And here we are again…at the end of the world. With Peter Zeihan
Peter Zeihan joined Burr & Forman to deliver a presentation on the geopolitical, economic, and technological forces reshaping the global landscape.
Transcript
INTRODUCTION
[Host]: I, and the general commercial litigation group, want to thank everybody for attending — friends, families, clients, everybody. If you’ve never heard Peter Zeihan, you are in for a real treat. He’s a New York Times bestseller, and he consults with governments, militaries, private groups, hedge funds — you name it. If you’ve read his materials, you know why. I think everybody will get something they can take with them.
We talked with him, and he’s got topics everybody’s going to love — you’ll probably be able to guess most of them, because they change daily, and he’ll hit almost all the current topics: wars, and whatnot, that are going on right now.
When we finish, some of his books are up here — this is his latest, a redo of The Accidental Superpower ten years later. Feel free to grab one — they’re autographed. If we run out, Bri has more at the desk.
Ask questions — love it. We’ll have microphones available; when he’s finished, just raise your hand and someone will bring you one. For people online, we’re monitoring the chat — type your question and Bri will read it aloud so everyone can hear. We’d like you to use a microphone so the questions can be heard by everybody. It’s being recorded, too, so if you registered — online or here today — you’ll receive a copy of the recording later. Feel free to take notes, or just know you’ll get a copy. If you’d silence your cell phones, that would help.
One more plug: Peter’s got a new book coming, totally different — his first fiction book, coming out this January. I’ve been lucky enough to read part of the manuscript, and it’s going to be exciting and interesting, following some of the same theories from his other books. I had the pleasure of being killed off in the first eight pages — I’d wanted him to name a character after me, and I didn’t even make it out of the first chapter. Anyway — thank you, Peter. Everybody sit back and enjoy. And there’s more lunch in the back if you need it. Peter, the floor is yours.
PETER ZEIHAN
Okay, we’ve got a lot of ground to cover. I don’t know if you’ve noticed, but things have gotten more than a little crazy this year. I’ve got a few personal anecdotes we’ll throw in, because I’m kind of living in the churn.
Two-minute backdrop. Everything about globalization was designed to fight the Cold War. We sent our Navy out after the war to patrol the global oceans so anyone could trade with anyone else — we secured the entire system to allow that to happen. It was the first true globalized system, and for us, it was a bribe.
When the Cold War ended, we started moving on, and in a series of ever-more-populist political contests, we went into less internationalist leadership — culminating in Trump one, Biden, and Trump two. Trump two is more nationalist than Biden economically; Biden was more economically nationalist than Trump one. Now we’re at the end of that model. Globalization was always going to end — the only question was when. It’s now.
We were also always going to get here from a demographic point of view. Until World War II, most of the planet was agricultural, and on a farm, kids are free labor. But once globalization opened up services and manufacturing jobs, people moved into towns to take them — and kids stopped being free labor. Fast-forward eighty years, and we just stopped having them. A demographic structure that originally looked like Nigeria’s today evolved into what Japan’s is today. The whole world has been going through this process.
Countries that industrialized later could learn from the countries that did it earlier, and their birth rates dropped even faster. Anyone know the fastest-aging society in the world right now? Yemen — their birth rate has dropped 90% in fourteen years. The whole world is doing this. Between 2025 and 2035 was always going to be the window when most of the rich world — and a substantial chunk of the developing world — aged past where Japan is now. And if you don’t have enough people to buy stuff, there’s no trade, and without trade, there’s no globalization. So we were always going to get here, and this decade. What Trump has done is forward-load the process and condense the adaptation time. It’s happening right now, and what’s left is single-digit years — low single digits — assuming no more policy changes. And oh my god, there are so many more coming.
Two countries are partial exceptions. First, the United States. For all the grief I give the baby boomers — and they deserve it — they did one thing none of their cohorts around the world did: they had children. The millennials, same — as much grief as I give them, they have something none of their global cohorts have: they exist, which means there’s a generation to work, pay taxes, and have kids for the coming decades. Our birth rate has ticked up recently, but it’s still dropped to about half of what it was twenty years ago, which is a problem — if we keep aging at this rate, in four or five decades we’ll have a major crisis. But that’s nothing compared to the rest of the world.
Generally, when a society stops having kids for a quarter century, economic growth actually accelerates — money once spent on daycare, diapers, and school instead goes to pizza, cars, homes, vacations. That’s higher-octane growth that drives more investment. But you can only do that once, because after twenty-five years you’ve lost the people who’d generate the next generation. China and Europe have done fairly well over the last 35 years because they spent on themselves instead of the next generation — but that’s over now, and their growth rates have plummeted. In China, conservatively, growth has dropped 75% over the last decade. That’s not coming back.
The other exception is the People’s Republic of China — for the opposite reason. I showed this slide last year: the “missing millions” — about 100 million people that outside observers say don’t exist. Chinese statisticians now publicly admit they overcounted by 100 million people under age 40. Internally, they’ve now settled on something closer to 300 million, and are even floating the possibility it’s as high as 500 million. Even at just 100 million, this is by far the fastest-aging demographic in human history. By 2030, there won’t be an economic model humans have ever invented that works with that population structure. Regardless of what you think about the Washington–Beijing competition, know this: whatever we import from the PRC is going to zero. That should be the starting point for any policy.
Manufacturing. Everyone’s obsessed with manufacturing right now, and rightly so — we need to roughly double the size of the industrial plant. That normally takes thirty years; we maybe have three. We’re going to fail at that.
Here’s a supply-chain matrix I’ve shown before: if you’re on the right, you’re in NAFTA; on the left, mainland China; toward the bottom, a short supply chain (half a dozen steps); toward the north, hundreds or thousands of steps. Here are the manufacturing processes everyone obsesses over — this one slide alone could take all day, so we’ll just look at one piece: high-end semiconductors, seven-nanometer nodes and smaller. Chips like this are made almost exclusively using extreme ultraviolet (EUV) lithography — the Dutch company ASML makes the machines — but that’s just one part of the process. There are roughly 100,000 manufacturing steps and 9,000 companies involved, spread across more than 50 countries. Half of those companies do only one thing for one end customer. Lose a handful of them, and you don’t get high-end chips at all. Every smartphone, every laptop, EVs, satellite communications, and above all, artificial intelligence — all in this category. We’re going to lose all of it when globalization breaks. That’s unavoidable.
Both the Trump and Biden administrations focused on bringing in high-end fabs — essential, but not the highest-value-add part of the process, and by itself it won’t matter, because the fab is just where everything (like EUV) comes together. Roughly 40% of the supply-chain steps are on the front end before the fab, and 30% on the back end. You need the EUV machines — the most sophisticated machines humans have ever built — and you need the logistics for thousands of inputs to converge at the fab. The back end is the hard part.
The process, roughly: melt a vat of silicon, lower in a perfected seed crystal, and over a couple of weeks draw it up as it grows into an ingot about the size of a Volkswagen. Slice it into thin discs, dope them with chemicals, and run them through the EUV machine to etch them. Bake, dope again, repeat — somewhere between 9 and 90 times over 8 to 12 months. That gets you semi-finished semiconductors — the easy part. Then they go elsewhere to be cut into individual chips, tested, wired, and built into products like thumb drives, motherboards, and circuit breakers, which then go into final products. From crystal growth to finished product purchasable by a consumer is typically about two years. This does not move quickly.
The chip we’re most concerned about for AI — especially large language models like Claude and ChatGPT — is the GPU: graphics processing unit, about the size of a postage stamp. GPUs sit in every gaming console, designed to parallel-process graphics. Several years ago, when large-language-model AI became the technological breakthrough, GPUs were the only hardware capable of large-scale parallel processing — so instead of one per gamer console, we put several (four, in a typical case) into a server blade, and tens or hundreds of thousands of blades into a server farm. Parallel processing generates a lot of heat, so each blade has copper cooling plates: roughly 100°F water goes in, 170°F water comes out. These run 24 hours a day until heat destroys them — usually four to six years.
So when this hardware system breaks, we get a two-year window where chips already in the pipeline finish reaching end products, and four to six years where what’s already installed keeps working — and then we have none of it. Whether that’s enough time to build an alternate supply system or adapt the technology, I can’t say. Building a high-end fab typically takes a few years, but that’s not the sticky part — the sticky part is everything else on the front and back end, which here in the US we’ve largely ignored because it’s not flashy like a fab. It takes all of humanity to build this supply chain. If we’re looking at a more constrained trading environment, with fewer partners closer to home, rebuilding it from scratch is, if we’re lucky, a 10-to-15-year process — probably longer. At some point this won’t be subtle: it’ll just stop, and we’ll have a few years while existing equipment wears out before we have to find something else.
So what’s “something else”? Here are the various AI types: LLMs at the top, plus images, video, autonomous vehicles, smartphones, robotics, and classic machine learning — and underneath, the hardware each depends on. Everything in red is something we’re not going to be able to do much longer; purple gets dicey. We’ll still have machine learning — yay — but the rest, no. We’re close to a plateau on what we’ll be able to achieve with AI in our current lives.
The only other path forward is redesigning what a semiconductor is, using less-advanced lithography than EUV. That’s possible — a 14-nanometer chip made with older lithography can run large language models, but you need five times as many chips for the same processing power, and they generate roughly five times the heat per calculation. That’s a more than 15x increase in the cost of running a data center. In that environment, this isn’t a technology available to everyone — we’ll have countries, governments, and rich individuals hoarding chips. That’s a new level of geopolitical competition nobody’s prepared for.
I showed this “what to watch” slide last year too: total US construction spending on manufacturing plant — building the things that build things. I told you that with Trump coming in, it had flatlined — not dropped, but flatlined — and that if it dropped, given how little time we have left, we’d be in serious trouble. That’s what 7,500 tariff changes in 16 months does to business confidence. People are finishing projects they started, but outside of data centers, almost nobody is breaking ground on anything new — and now almost half of that spending number is data centers. We are deindustrializing at the fastest rate in American history right now.
Here’s every high-voltage power line in the country, including the long-range ones. One big challenge: no community can really rely on another for electricity. Before even getting to data centers, just doubling the industrial plant means we need at least 50% more electricity — and with one exception, there’s nowhere in the country you can put a power plant and supply the next town over. The exception is Appalachia, because in the ’60s, ’70s, and ’80s, several administrations decided it was cheaper to generate power where the coal is and wire electricity to population centers. That’s the only place in the country where that’s ever happened.
Under Trump one and then the Biden administration, the regulatory structure loosened to make cross-state and cross-jurisdiction power lines easier — that’s in place now. But because of the general degradation of the business environment — which I’d argue is the most hostile in my lifetime, largely due to federal regulation and erratic policy — nobody’s breaking ground. Several projects that started three years ago have all stopped.
What does work in an environment of strained manufacturing, limited power, and limited chip access? A lot of things that can blow up.
Here’s a series of drone technologies — some are successors to others, mostly parallel advances. First-person drones: the drone has a camera and a digital tether — satellite, cellular, radio, or physical fiber — letting you see and control it in real time. That lets you change mission parameters on the fly and hit very specific targets — don’t hit the tank, hit the tread on the right side in back, where the ball bearings are, so it’s taken out completely. These are the Lucas — the American equivalent of the Shahed; they look similar but function very differently. The Ukrainian naval drones used to sink Russian battleships have worked well — each carries about half a ton of explosives. These have been flying back and forth over the front line, making everything within 20 kilometers of it inaccessible for either side.
Remember the Predators and Reapers? They’re old enough to drink now if they were people. And the newer Tomahawks — similar, we control them as they go. What all of these share is that because of the tether and active control, if it’s cut, the drone crashes, returns home, hovers, or “Peter Pans” — military term for “second star to the right, straight on till morning”: it just flies off and is lost, which is humiliating for the operator. The Russians are the best in the world at electronic warfare, and that’s been a major limiting factor for all these systems.
To get around that, you have memory drones — built around a chunk of NAND flash memory, the kind in a thumb drive: cold storage, holds data with the power off, not fast or fancy, but it doesn’t need cutting-edge chips — more like 30 nanometer, with vertical layering for density instead of speed. You preload it with information; the drone doesn’t even need a camera, just the ability to listen for a positioning signal — GPS, or Russia’s GLONASS, or China’s BeiDou — and it follows a preloaded route, ignoring everything else because it can’t see or think. A helicopter could sit right beside it and it wouldn’t notice; drop something in front of it, it has no idea; if the drone ahead destroys the target, it’ll plow into it anyway because it can’t look or decide. The upside is the tether is much harder to jam. The downside is a skilled electronic-warfare operator can spoof the signal and send it somewhere else entirely — in the Ukraine war, the Russians have hacked GPS signals and redirected Ukrainian memory drones into the Baltic states.
Next generation: more memory means a decision tree rather than just a path. You might still fly it manually at first, but it has a camera, and once released it looks around using a priority list — primary target anti-aircraft; if none, a power plant; if none, a tanker truck; if none, a tank — and self-assigns a target. Because the decision-making is internal with no external link, it can’t be jammed; the only way to stop it is to shoot it down. Starting in March, Ukraine began fielding these by the hundreds per day, and almost all the damage to Russia’s energy sector in the last three months has come from this type. In the last few weeks, they’ve gotten cheap and numerous enough to go after highway supply trucks carrying troops, fuel, ammo, food, and water — even better decision trees now: fly until they spot a road, then follow it until they find a truck. Everything within roughly 150 kilometers of the front is now close to a logistics-free zone for Russia. The Russians will innovate in response, of course — none of this technology is difficult, it predates 2014 — and in a world where high-end chips disappear, everyone will still have access to this tier. This is remaking what warfare is.
The Iranians have a few of the equivalent “super Shahed” drones, and most of the damage to US military facilities in the Gulf during the war came from just a couple of them.
The one that should really scare you is the Jaedan (a South Korean design). Instead of NAND, it uses DRAM — fast, “hot” memory that loses data when powered off but, as long as power holds, can process and feed the system in real time at high speed. Its decision tree isn’t six or seven steps — it’s 60,000 or 70,000. It can carry programmed pilot data that lets it maneuver like a human pilot evading threats; if a target’s already been hit, it reassesses and reassigns; it can coordinate with others; and like the rest, it’s unjammable because everything runs internally. It’s small — can only carry something the size of a grenade — and this prototype was only announced in April, but again uses no genuinely new technology. We’ll likely see these on the Ukrainian battlefield within a year, because Russia recruited North Korea to help in the war, and South Korea is furious — and South Korea also produces 90% of the world’s high-end DRAM. Now they’re working on the hardware too.
We’re in the second generation of military-affairs technology. The first was in the 1980s and ’90s — satellite communications, cruise missiles. Now the technology is cheap and common enough that people in garages are mass-producing new weapon types, and nobody’s ready for it.
One more thing — pure airframe hardware, not computing hardware. Quadcopters are highly maneuverable, you can change what they’re doing second by second, but they’re limited in range and payload, aerodynamically inefficient since rotors must run constantly, and they typically carry a small payload — something like a six-pound mortar shell, a grenade, plastic explosive, a stick of dynamite, whatever. We have footage of these flying into aircraft hangars and hitting back rooms storing ordnance they didn’t even know was there.
Or you can have a fixed-wing aircraft, like a Predator — four times the speed, five times the range, but it can’t loiter; if you have to abort, it has to circle and try again, where the quadcopter just descends. A fixed wing can carry maybe a 100-kilo bomb, so you don’t have to be quite as precise. Five days ago, the Ukrainians started hybridizing the two — putting a fixed wing on a quadcopter — so it launches, then glides. In five days, they doubled their quadcopters’ range. Quadcopters used to be limited to 20–50 kilometers; now they’re regularly hitting targets 100 miles from the front. Unlike the truck-hunters, these primarily go after personnel in bunkers, and we’re seeing the entire Russian side of the front line getting “denuded” in real time — and we’re early in the fighting season, which usually runs late May to early October. We’ll probably see significant Ukrainian gains in the near future as these technologies compound. Russian battlefield deaths — not casualties, deaths — now exceed 1,000 a day; more Russian men are being killed at the front than Russia is generating in new births.
Does this mean Russia loses? Not necessarily — they’ve spent three years building defensive works, especially minefields, and stopping reinforcements is very different from clearing minefields under fire. But for the first time in three years, that’s a real possibility. The technology isn’t sophisticated, and there’s no reason Russia — whose military-industrial strength is taking a proven innovation and mass-producing it — can’t copy it at scale. And it won’t just be Russia: Ukraine, Poland, Germany too. The face of warfare has changed.
Missile defense. The PAC-3 is our primary missile-defense interceptor, fired by Patriot batteries. We can make 700 a year, at $3–4 million apiece. An Iranian Shahed: Iran can assemble 700 in a week, at $4,000–$60,000 each. We’ve already used two-thirds of our global Patriot supply — we can’t fight another Iran war the way we fought this one.
A Ukrainian “Sting” missile is limited in range and speed, roughly half as effective as a Patriot interceptor, but costs only $2,000–$3,000, and Ukraine is already making 700 of them every two days. Poland, Germany, France, Romania, Finland, Sweden, Kuwait, the UAE, Qatar, and Saudi Arabia have all broken ground with Ukraine on facilities to mass-produce these. Notice which country isn’t on that list — because Trump doesn’t like Zelensky.
Any hardware within roughly 600 miles is now vulnerable to these technologies; Ukraine has proven it can reliably hit 800 miles, with smaller warheads. That radius covers the traditional hostile triangle of China, Iran, and Russia, plus their satellite states — Belarus, Yemen, Cambodia, Laos. Any hardware in that zone, we basically have to write off.
Our hardware problem. I’m not going to pretend to fully understand the planning, but every military conflict we’ve fought since 1988 has relied heavily on the JDAM — a “dumb bomb” with a guidance fin kit, laser-guided in. The downside: you have to get within double-digit miles, and it glides. The Trump administration didn’t want to engage Iran at that range — not necessarily a bad call, but it forced other decisions: we put one carrier in the Red Sea, not the Persian Gulf, and another off Pakistan in the Indian Ocean, relying exclusively on long-range munitions — we’ve used 40% of our Tomahawk supply, two-thirds of our standoff missiles, and two-thirds of our PAC-3 and THAAD interceptor stocks (designed to shoot down ICBMs).
If there’s another conflict this administration or the next, we cannot do that again — we don’t have the hardware. The current budget plans to quadruple production of all of these. Even if that capacity comes online at record speed, we’d only get back to where we were on February 1st sometime in 2034. The US is no longer capable of power projection in the Eastern Hemisphere. If we get into another fight, it’ll have to be at close range, and it will look very different.
Energy. Here’s the oil-flow map from the dawn of the shale revolution, when the US produced a lot of natural gas but not yet liquids — three big demand centers: North America, Europe, East Asia. After a few years of shale, US oil production more than doubled, and the US stepped back from being an active consumer in the market. Persian Gulf flows went mostly to East Asia. Right before the Iran war, the US had become a significant exporter, Europe was taking small amounts of crude from anywhere it could, and East Asia was dependent on everything else. (The blue lines are Russia’s shadow fleet.)
This year, Persian Gulf flows have dropped roughly 75% — about 13 million barrels a day offline. Suppose we got a peace deal tomorrow and Iran became, hypothetically, the 51st state — you still wouldn’t get new crude from the UAE this year. The UAE is the only country that shut in production in a way it had actually prepared for, since it actively participated in OPEC quotas, so it can turn production back on — but that takes upwards of six months. Nobody else who shut in production has that capacity. Kuwait would take a minimum of two years; Iraq, a minimum of five; and in Qatar’s case, where some facilities were destroyed, 10 to 20 years. The shortages we’re seeing aren’t short-term — and that’s before counting the Russian oil infrastructure Ukraine is likely to destroy with these new drone technologies this year.
We’re looking at roughly 18 million barrels a day offline for the rest of the year in the best case. We’ve been drawing down global government and corporate reserves for three months at the fastest rate in human history, and within the next two to five weeks we’ll drop below minimum operating levels in East Asia, with Europe shortly after. Oil prices will probably at least double, and things will get real — we’ll see genuine demand destruction as people simply can’t afford the prices. When that hits the US, Trump will do what any American president would: stop exports. That would cause real damage to a lot of producers, but it would put a lid on North American energy prices — at the cost of everyone else losing access to Persian Gulf, Russian, and American crude simultaneously. This is, in my view, the last year for globalization, because a lot of countries simply won’t be able to keep the lights on. And that’s the best case, assuming Iran effectively joins the US side. A more likely scenario is less optimistic.
Iran negotiations. Field Marshal Asim Munir is the most powerful person in Pakistan — he runs the armed forces, led the last coup, and is in practical terms the country’s leader. Ten years ago, he was working with the Pakistani and Afghan Taliban against US forces. Now he’s Trump’s point man for negotiating with Iran. Yes, that’s a problem. There is not a single American actually in the talks. JD Vance, Steve Witkoff, and Jared Kushner went to Pakistan on day one; Kushner tried to turn it into an investment pitch, and the Iranians reportedly told him if he came back, they’d shoot him. Kushner — say what you will — isn’t dumb; he hasn’t gone back.
Witkoff, I’d argue, has been the least effective of the three — he’s worked the phones since but hasn’t stepped back into the room. That leaves JD Vance. I’m not a fan of his for various reasons, but he genuinely tried — the first round of talks ran 21 hours, and he did a reasonably solid job. It didn’t solve anything, but given that the US and Iran have been negotiating various things, on and off, for 40 years with little success, the idea that this would get resolved in 20 hours was always wishful thinking. Credit for the effort, though. He hasn’t gone back either. In Trump’s own words, the negotiations have “gotten boring,” so there are no Americans currently involved — it’s all being routed through Munir, who has his own goals. Don’t expect movement until the American president changes his mind and puts an American back in the room.
Netanyahu was at the White House with Israel’s Mossad chief on February 11th, presenting the case that convinced Trump to act — the number he used was 96: this would all be over in 96 hours, the government overthrown, leadership killed, an uprising. Every defense secretary, CIA director, and joint chief for the last 45 years has warned an American president that we cannot contain the aftermath of an Iran strike — and to their credit, the current CIA director and joint chiefs gave Trump that same warning that day. He committed immediately; carriers started sailing within six hours. That’s how much planning went into this.
The carriers haven’t gotten within a thousand miles of the coast, because the Navy understands what drones can do. There shouldn’t — and likely won’t — be an amphibious landing, because a troop ship sitting still is a target. One weakness of current drone tech: unless it’s a first-person drone under real-time control, they struggle to hit moving targets, since they don’t know where the target will be. The handful of ships hit by drones in the Gulf fall into two categories: those controlled by first-person drones, or those that stayed anchored in the same spot for days and got hit by Iranian memory drones cued by Russian satellite reconnaissance. The US Navy doesn’t want to be in that position.
Which means the people we actually need to talk to are the IRGC — the Iranian Revolutionary Guard. These are the real power in Iran today; I’d say they make the SS look almost mild by comparison. There used to be two power centers in Iran: the IRGC, who control the missile program, the nuclear program, and regional paramilitaries, and the clerics — roughly 10,000 of them, the religious leadership, who were, relatively speaking, the more moderate of the two. Israel and the US largely killed off the clerical political leadership, including the supreme leader. But with 500,000 people, the IRGC simply absorbed all the leadership roles in the state. Talks with Iran were never going to be quick or easy, but the people who were capable of negotiating a political settlement are now dead — leaving only the IRGC. We’re in the most difficult negotiating environment in US–Iran history, and there’s no American even in the room.
Domestic politics and industrial policy. We need to double the size of the industrial plant, starting not from a standstill but from a step backward — we don’t have the power, the labor, the capital, or the inputs. This is hard. No company, no coalition of companies, can make this happen. Capitalism probably can’t make this happen on its own. The only way forward is fusing state coercive power with corporate America — the nice phrase for that is “public-private partnerships.” The less nice, and probably more accurate, phrase is corporate fascism.
Here’s the problem with that, beyond the obvious ones: remember the old political matrix — hippie liberals up top, spendy liberals on the left, “get out of my wallet” on the right, “control your life” at the bottom? Here’s how all the American political factions looked before Trump, and here’s where we are after the last election. In this environment, neither Republicans nor Democrats can win an election outright — there are too many self-identified independents now. The latest Pew poll, from last month, found the MAGA share of identified Republicans had nearly doubled to an absolute majority, while the overall number of people identifying as Republican had roughly halved as people left the party. We’ve got a small, ideologically distinct core controlling what’s left of the Republican brand — and something similar, though different in character, has happened on the Democratic side.
For public-private partnerships to work, you generally want four groups involved: the national-security community (as consumers), unions and businesses (economic interest), and fiscal hawks (people who can actually do the math). None of those four are in the room with this administration — or were with the last one. They’ve been out of power for six years, and Trump has personally campaigned against their champions across the country. They’re gone.
The last country to attempt large-scale public-private partnerships without those four groups involved was Argentina in the 1940s, under Juan Perón — whose policies spawned “Peronism.” Argentina went from being a top-ten country in per-capita income, a position it had held for 60 years, to what it is today. I’m not saying that outcome is unavoidable for us — I’m saying it’s the path we’re currently on.
Cabinet roundup. Lori Chavez-DeRemer, the Labor Secretary — does anyone even remember her? She’s basically a smaller-scale version of [Kristi] Noem: similar scandals, hiring a romantic partner, public corruption allegations, theft — minus the more theatrical drama. Pam Bondi — I feel for her. Part of the problem governments are running into is that so many MAGA-era claims are based on falsehoods, like the 2020-election narrative, and she was tasked with turning those into actual lawsuits. Of course she failed — and then was fired for trying to do exactly what her boss asked, which was, if not illegal and unconstitutional, simply impossible.
Tulsi Gabbard — if she wasn’t compromised by Russia, she was certainly interviewing for the role. I could give you any number of examples of doctored intelligence reports that left Trump under-informed, but the clearest is the Alaska summit that collapsed after 90 minutes. The day before the summit, Gabbard called in sick with the flu, and for the first time, career CIA officers were able to brief Trump directly. The CIA is, unsurprisingly, good at reading other people’s communications — they showed Trump Putin’s actual talking points for the summit: how he’d open, what photos he’d use, the order he’d present things in, and where he was counting on Trump being uninformed enough to be steered. So when Trump arrived with Putin’s own playbook in hand, facing Putin running that exact playbook, he called off the summit after 90 minutes. Good for him. Gabbard fired the relevant officers the next day — and she’s done that across the intelligence community. She’s now gone, and the nominee to replace her is Bill Pulte — yes, of Pulte Homes. He currently oversees Fannie Mae and Freddie Mac, and reportedly used that position to enrich himself, while more notably using it to generate accusations of mortgage fraud against Trump’s political opponents. He’s now been nominated to follow Gabbard, and there’s pushback in the Senate, even among Republicans, over the idea of using intelligence functions to manufacture cases against political opponents — very Peronist, I’d add.
A few more to know, not necessarily because they’re about to be fired, but because of where they stand with Trump: RFK Jr. — Trump’s frustration with him isn’t his incompetence, or that the American death rate has doubled under his watch, but that he’s so popular with the anti-vaccine crowd that Trump feels eclipsed.
Pete Hegseth — in his own words, the least qualified defense secretary we’ve had, and he’s found new heights of dysfunction in office. Fifteen months in, he still has no chief of staff or secretary, reportedly out of fear they’d undermine him — which is almost comic, since he doesn’t even know his own staff. His daily routine, by most accounts: walk in, do a video, work out with the troops. The reason he might get fired isn’t incompetence — it’s that Trump is looking for someone to take the fall for the Iran war, and as Defense Secretary, he’s a logical scapegoat. General Dan Caine, the Joint Chiefs chairman, actually knows what he’s doing — and because Hegseth is so dysfunctional that he can’t even schedule a Pentagon meeting, Caine has had to effectively live at the White House, spending more time there over the last three months than any Joint Chiefs chairman since World War II, just to make sure that orders coming from the top don’t get people needlessly killed.
Kash Patel — less competent than corrupt-adjacent, more of a conspiracy theorist running the FBI. He’s at risk of being fired for the same reason as Bondi: he’s been unsuccessful at manufacturing cases against Trump’s opponents. Has nothing to do with whether he can actually do the job.
And the most genuinely qualified of this group, Howard Lutnick — actually well-suited to be Commerce Secretary. Unfortunately, before taking the role, his unofficial reputation in New York was “the most corrupt person in New York,” which, candidly, is part of why Trump likes him. The issue that might get him dismissed is that he vacationed with Epstein. Fun fact: before he became Commerce Secretary, one of his companies once hired me for consulting work — I always vet clients before taking a job — and I discovered, early in the administration as tariff policy was being formed, that his children were effectively selling access slots at the Commerce Department so people could pay to get tariff exemptions. Lovely family. I fired myself from that engagement.
The “YOLOs.” Short for “you only live once” — these are Republicans Trump has aggressively campaigned against; all but one are senators (Massie is a representative from Kentucky — he’s included because Trump dislikes him more than almost anyone). They’ve clashed with Trump over Pulte, over slush-fund issues, and other matters — but the key point is seven of them are senators, which is enough to block any new, incompetent, or corrupt nominee from being confirmed. They’ve used that leverage 37 times in the last three months.
That creates an odd dynamic: Trump can’t fill vacancies he creates by firing people, because there aren’t enough qualified people within the MAGA coalition, and the loyal-enough people he can find aren’t qualified — and this group keeps blocking them. Two of these senators have already lost primaries and will be gone in January; two more aren’t running again when their terms expire (assuming Grassley of Iowa, who’s even older than Trump or Biden, makes it that far). But three of them have terms that don’t expire until after Trump’s term ends. So two things follow: this administration is about as capable as it’s going to get — it can’t bring in fresh talent, partly because of the YOLOs, and partly because when Trump took over the party out of power, he ended normal recruitment of candidates, staffers, and policy people. What’s in the White House now is what we’re going to get. Second, this is the YOLOs’ last stand — whatever you think of them, they’ve shown their hand, and Trump is working to politically destroy them, with some success already. These are the last “old guard” Republicans left in Congress, and in three years they’ll all be gone.
It’s probably more accurate to think of Trump the way you’d think of Xi Jinping in China — not really a “Republican” in the old sense, any more than Xi is really a “communist.” He took over the institution and remade it to serve his own needs, and it’s become something else. The sooner we accept that, the sooner we can move to whatever comes next. The political system has already evolved — we’re at one of those moments where the floor and ceiling are both gone. That’s frightening in a lot of ways, but it’s also freeing. We’ve reinvented the party system five times in our history; this is the sixth. We’ll get through it. But it’s time to start thinking about what we want it to look like, because the old Republican and Democratic parties are both gone, and they’re not coming back — and this current moment won’t survive the current president, because he has no real successor and won’t have one.
I don’t know if that makes anyone smile, frown, scream, or cry — but that’s the moment we’re in. What we do with it is up to us.
That’s it for me — the QR code takes you to everything that matters, and this is the book Vix so valiantly “died” in. I’ll take questions.
Q&A
Q: We haven’t heard much about the Keystone pipeline lately. Why isn’t anyone talking about it?
Keystone runs from the Albertan oil sands through the Great Plains and southern Midwest into Texas. It was first proposed ages ago, and part of it was built before the Biden administration killed the permit and stopped construction. The Trump administration has since restarted what I’d call “Keystone 2,” and construction is underway — I’d be shocked if it isn’t finished at this point. Once done, it’ll bring roughly 1.5 million barrels a day of Albertan heavy crude to the Gulf Coast for processing; most American refineries actually prefer Albertan crude to the light, sweet crude from shale. But it won’t be finished before this summer’s energy crunch — best case, near the end of next year. So American refiners, geared toward heavy crude like what comes from Venezuela and Canada, are going to be one of the two groups hit hardest by the energy crisis in the short run, since neither country can get crude here fast enough. Best case for Keystone is full capacity by the end of 2028; best case for Venezuelan volumes reaching 2 million barrels a day is 2038. It just takes that long — but I’m fairly confident it’ll get built.
Q: You said there’s no logical successor to Trump after 2028. Why wouldn’t someone like JD Vance win?
Two reasons. First, Vance was a junior senator, and in Trump’s eyes — which are really the only ones that matter for succession — he hasn’t performed the way Trump wants. That may not be a fair read, but the president has exactly as much authority as he chooses to delegate, no more, and Trump is inconsistent about who he trusts. Second, and more importantly, Vance doesn’t have a following independent of Trump — so if he becomes the successor, it’ll be because Trump personally hands it to him. Getting and keeping that blessing requires an enormous amount of deference through the primaries and the general election — and for independents and old-school Republicans, that kind of visible deference is disqualifying. Whoever wins Trump’s blessing will likely lose credibility with most voters in the process of earning it. So the only paths I see for someone like Vance: Trump dies at a convenient moment (he’s quite old, so it’s possible), or the Democrats nominate someone like Bernie Sanders, leaving voters with another deeply unappealing binary choice. That second path seems more likely.
Q: There are a couple of potential trillion-dollar IPOs coming — SpaceX, Anthropic, OpenAI. Thoughts, especially given the chip presentation?
SpaceX first — I’m a big fan. Cutting launch costs by 97% is spectacular, and that’s just phase one; it opens up possibilities we couldn’t previously consider. Space manufacturing becomes viable, and it doesn’t require the high-end chips so many other things need, so there’s interesting potential in biomedical and construction applications. A lot of that is still prospective — until there’s a platform for actually manufacturing in space, it remains a long-term play. I’m not telling anyone to buy or not buy SpaceX, that’s not my domain — but I am genuinely excited about it. For what it’s worth, I’m no fan of Elon Musk for various reasons, but SpaceX itself is brilliant.
Anthropic — by far the best platform right now, the strongest safety guardrails, and clearly the most trusted by the Defense Department, even amid friction with Hegseth and Trump. But it’s constrained by hardware. I don’t doubt the IPO will be a success — but we’re heading toward a point where chip supply simply stops growing, and the entire LLM ecosystem has to reorganize around hardware that doesn’t exist yet. Even if we designed the ideal chip today — one with cooling built into the hardware itself — meaningful adoption wouldn’t happen before 2031; the manufacturing pipeline just takes that long. The lithography you’d use for that kind of chip (deep ultraviolet, DUV) is more manual than EUV, so fabs take months to retool rather than days.
Q: A lot of the news cycle focuses on oil prices and the economic fallout. But steel prices are up 50–60% since November, and there’s a supply shortage that isn’t getting much attention. Any thoughts on where that’s heading, and the implications?
It’s multifaceted. The expected driver: tariffs of around 50% are pushing up steel prices, and good luck building anything without steel — same story with lumber and copper. There are roughly 8,000 grades of steel, and the US essentially dominates the top third — specialty steels nobody else can make — but that’s not what you use for construction; that’s rebar and hot-rolled steel. Three problems there: almost all of it is imported, because domestic low-grade production isn’t competitive; the two largest suppliers into the US market until recently were Ukraine and Russia, so global supply has tightened; and essentially every country is now bracing for deglobalization, which means a construction boom everywhere at once. Don’t expect this to ease in our lifetimes. The one variable that could change it is tariff policy — but Congress specifically granted the executive branch authority over that category of tariff, so unless the president changes his mind, it’s not going away. It’s different from, say, the recently imposed “child labor” tariffs on 60 countries that happen to be our 60 largest trading partners — those won’t hold and will likely disappear within months. Steel tariffs are real, and they’re going to last.
Q: Does the US blockade choke Iran into submission? Does quantum computing offer any hope for the chip crisis? And any early predictions?
Quantum first: until we can stabilize a quantum reaction for more than a tiny fraction of a second, no. Quantum computing, in essence, means reading and encoding information into the subatomic particles of an atom — in theory, near-infinite processing and storage capacity. Right now there are roughly 130 quantum computers in the world, each one handcrafted piece by piece. For it to be part of the solution, the reaction would need to hold stable for minutes, and we’d then need to build an entire manufacturing system around it. So no, even assuming it’s ultimately possible, it won’t arrive in time to help with this current cycle.
On Iran: the blockade is the lever you need if you want to bring the IRGC to the table, because the IRGC controls oil exports — that’s their income. Until the blockade, nothing about the war had actually bothered the IRGC; from their perspective, Trump and Netanyahu had removed all their domestic political rivals, a pure win. With a ceasefire in place for three months now, they’ve rebuilt rail links to China and actually have more drones than they did at the start of the war. Militarily and politically, they’re satisfied with how this has gone. Economically, they’re getting nervous, since oil exports are roughly 80% of their income. The theory behind the blockade is that if they can’t export, storage fills up, pressure backs up through the system to the wellheads, and they’re forced to shut in production — no income now, and none later. The IRGC, having anticipated this for decades, is running refineries at maximum capacity and exporting the refined fuel they don’t use, mainly to Pakistan and Iraq, while reducing pipeline and well pressure to the bare minimum needed to avoid reservoir damage. So it’s a race: how long can that strategy hold versus how long the administration tolerates the blockade and its consequences. We don’t have a clean answer, because Iran doesn’t report its production data to international bodies. If they’d run at full output, they’d have hit the breaking point within two weeks of the blockade starting — we’re well past that point now. I don’t have a confident prediction, but this kind of pressure seems to be the only real lever available if the goal is to force IRGC behavior change — and the fact that it’s been months without visible effect is a bit concerning.
Q: Given China’s progress in robotics, industrial automation, and control of critical minerals, could China become the first great power to build an artificial labor force that offsets its demographic decline from the supply side? [Multi-part follow-up on critical minerals, robotics value-add, and export-led demand.]
Starting with critical minerals: roughly 70% of global materials processing happens in China — more for some minerals like gallium and rare earths, less for others like lithium, but about 70% on average. Any industrial policy that’s going to succeed has to prepare for China’s exit from that role. These are things we need to relocate — it’s not expensive or technically difficult, these are century-old technologies — but we haven’t actually started building it out yet. It belongs on the to-do list, and it’s the kind of thing industrial policy should be able to do without tipping into outright corporate fascism.
On the robotic workforce: China has made real advances in robotics, but Chinese and American robotics are functionally different things. The value-add of a Chinese industrial worker is roughly one-fifteenth that of an American worker, so when an American works alongside a robot, it’s a very different productivity story than when a Chinese worker does. China has several hundred thousand more robots than the US, but collectively they’re probably only about a fifteenth as productive as the American robot fleet — not an insult to China, just a different starting point.
If China succeeds at replacing labor with automation, that addresses the production side of its demographic problem, but not the consumption or tax-base side. In that scenario, China becomes entirely dependent on American consumers buying robot-made goods — and if that demand doesn’t materialize, the entire economic model collapses overnight. So yes, it’s significant, and yes, I watch it, because automation is a real potential substitute for lost labor — but if it’s your entire strategy, you’re setting yourself up to fail.
On the export-demand follow-up: maintaining demand through dominant high-value exports requires unrestricted international market access, and China’s strategy for thirty years has been to manufacture almost anything, even when it isn’t the most efficient global producer — which has hollowed out manufacturing in country after country. That strategy depends entirely on a free, secure global trading environment, and that’s exactly what’s disappearing right now. So no, I don’t think that path holds.
Q: Is achieving a new American industrial revolution at this scale and timeline realistic without fundamental changes to the current campaign-finance-driven two-party electoral system?
I’d set elections aside — the election cycle is too slow to meaningfully affect this timeline. Best case, I think we’re at least six years from a new stable two-party alignment — through this midterm and the next general election, and probably settling out somewhere after that. It could take longer; there’s no obviously charismatic figure on either side positioned to drive that change quickly, which means it falls to ordinary people showing up to unglamorous organizing meetings to build it from scratch — something we’ve done three times before in our history, and it’s genuinely tedious, but nothing changes until people start.
On the industrial buildout itself: the presentations I gave two years ago, for those of you who were here, were very different — much more about sector-by-sector mechanics and what needed to happen where. We’ve run out of that runway. Back then I argued we had five to eight years; now it’s probably under two. We’re not going to pull off the full buildout — we’re in mitigation mode now. Part of that is the inherent nature of the problem, part of it is China’s faster-than-expected collapse, part of it is domestic regulatory friction, part of it is the Iran war and the disruption to the energy sector, and part of it is Ukraine’s targeting of Russia’s energy sector. A lot of forces are converging to break the system in the near term rather than the long term. I used to say “inevitable doesn’t mean imminent.” I’m not saying that anymore — we’re almost there.
Q: On the deficit-spending slide — US debt was high, but China’s was off the charts. Can you comment on the American deficit and where it’s headed, and the Chinese comparison?
There are different ways to measure debt. For China, the figure to watch is corporate debt, since most large Chinese companies are state-owned — total corporate debt runs around 400–450% of GDP. Personal debt adds roughly another 100% of GDP, similar to ours. Federal debt is only about 40% of GDP, less than half of ours, but local government debt is over 100% of GDP, more than double ours. It’s not an apples-to-apples comparison, though, because the US dollar is the global reserve currency and the yuan is largely locked within mainland China — so China has a political option we don’t: they can essentially wipe the slate and start over. Japan has done exactly that eleven times in its history, and it can work if you’re willing to accept significant casualties along the way, which China generally is.
On our side: our budget deficit, outside the COVID-era special case, is now the highest it’s been since World War II. Essentially, Trump has continued most of what Biden chose to spend on, while the people who would have trimmed the budget were the first ones fired.
Q: In a deglobalized world, what’s the risk of a World War III scenario — China making a move on Taiwan, Russia unable to function except through war? What restrains that?
The risk in answering this is that we’ve moved past the point where long-term trends predict outcomes, into the daily churn of what actually happens day to day. Demographically, this is “the decade” — but it might also just be “the year,” and decisions made locally and regionally now matter an order of magnitude more.
On China and Taiwan: under more normal circumstances, I’d say China effectively gave up on Taiwan years ago — roughly nine years back, Xi’s last independent senior adviser (before being purged) told him directly that invading Taiwan would trigger a dozen countries, without even engaging their militaries, that could collectively wreck China’s economy and cause famine and energy shortages severe enough to kill a significant share of the population within a year. Those countries include Australia, Singapore, Saudi Arabia, India, the US, and Japan. Since then, Xi has had little to do but produce ideological writing and propaganda — reportedly some 300,000 pages of it. We don’t have great insight into what he’s actually thinking, because by all accounts he doesn’t discuss it with anyone close to him. But if you assume his published military planning reflects intent, China could likely take Taiwan within two weeks militarily. The problem is what comes after — China still wouldn’t control global consumer markets, energy, or food, and needs unrestricted access to survive. Australia alone could meaningfully damage China’s economy within a week if it chose to.
Russia is a different story. Russia genuinely believes that without re-expanding toward something like the old Soviet borders, its declining demographics make it unable to defend itself long-term — and they’re probably right, strategically and demographically. But Russia started this war with roughly 20,000 tanks (by Russia’s broader definition, including APCs and mobile artillery) and now has about 7,500 — having burned through well over half its supply without even conquering Ukraine. I’m not ready to call the war’s outcome, but it’s increasingly hard to see Russia conquering not just the rest of Ukraine, but also Finland, Estonia, Latvia, Lithuania, Belarus, Moldova, Poland, Romania, Georgia, Azerbaijan, Armenia, Kazakhstan, and Uzbekistan — territories some Russian planners have discussed. I’m close to saying Russia simply can’t do that anymore, which means they’ve expended nearly everything on a war they may not win, accelerating their own collapse — and that puts the nuclear question back on the table.
We’re at a moment where today’s leadership decisions will shape the rest of the century — it feels a bit like September 1945 again. Imagine if the US hadn’t built the globalized system back then. We’re at one of those inflection points, and right now there’s really only one person whose decisions matter most.
Q: What’s the outlook for the dollar over the next four to six years?
There’s no real competitor, in short. Regardless of US debt or Federal Reserve policy, there’s nowhere else for the world’s reserve currency role to go. As globalization fractures, the yen becomes increasingly insular — it’s already a fairly closed currency, and Japan trades less as a share of GDP than the US does at this point; they’ve largely turned inward. The euro’s days as an expansionary international currency are over — it’s still the world’s number-two trade currency, but at meaningfully lower volumes than a decade ago, and usage outside the Eurozone is limited. I don’t think the euro disappears outright, but the larger risk is what happens to the European Union itself as globalization fails, since Europe has a milder version of China’s demographic problem — an aging population with fewer domestic consumers, leaving exports as the main lever. If the EU can’t secure a stable trade arrangement with the US, that model runs into trouble quickly, and it’s hard to see the euro surviving the EU’s potential unraveling. People sometimes mention the Canadian dollar as an alternative, but not at a scale that would move the needle. I don’t see another real destination for global capital.
Q: Following up — what’s your take on Lutnick’s plan to route national debt through a stablecoin/cryptocurrency vehicle, and how that might affect the dollar?
It’s possibly the single most corrupt mechanism I’ve seen proposed for this purpose — almost an achievement in its own right. There’s no version of crypto that functions independently of the US dollar system. If you can convince people to park funds in a vehicle the US government effectively controls through one specific individual, more power to him — he could make himself extraordinarily wealthy doing it, and since the space is largely unregulated, it likely wouldn’t even technically be illegal. But I can’t see serious institutional investors going for it; there’s no real advantage to them. That’s one of the broader problems with crypto generally — it can benefit the holder, but typically nobody else, and only for as long as the US government allows full convertibility, which, as we’ve seen under this administration, isn’t guaranteed. One of the bigger blows to the IRGC, in fact, came when Treasury identified some of their crypto wallets and simply seized the funds. It’s actually more vulnerable to state intervention than hard currency.
[Closing] Peter, thank you so much.
