the money that they used to have in stocks and bonds, that fueled the sort of capital environment that we had ten years ago just no longer exists. 

They’ve all been liquidated and they’ve gone into T-bills in cash, which is driven up the cost of financing for almost everything, including power plant expansion. Well, if you’ve seen the cost of capital increase by a factor of 4 or 5, and you have to finance three times as much for wind and solar as you do for core natural gas, you can see where the problem is. 

This is normally where the government would step in with concessionary deals on whether it’s on taxes or directly on financing in order to help bridge that gap. And so all Donald Trump has to do is say, I’m not going to finance this stuff anymore, and a lot of it is going to go away, even if, as isn’t the case in the desert southwest or in the Great Plains, solar or wind are already cheaper on an all in cost basis over the entire life of the project. 

But that’s not the number that matters. Part of the problem that I’ve always had with the green communities, they keep using this thing called levelized cost of power, which shows how over the life of a project, the cost of solar and wind has gone down and gone down and gone down. And it has. But they assume that there’s no problem with intermittency. 

So like when the sun sets, solar doesn’t work anymore. If you pair a more realistic cost structure because you know you want electricity after the sun goes down. Hello. With financing the issue, then the federal presence in the financing world really is critical. And even in projects that make a huge amount of sense, not just environmentally but economically. 

If you can’t get that financing right, you can’t have the project. Private industry can step in, but it’s going to be a hard sell to do financing for something on concessionary terms, for something that it’s going to take longer to pay out as compared to a colder natural gas plant. And you might get local and state governments kicking in some for political and environmental reasons. 

But there’s no way that they can compete with the sheer volume that the federal government can come up with. So we should expect a lot of these projects to slow down quite a bit. Even if Donald Trump doesn’t call them out by name is something that he doesn’t like. You interrupt the financing and you simply don’t get much new construction.