伊朗战争 + 五角大楼与 Anthropic 的冲突：战争部副部长 Emil Michael

All right, everybody. Emergency podcast time. Episode 263 of All-In. We have Emil Michael, the Under Secretary of War for Research and Engineering, working directly for Pete Hegseth. We had to get this out to you on Thursday night because it is an emergency pod. One of my old besties, Emil Michael, is here. Emil and I were part of Team Uber back in the day. He was Travis's right-hand man. Some might say fixer. And Emil Michael is now the Under Secretary of War here in the United States, serving his country like our bestie David Sacks. Welcome to the program for the first time, Emil Michael. How are you doing, brother?

I'm doing good. I hope it was more than a fixer, but, you know—

A $20 billion fixer. I mean, you got it done. You got it done. He would give you the hardest things.

That's right. Fair enough.

If it was hard, that's what a fixer is.

An operational ace. That's what they call them.

All right. In Brooklyn, we call them fixers.

A rainmaker.

There's that, too. There's that, too. Making it happen.

With us again, Chamath Palihapitiya. How are you, brother?

Great.

Yeah. Look at that smile. What do you got going on? You got some pokers in the fire.

Okay. I'm not going to say. In the coming weeks, I think some news is going to drop. That's my prediction. I don't have any information.

Are you loving the tweet-mogging that's been going on this week? So good. So good.

Oh, my God.

So good. He's looksmaxing by default, but he's been mogging—

The gooners.

Yeah. So funny. What was your favorite?

The one I sent you this morning that you said—what did you say?—“So funny.”

“Are you collecting your losses? Tax harvesting.”

What did you say? “Oh, my God.”

Chamath said, “Oh, my God.” It was just like, “Yes, I did. Yes, I did.”

“Yes, I did.” Yes. Someone said something to Chamath. He was like, “Dropped in.” “Why is everyone so mad at Chamath? All he did was lose billions in retail investors' money by proposing one-page SPACs. It's not like he then told them to enjoy their capital losses or anything. Give the man a break.” Chamath's response: “Yes, I did.”

All right. Piling on is your sultan of science, everybody's favorite. He brought some great science to the show last week. David Friedberg, how are you doing?

Oh, yeah. I've been traveling this week. Back at home.

All right. Sacks is out today. He's very busy on Capitol Hill. We'll talk about what he's up to next week.

Let's go. Come on. Let's go. Let's go.

Let's go. Go, Jason.

The U.S. and Israel launched a joint attack on Iran on Saturday. Today is Day 6 of Operation Epic Fury. Iran's Supreme Leader, Ali Khamenei, was killed within hours of the operation. Forty senior officials have also been killed, and the death toll so far is about 1,000 people, according to reports. Tragically, 6 U.S. Army Reserve soldiers were killed following a drone strike on a base in Kuwait. A U.S. submarine sank an Iranian ship off the coast of Sri Lanka. This is the first torpedo kill since World War II.

Why we are at war has been a bit of a moving target in the debate. The first explanation from Rubio was that Israel was going to attack and the U.S. had no choice but to participate. He later walked that back. Trump made it clear this is not a regime-change effort, but we're doing this to stop terrorism and the development of ICBMs by an obviously pretty crazy group of individuals, and nuclear bombs, which we blew up a couple of weeks ago. Trump also mentioned that the people of Iran should, quote, “seize the moment” and “take their country back.”

Pete Hegseth, your boss, Emil, said, quote, “This is not a so-called regime-change war, but the regime sure did change, and the world is better off for it.” Here's an interesting Polymarket: “U.S. forces enter Iran”—boots on the ground—by the end of March, a 40% chance; by the end of the year, a 59% chance. The sharps on Polymarket believe we will have boots on the ground. Will the Iranian regime fall by June 30th? A 39% chance, according to Polymarket, and by the end of the year, a 51% chance.

So, Emil, I guess there are 2 questions people really want to know. I'll leave off why we're doing this—I think President Trump has been pretty clear now—but how long is this going to take? And are we going to have to have boots on the ground? What is success here?

I think the president talked about this as a weeks-not-months kind of operation, and it's aimed at essentially disarming the regime, or the country, in such a way that they can't supply Hezbollah, Hamas, the Houthis—all the kinds of terror groups that get sponsored with weapons and money from Iran—not to mention the nuclear bit. That's why, from the reporting, you see that they're going after the depots. We went after the nuclear sites before. They're a prodigious drone maker, with these huge one-way attack drones that can go hundreds and hundreds of miles. They also have lots of ballistic missiles aimed at every country in the Middle East, as you've seen them attack.

That's one. In terms of boots on the ground, there's no scenario where we have some protracted Afghanistan- or Iraq-2-like situation.

Friedberg, your thoughts on this war? Obviously, a lot of people voted for Trump in order to have the peace dividend. He was, in his first term, absolutely the peace president. And now here we are: 8 countries have been bombed, and we've had 2 leaders deposed, 1 of whom has been killed. Your thoughts, Friedberg?

I think the president and the administration have probably the biggest meeting of the term coming up with China in April. My estimation, based on the conversations and the comments made by the president before he came into office and since he's been in office, is that finding a grand bargain or a deal with China is probably one of his top priorities.

If you think about the importance of that, is the U.S. going to wade into a giant global conflict led by a U.S.-China rift, or is the U.S. going to find some grand bargain? I think he would probably have a preference for the grand bargain. And, that being the case, I think you could look at the actions in the context of Maduro and Iran as creating maximal leverage going into those negotiations.

The reason for that, Friedberg?

Ninety percent of the oil that comes out of Iran goes to China, and there's been a long-developing relationship between Maduro's government and China. These are big economic drivers supporting economic growth in China. So, creating leverage by having significant influence, damage, or destruction to those supply chains for China gives the United States the footing to negotiate a better deal for America.

I would imagine that the president's intention here isn't to go and decide who should be in charge, drive regime change, and end up in a multiyear conflict with Iran. But ultimately, if there's some transaction with China that gets everyone out of this and puts the U.S. on a strong footing, where American businesses can sell into China—which is very challenging, as everyone knows today—and there's regulatory, economic, and trade parity between the U.S. and China, as well as a point of view on what happens with Taiwan and the availability of key technologies like semiconductors, I think it could be a win-win.

I think that a deal with China could be the crowning achievement of this administration, particularly going into the midterms. The timing is right, and I think that's probably a core part of the motivation here.

Chamath, your thoughts on this action and why we're doing it? You've heard, obviously, the president's position: We're not doing regime change. It's a secondary effect, obviously, but we want to stop those ICBMs and nuclear bombs from being developed, and we want to stop terrorism. Additionally, Friedberg says, “Hey, we're framing this great discussion we're going to have with China, and oil is part of that.” Where do you think you stand on all this?

I'll build on both what Emil said and what Friedberg said. I don't think this is about regime change, and I don't think it's about a local or regional conflict. I think if you take a step back and zoom out, the most important thing that we did in the last 3 months was, by taking out Maduro and by taking out the Iranian leadership, we created enormous leverage, as Friedberg said, with China.

Now, why is that important? Because I think all of this centers around that geopolitical discussion. Last night, something important happened: The official Chinese bureaucracy posted what its GDP targets were, and it was shocking to anybody reading it because what we saw was that they guided to a range of 4.5% to 5%, which, if you look at the historical context of that growth, is the lowest that it has ever been in about 30 years—3 decades. So, this is before they entered the WTO.

The question that one should ask yourself is: When a country that's growing at 8%, 9%, and 10% starts to grow at half that rate, yet has double the number of people and double the GDP, what happens? You already have incredibly high domestic unemployment, especially youth unemployment. Does it become more or less chaotic? I think the historical record of every other country would show that it will become more chaotic.

If you have that as a starting point, what is it in China's best interest to do? I think it becomes obvious that the right thing to do would be to invade Taiwan. Why? Because you start to create a sinkhole that occupies your people, occupies resources, can get domestic production up and running, and can start to generate a war machine. And you see the economic impact of war machines in any country during any conflict.

And if I had to guess, just to build on what Emil said, the president saw that, and I think what they did can be summarized in this chart, which I sent to Nick.

So, if your goal is to prevent war with China—a massive global conflict that could be nuclear and cataclysmic—how would you do it? This chart paints one way to do it. If you look at the conditions inside the Chinese economy, the most interesting takeaway is that they are enormously dependent on imported oil. It’s about 20% of their economy, but it’s not 20% of their economy because it’s 100% of these critical things that create GDP: logistics, transportation, aviation, and feedstock inputs.

And of that 19%, about a fifth of it comes exclusively from Iran and Venezuela. Now, all of that is off the table. So, if you take that and then you see what Steve Witkoff, Jared Kushner, and Josh Gruenbaum have been doing, which is trying to get a deal done in Russia, and you put all of these things together—because, by the way, if you add Russia into that mix, it’s about 40% of China’s oil—not only do you re-dollarize, not only do you stop the funneling of all of these illicit oil funds into creating chaos all around the world, but you hem in China going into a massive moment at the end of March, beginning of April, where, as Friedberg said really astutely, there is the potential for a grand bargain. I think that secures global safety, and that is a huge thing for America.

Emil, how much does this have to do with China? I think you know my instinct is—and I’m not speaking for the administration on this—that that’s a second-order benefit to some of these things. You said 8 conflicts. There have not been 8 conflicts. We inherited Gaza; we inherited Russia-Ukraine. Venezuela was its own operation, and then you could sort of attach to it the drug boats that were coming out of that as one big operation.

The Houthis—Biden was ignoring the Houthis. They were just shooting at our ships. That was very limited in terms of, “Stop shooting our ships. We need freedom of the seas.” That’s something any president should be doing generally. I think Iran is the one material conflict outside of Venezuela, so it’s not that many. How long did Venezuela last? It was one raid. One.

A few hours. I guess that’s a really important note, Emil. This is an important note, I think, for you to explain to us: There’s a new approach here with regard to these actions. It’s no boots on the ground, and we seem to—and you, of course, have better information than anybody else does—I don’t think anybody would have known Venezuela would have gone as well as it did so far. Listen, we’ve got a long way to go with Iran, but this has gone very well as well. Explain to us what you know, and what you, the president, and Hegseth know that we don’t, that makes these two operations go so smoothly. What is it? There’s obviously some new technology here in the case of what happened in Venezuela.

Yeah. Besides the discombobulator, what we’ve got is a very well-trained military. The Global War on Terror was a disaster in so many respects, but the people who are fighting that are generals now, so they’ve learned a lot of lessons.

You compare that to the Chinese military: They don’t have a lot of experience. In fact, the decapitation they did in the Chinese military—the one guy they took out was the one guy who had experience in Vietnam. They don’t have conflict experience, and that matters because you understand going in what the things are that could go wrong.

Then you have incredible technology: space, air, land, sea, cyber, all kinds of effects that you can bring together. Imagine a hundred guys going into the most fortified compound in Venezuela, where the president is, taking him and his wife out safely, and getting out with no KIAs.

Incredible. I mean, it’s just incredible. So stunning.

Yeah. These war games have been on the shelf for a long time. Every scenario has been planned years ahead of time. Midnight Hammer in Iran was planned years ahead of time in terms of how you would do it if you were going to do it.

And then you keep refreshing the tactics, techniques, and procedures, and you’re updating them. We have a very sophisticated way of doing these things to minimize loss of life and maximize success.

Can I ask a question? I don’t want to derail this conversation, but is the discombobulator real? What can you say about the discombobulator?

I was obsessed with this when I saw it on X. I was like, “What is this thing?” I need it in my house. I just want to push a button on all of my kids. And that’s just for when the Houthis show up. Oh my God.

It’s not meant for your kids or if they’re behaving badly. No, I can’t talk about it.

Emil, do you think we would have been able to pull off that mission as successfully 5 years ago or 10 years ago? Has the technology improved that quickly, that this is not something that’s been possible historically? And how does that change the pacing and the face of war for the next couple of years?

I’d say no. It wasn’t only a technology maturation from 5 years ago. It was also the maturation of the rules of engagement.

The rules of engagement that we used to have—if you read about them, some of them were insane. In Afghanistan, if the guy had a small gun, you had to have a small gun. There was this parity in weird ways. When you’re like, “Is the objective to have a fair fight or an unfair fight?” if you’re on our side, you want it to be unfair. So, the rules of engagement were relaxed.

Who writes those crazy policies? Who sits in an office and says you can’t shoot back if a combatant is shooting at you if you aren’t matched gun for gun?

Yeah. Those crazy policies are written into military departments. That’s why, when Secretary Hegseth talks about this kind of thing and what was happening with him when he was in Afghanistan—if you ever read his book about Iraq—he says the rules of engagement were so punishing that we were at risk all the time, because you had to have a legal understanding of what was happening every minute on the battlefield.

That’s as opposed to, “Your job is to take out these guys and protect these guys. Here are your munitions. Here are the red lines. In the middle of that, use your judgment. Your commanding officer, use your judgment on how to win.” We’ve kind of gotten back to that: Use your judgment, push responsibility into the field, still have your red lines, but other than that, the objective is the objective.

It’s more of a Colin Powell approach: Go all in, have a clear objective, come out, use overwhelming force. We were not doing that for the last 4 years.

Going back to the face of war going forward, my understanding is that there have been more drones deployed by the United States this past week than we’ve done in the history of military activity. Is that right? How does that really change things going forward here?

It changes it big. The Predator drone was the first big drone program 10 or 15 years ago. It was this big, honking drone. If you remember, Obama would take out some of these Al-Qaeda leaders with drones on their balconies and things like that. I think President Trump took out Soleimani with a drone near his car. That was the beginning.

Then the Russia-Ukraine war happened, where it was drone on drone. Seventy percent of the casualties are from drones. Drone-on-drone warfare, robot-on-robot warfare—those things are the future for sure.

Right, and that’s why companies like Anduril exist: They’re making unmanned systems. This has been something you’ve specifically been very focused on, and you tweeted today a little bit about a competition. We’ll play a little video here. This LUCAS—Low-Cost Unmanned Combat Attack System—used to take a lot of time. It certainly wasn’t startup time to get a new product into the channel for our military to use.

Explain what program you’re running here. It feels like the DARPA self-driving challenge all over again. What do these drones cost? I know there’s a company making them for, I think, $35,000. Am I correct?

The small drones, like the one I’m holding right there, are way cheaper than that. The LUCAS one-way attack drone, which can go 500, 600, or 700 miles at the speed of an airplane and carry a big warhead, costs about $50,000 to $80,000, depending on what kind of equipment you put on it.

We have a Drone Dominance program, and really, we basically have to build an arsenal for drones. Now, are we likely to have a territorial conflict like Russia-Ukraine with Canada and Mexico? No. But we do want to take out drug drones at the border, and long one-way attack drones are important for any kind of major conflict like you’re seeing in Iran.

They’re also important to protect military bases, for America 250, the World Cup, and the Olympics in 2028. There are a lot more uses of drones for surveillance, not just for combat.

There, you’re showing drones that are sort of human-operated, but how much of this should basically be AI, so that it’s just some computer vision and, again, back to what you said before, a model understands the rules and the red lines but otherwise is on task and accomplishes your mission?

How much of it is one versus the other?

I believe that a sophisticated drone war is going to be drone swarms controlled by AI to some degree or another, right? To what degree the control matters? For example, drones have decoys. They could spit out things, they could dazzle, and they could put out other things. So, how do you discriminate what’s a drone and how to hit it?

You could use AI for that because it’s learned how to do automatic target recognition, for example. Then, could it identify a person, and does that make it safer? It’s going after someone you actually want to get and not someone you don’t want to get. So, there are a lot of uses for AI at the edge, if you will.

In the future, the Ukrainians and Russians do something called a kill box. When they lose communications because the drone is jammed, it just starts going in a box and looking for the person they’re trying to get. They’re starting to use AI to do that.

Wow.

China already has this ability, probably by some magnitude. They have drone swarms because they can force the companies that make them—not just DJI—to interoperate.

Interoperating drones is called heterogeneous autonomy, right? You take different kinds of drones, figure out how they communicate with one another, and then make sure they’re not going after the same target. That’s a pretty complex thing that they’re definitely working on.

Let’s talk about the fidelity of these. Obviously, AI is a new technology, and it can make mistakes. Anybody who uses it on a day-to-day basis might experience a hallucination. How confident are you in the AI in the Ukraine-Russia conflict? They obviously are not going to be as thoughtful as we are in putting this together. They’re in a hot war right now, but we, as the United States, have to be very thoughtful about this.

How confident are you that this isn’t going to make a mistake? I think that’s the key to a lot of this debate. When will it be perfect, defined as much better? I guess this dovetails with the self-driving thoughts: It has to be a magnitude better than a human. So, when will this be a magnitude more accurate than when we make a mistake as a military and kill a civilian?

Yeah, no, it’s a good question. I don’t know when that moment hits—that FSD moment where it gets better. It’s certainly not there, and you wouldn’t want to take huge risks with that. There’s a gradation of when you would use it and what kind of risk you were trying to take or not.

If you were trying to take out a drone using AI and a laser or something, you’d be pretty okay with making mistakes because you just missed the drone. With a laser, the laser goes off and it’s all over. If you were doing something more sophisticated in a densely populated area, you’d take less risk.

We’re developing procedures and tactics for each scenario. This is part of the debate I had with Anthropic, which is that we need AI for things like Golden Dome.

A Chinese hypersonic missile comes up. You’ve got 90 seconds before it separates and all kinds of decoys appear, and you don’t know where the actual payload is. You want to hit it from space. A human doesn’t have the reaction time and may not be able to discriminate with their own eyes what they’re going after. That’s a pretty low-risk thing because it’s in space and you’re just trying to hit something that’s trying to hit you.

I think in the next 10 years, you’re going to see a lot of these applications develop AI to one degree or another, so long as we think it’s safe and it’s not going to make mistakes.

Before we get on to the Anthropic discussion, we really appreciate you coming here. My lord, this has been so informative. Thank you, Emil, for coming here and explaining to the American public and to us what you’re working on.

It really makes us—I think I speak for everybody—really confident in what you’re doing. It’s so great that you’ve left the private sector to do this.

What I would say, just very quickly, Emil, is that I think not enough people understand that the American military has had to fight with one hand tied behind its back. That little insight you just gave about Afghanistan seems so scary to me.

The men and women who sign up for the American military are doing this to fight on behalf of this country. They deserve a lot more than being sent there and all of a sudden being given this rulebook and told, “Do your best.” It’s like, “Oh, wait. You violated 19 rules trying to protect America. Do your job.” That’s insane.

Let’s just—

It’s really insane in some cases. My belief is that’s what the frustration for those soldiers who were out there in those wars was more than anything. There was the broader frustration: What are we doing here? And then the secondary frustration is: While I’m here, why can’t I do my job?

Yeah. Is there much of a debate internally—and sorry, JCal, before we move forward on this—regarding this idea of full autonomy in military action? I don’t want to speak ahead to the Anthropic point, but it was something that the media seemed to say was part of Dario’s concern: When you press the button and hand over complete autonomy, and there’s a kill action that you’re now giving to a robot or to some autonomous system, do we then have a moral issue at hand? Is that something that’s debated or discussed? Is that the right way to think about the framing of what goes on?

I mean, we’re not even close to there yet, right? These systems are not ready to launch if they have real risk for a civilian. So, we’re not even debating that.

We’re just trying to get basic autonomy in drones, basic autonomy in underwater unmanned vehicles, and basic autonomy in collaborative combat aircraft that fly along with a jet aircraft so that it has more firepower, but it’s still tethered to what the jet does.

Incredible. Yeah.

Yeah. So, we’re just at the very beginning of this stuff. But Golden Dome is a good example. Who can oppose that? It’s the only way to deal with a threat like that. Who could oppose having a laser at a military base with a bunch of soldiers sleeping there that can autonomously take down drones?

It’s pretty scenario by scenario, but we’re not having a lot of debate because the Skynet thing is not a realistic thing at this moment, right? Except, one thing I did tell the Anthropic guys—or I’d tell any company—is that your models are getting stolen by the Chinese. They’re going to unguardrail them and use them against us. Then you want our models to be less capable against your models. It’s sort of—

They’re not going to be thoughtful. In other words, they’re going to go for it. If we benchmark this against where we were 10 or 15 years ago, there was the WikiLeaks video “Collateral Murder,” I think they called it, where we tragically had an Apache take out some journalists. This technology, even applied today, probably would have avoided that, in my mind.

We have enough that when you’re targeting not drones but people on the ground with an Apache, this would have probably avoided that.

Yeah. Or the Kuwaiti aircraft hitting an American aircraft, making a mistake because it doesn’t have the identification. I mean, it’s the same self-driving argument to a degree. Self-driving could save lives, even though it’s scary to look at a car without a human behind the wheel. There are tons of scenarios where it’s a way better, safer option—more precise than the alternative.

All right, before we move on to the Dario thing and Anthropic and that brouhaha, there was one piece that we haven’t addressed with this interaction, Friedberg and Chamath: the Israeli government and its desire to take out this regime.

According to Tucker Carlson and a large contingent of the MAGA base, they feel that we are captured by this group. Does Israel have too much influence over the United States with regard to these actions in the Middle East? This is a big debate within the party, within the Republican Party, and within the MAGA constituency.

Number 1, we don’t want these wars. Number 2, is Israel driving this thing to the point of Rubio’s quotes that, “Hey, we’re doing this because Israel is going anyway”? I think we should address it here. Not that I have a personal stake in this. I’ll give my personal opinion at the end.

I don’t think the president is captured by Israel in the least. I think he decides what is in the best interest of the United States, and if Israel can be a part of that, then they’re a part of it.

Look, let’s be clear: They’re incredibly capable. In something like this, being able to incorporate the intelligence of Mossad—what you’re seeing today in Operation Epic Fury—we’re 4 days in, and Iran has been 90% depleted of all of its munitions. It looks like they’re just firing no more missiles out from Iran to anywhere else. There are fleets of drones and planes just waiting.

Everybody knew where the Iranians were. It’s great that when we make a decision on something that we need to do, we can rely on our allies. I think the opposite question should also be asked: What was the UK doing? Why is Spain pontificating? Why was Europe taking the weekend off before it could even issue a statement? Why don’t you ask that question?

Yeah.

No, it's an equally valid question. Friedberg, do you want to get in on this or no?

No.

I'm a Jew. No one's going to care what I have to say. They're either going to say, “This guy's a Jew. We shouldn't listen to him.” So let's move on. Go ahead.

Yeah, yeah, yeah. Emil, any thoughts on this? I do want to know from you, though: Is Iron Dome working? Is the laser system in Israel operational? If so, are there any success metrics you can share around it?

I think that the Golden Iron Beam was the first generation of the Israeli air-defense thing, and then they're building Iron Beam. I think it's still early, but it works. They're a technologically sophisticated country that's very small, that has a reason to invest in these things, and they have a lot of smart people to do them. So I think it's good.

Does it primarily work on rockets? I want to understand the logical evolution of this because, in the 1980s and 1990s, there was a lot of conversation about space-based lasers that could shoot ICBMs out of the sky to avoid global nuclear war. We could always take out every nuclear warhead delivered on an ICBM. Is that technology feasible? Is there a place in the near future where we could see basically maximal global deterrence using these systems, either ground-based or space-based, to take out hypersonic missiles?

I think the harder but more valuable problem to solve would be the space-based way of doing it, because then you could get at almost any kind of threat that hits space. But you still need a ground layer because there are cruise missiles that could come at you, drones, and so on. We call them multilayered systems: How do you get every kind of weapon at every layer?

Directed-energy lasers, as they get more powerful, could take on a bigger weapon farther away. As those technologies improve, they get more and more capable, and I think all these defense systems are going to get more and more capable of getting more and more varieties of weapons at farther and farther standoff. That's what you want. You don't want to shoot it when it's right over Tel Aviv. You want to shoot it when it's still over their land, ideally.

Are the laser interceptors in the field today? There are reports that they are.

I think there are some. I think they've demonstrated some of them.

Got it. Is this our technology or Israel's technology? President Trump said, “Hey, that's actually our technology.” Is there any insight there?

We have collaborations with Israel on some of this stuff. They have their own, and we have our own. It's not one or the other. They're good at technology, and we're good at technology. There are certain ways you get part of our system and part of their system because it's a quickly evolving part of science right now.

How do you cohere beams of light to get distance? How do you use high-powered microwaves to just drop drones in their tracks? There are lots of different ways to get at some of these things. A lot of it's ours, and some of it's theirs.

To the earlier question, I am pro-regime change if it can be done thoughtfully. Obviously, isolating a dictator is the best thing you can do. We've done that successfully with Putin, Kim Jong-un, and others. Keep diplomacy up. But if there's a moment in time when you could free the people of Iran after 50 years of being subjugated by these lunatics and dictators, I'm all for it.

I actually trust President Trump to make that decision. I know this may sound crazy. People think I'm a libtard or something because of the way my besties frame me on this program, which is completely inaccurate. I'm an independent.

You are.

I'm completely independent, based on my voting, and I'm not on either one of these sides. I am pro-President Trump, and I trust his judgment. I think he has more information than us. I think you have more information. I actually trust you guys to do it thoughtfully.

There obviously was a window here. Israel can have its own motivation. That could be the China motivation, but there's also spreading democracy, which might be the least of people's concerns here, but it's at the top of my list. I would like to see the people of Iran free.

Just to build on your point, Jason, the thing that Emil said before, which I think is important as well, is that we have an enormous amount of learning from what happened in Iraq. We also have a ton of learning from the Iran-Iraq War and a ton of learning from 1953, when the U.S. and the British deposed Mossadegh—or at least fomented that—and put the shah in, and then the shah was booted out.

If you take those 3 chapters in Iranian history, or that regional history, there's a ton to learn. To your point, there is a way to affect what we need to do without creating some 20-year forever war.

There was an incredible tweet. I don't know if you guys saw this. Somebody said, “So every war doesn't have to be 3 decades and trillions of dollars to your friends in Virginia, Maryland, and D.C.” Did you guys see that tweet? It's true. These things can be one and done, in and out.

If President Trump succeeds here, I just want to also give him some flowers. The people of Venezuela and the people of Iran being free represent about 5% of the people in the world living under an autocracy, under a dictator. If both of those flip back to democracies, he'll have done more for the spread of democracy than any president for many decades, perhaps in our lifetime. This would be incredibly noble, incredibly noble, incredibly just.

Would you, in the human-rights set, want him to get the Nobel then?

Absolutely. Give him all the Nobels. Literally, if you can free people—

All of them.

Give him every prize. Give him an Oscar.

Physics, chemistry—[laughter]—he can have everything.

Physics, philosophy.

Jay Cal's an independent. When's the last time you voted for a Republican presidential candidate? Just curious.

Yeah.

Say it. Say it.

No, no, no, no. Montdale. I would have voted for—if I was of age, I would have voted for— I wouldn't have voted for the Bushes. I voted for the moderates, obviously Clinton and Obama.

Oh, we're playing the would-have, should-have game.

I would have voted for Reagan in—

I would have bought NVIDIA at $4.

Well, no. I didn't vote for Kamala, so I'll leave it at that.

Why don't you say that you voted for President Trump? Just say you voted for President Trump. I don't want to complicate things, but you did, so just say it.

I didn't vote for Kamala. I'll leave it at that. All right.

It's so weird that you'll say you're a moderate, but you won't say that you voted for President Trump.

I am supporting President Trump in about 60% or 70% of what he does. Let's leave it at that.

3, 2— all right. Let's talk about the economic impact of oil and insurance. Oil has risen to $84 a barrel Wednesday, with the Strait of Hormuz basically at a standstill at this point. Here's the clip. You can see the traffic slowing down, and then some of the dots are even going away. That could be ships being taken out.

Unless the Strait opens, 3.3 million barrels of daily production would be lost early next week. Then there's insurance companies: They've all canceled the war-risk coverage of vessels in the Gulf, effective March 5. Supertanker traffic dropped 94% within the first 48 hours. Trump said the U.S. would provide political-risk insurance for all maritime trade through the Gulf, especially energy.

Friedberg, your thoughts on the economic second-order effects that we're starting to experience here and that, over the next 4 weeks, could be intense and acute?

The modern insurance market emerged specifically to solve the risks of maritime trade. In the 17th century, Lloyd's of London was a coffee shop in London where all the maritime traders would get together and talk about, “What's the safest route so pirates don't get our ship, and so we don't run into weather?” That's where they would have these conversations, and eventually they started underwriting the risks of the shipping routes and giving each other guarantees.

They said, “If you make this route, great. You pay me a certain amount, and if you don't make the route, I'll pay you the lost value.” That's how Lloyd's of London, which is the world's biggest reinsurance market, started.

Today, Lloyd's of London has 78 what are called syndicate members. These are pools of reinsurance that underwrite big, crazy risks, like maritime insurance for folks moving oil tankers through the Strait of Hormuz, which the IRGC just announced they're shutting down.

When the IRGC announced that they were shutting down the Strait of Hormuz, there was a significant risk of mines being placed in the Strait and ships being attacked and blown up. So, loss of value. The insurance premium spiked initially from 0.25% of the value of the ship to 1.25%. It went up by 5×, so folks had to pay a lot more of the value of their ship in order to continue the routes and get guarantees that they'd make it through.

Then all of the markets started to shut down. Once the conflict got heavier, everyone said, “Let's shut this thing down.” That's obviously a massive risk to energy prices globally, which drives inflation and puts U.S. economic security at risk.

And so this is a brilliant move. I would say the U.S. government stepped in with the U.S. International Development Finance Corporation, which was actually, funny enough, started a couple of years ago, in 2019 or something like that, as a kind of outgrowth of one of the agencies that provided credit from USAID, much talked about, USAID. And so they're leveraging the credit capacity of this old USAID agency to go out and say to all the shipping companies, “Hey, we'll give you insurance on your routes.”

And the reason they need it is the shipping companies are leveraged. They take on debt to buy the ships, and the lenders require that they have insurance, or else they're not allowed to take the routes, because the lenders are ultimately going to be out the money. And so the shipping companies themselves need to have insurance. This provides a market that has now gone away.

Very smart.

And ultimately, a lot of people are saying this could actually reshore or onshore maritime insurance back to the United States and create an entirely new insurance industry here in the U.S. that has historically been served almost exclusively by European syndicates and European partners. And it actually creates a big economic opportunity as this war dies down for American insurance companies and American brokers to basically be the underwriters and the guarantors of this sort of insurance and create a new industry. So that's a super interesting kind of side story on what's going on here.

All right, some breaking news here, folks, via Bloomberg. The Pentagon has formally notified Anthropic that it's been deemed a supply-chain risk. This has never happened to an American company; it has happened to Russian companies and Chinese companies, including Huawei. For background, the Department of War canceled Anthropic's $200 million contract on Friday and said they would do this.

The dispute came down to 2 clauses, according to sources, and we have 1 of the principals here, so we will hear directly from him in a moment. Anthropic had 2 concerns. Number 1: fully autonomous weapons, aka murderbots. As we previously discussed, Dario didn't feel that their technology was reliable yet and wanted some assurances.

The second thing Anthropic said was they were concerned about mass surveillance of Americans, because they believe this technology is uniquely powerful and can do things beyond what a series of webcams or a network of 7-Eleven cameras can do. The Pentagon said they wanted all lawful use. Dario, you're welcome to come on the program next week or any time to give your side of the story, but this week we have Emil. Emil, your thoughts, and explain to us what happened here and how this broke down.

It's worth a little history—a short history. If you remember the Biden executive order on AI, it was this crazy executive order that limited the amount of compute any model company could do and essentially grandfathered in a small number of AI companies that they were going to designate as the winners, and everyone else was out, so they could have more control over what they did. Anthropic was one of those winners.

And then they were smart. It was a good sales strategy to sell into the most sensitive parts of the U.S. government, like all of our combatant commands—CENTCOM, that's doing the Iran fight now; INDOPACOM, which is sort of responsible for China; and several of the intelligence agencies. They did forward-deployed engineers, Palantir-style, so they got very sticky to the workflows and all that.

So I came in and got the AI portfolio for the Department in August, and I said, “I just want to see the contracts.” The old lawyer in me. I looked at the contracts and was like, “Holy cow, they say you can't use them to plan a kinetic strike. You can't use their AI model to move a satellite.”

You can't do a war-game scenario with it.

You could do a scenario, but you can't say, “Let's suppose you're writing a plan saying, ‘If this happens, here's what we would do,’” if it might involve a kinetic strike that causes harm to a human. What do you think these folks do? We're the Department of War. This is what we do.

So I said, “Okay, well, I've got to, number 1, have direct relationships with these companies, not just through Palantir, because I want to use it more broadly. And number 2, I need to have the terms of service be rational relative to our mission set.”

So we started these negotiations, and it took 3 months. I had to give them scenarios, like this Chinese hypersonic missile example. They're like, “Okay, we'll give you an exception for that.” I said, “Well, how about this drone swarm?” They said, “We'll give you an exception for that.” And I was like, “The exceptions don't work. I can't predict for the next 20 years all the things we might use AI for.”

And so all lawful use seems like a good thing. If Congress wants to act, great. We have our own internal policies; we'll follow them. We're not knuckle-draggers here. We don't want to hurt people unnecessarily. So it's our province to decide how we fight and win wars, so long as they're lawful.

And I think at some point it turned into a PR game for them, because they were not going to win this intellectual battle of, “Well, we're going to stop you. We're going to use our judgment because we think Congress is behind us and impose it on the U.S. military.” And it became this: let's find the issues that are most inflammatory—robot weapons and mass surveillance. I mean, we're the Department of War. We're not the FBI. We're not Homeland Security. We're not—

You're not allowed to legally spy on Americans.

Yeah, you're not. And then what it came down to on that issue, just as an anecdote, is they didn't want us to bulk-collect public information on people using their AI system. They wrote it in a way that I was like, “So you're telling me, before we got to bulk-collect, if someone types in, you know, ‘Chamath's at LinkedIn,’ and I'm using publicly available information, I would be violating your terms of service?” They said, “Yeah.” I said, “Well, okay, let's rewrite it.” That was months of this stuff, which was sort of interminable.

And then the trigger point was after the Maduro raid. One of their executives called Palantir, the prime contractor we buy it through, and asked them, “Was our software used in that raid?” Which is, by the way, classified information anyway. So we're trying to get classified information and implying that, if it was used in that raid, that might violate their terms of service.

So they wanted to enforce their terms of service. They went behind your back to try to collect information to then maybe pull your license for their technology.

It wasn't behind my back. I don't want to accuse them of that. Palantir is the prime contractor, and Anthropic is the subcontractor. But it raised enough alarm with Palantir, which has a trusted relationship with the Department, to tell me. And I'm like, “Holy cow, what if this software went down, some guardrail kicked in, or some refusal happened for the next fight like this one, and we left our people at risk?”

So I went to Secretary Hegseth and said this could happen. That was a “whoa” moment for the whole leadership at the Pentagon: that we're potentially so dependent on a software provider without another alternative, and that provider has the right or ability not only to shut it off, but maybe a rogue developer could poison the model to make it not do what you want at the time, or sort of trick you because you have to trick it.

I mean, all these things that we worry about with models: hallucinating purposefully, not following instructions, and some insider-threat stuff. Then that culminated in the Tuesday, kind of dramatic meeting with Secretary Hegseth, Dario, and me, with the Friday deadline that got blown. I never thought they really wanted to make it.

Is the model entirely hosted by Anthropic, or—just explain to us technically—does this sit in a cloud that Palantir runs for you guys? Is there really technically a way that employees at Anthropic could intervene in the use of the model?

Yeah. They put their model in AWS GovCloud—

GovCloud. Yeah.

And then Palantir serves it from there, and they refresh it. They hold the control plane for the model.

So, yeah, they can change the model weights if they want. They can do whatever they want.

Yeah.

The insight into this thing is unbelievable, not just for governments, but now if you're running a company. The reality is that what Anthropic showed—which, by the way, is their right at some level—is that they're going to have a political perspective and a set of terms that reflect their philosophy, and that philosophy can change on a dime. But what the government did was also completely reasonable: we can't rely on you if you're going to be completely unreliable and disallow things that are reasonable.

I'll give you a different example to make the point. There's a state that wants to run some healthcare program, but they're a pro-life state. You can't conduct abortions in that state. Does that mean that Anthropic engineers can decide, “You know what? We're pro-choice, so we're going to change the access model and the capability of that model inside of that state”? Is that allowed? Should that be allowed?

At 1 level, you'd say, “This is a private company. They're allowed to choose.” But what that really means is, for the government, for all the states, for any city, and for every company, you cannot choose to only use 1 of these things, because it is just a matter of time until some person inside one of these companies goes on some lunatic moral tirade and then jeopardizes your business over something that is nothing about law, but is everything about subjectivity.

That is the huge thing that this tore open this weekend. If you're not figuring out how to be multi-model and agnostic across these models, you're taking on enormous business risk after Friday, because you can't tolerate that these folks will do that. It's too critical a technology.

By the way, this is deplatforming all over again. Remember what happened when you didn't like what was said? All of a sudden, you were deplatformed. This is that times a thousand, because this is not about posting on social media. This is about using fundamental technology to either advantage or disadvantage your business. Emil.

Yeah. I think I described it the other way the other day: the CEOs of these companies say they're going to cause 50% white-collar unemployment. This is as powerful as a nuclear bomb. It's like 50,000 geniuses in a data center, so you could have a small country coerce the world into its will. You're like, “Holy cow.”

All right. So this is a general substrate of intelligence, of technology, that's applicable to a lot of things. Very generalized. It's not like Workday, like HR software, where we could just use a competitor. This is going to be part of our everyday life in so many different ways, and controlling whether it has a moral conscience—Anthropic has its own constitution. It has its own soul. It's not the U.S. Constitution.

So you're subject to that, plus whatever whims and how that changes. That's a scary thought for Americans generally. I think that did come through a little bit today. In the coming years, it's going to be a bigger and bigger deal.

So take us through OpenAI's software, Gemini's software, and Grok's software. Have they pushed back on any use, or are they like Dell or Apple? They sell you a computer, and you have the computer and can use it as you will. Have any of those given you any pushback?

Grok's all in for all lawful use cases across all classified and unclassified networks, as you'd expect, because Elon is truth-seeking. We want truth at the Department of War. We don't want ideology.

Because ideology will mess with operational decisions. You don't want anything to be fake or tilted.

We're surging Google, and we have Google for all lawful use cases on classified networks, and we're trying to move them to classified networks. They just have to build out infrastructure because this stuff's complicated.

So they're in compliance in terms of what you're looking for as a partner. I guess the last one is OpenAI, and Sam seems to be—

Just characteristically playing both sides a bit. To his credit, I called him and said, “I need a solution if this thing goes sideways. I need multiple solutions. I'd like you to be one of them.”

And he's like, “Okay, well, what can I do for the country?” I was like, “I need to get you up and running as soon as I can.” He was trying to protect Anthropic, to his credit. He was like, “Don't call them a supply-chain risk. That's bad for the industry. Let me maybe negotiate terms that they'll find acceptable.”

But he's in the middle because they compete for the same researchers. A lot of this comes down to these 1,000 researchers, like baseball players, that get traded between these companies.

Moneyball. Yeah. These are the best of the best. It's a very Moneyball-ish sort of thing, and there aren't that many of them, and you lose—

20% of them, and all of a sudden, they launched Claude Code before you launched Codex, or something like that, and then the numbers changed pretty dramatically.

Sam was being a real patriot, to his credit, and trying to help Anthropic while they were trashing him and recruiting from his company. I'm not biased. I just want all of them. I want to give them all the same exact terms, because I need redundancy.

I want to see if they diverge or not, or whether they converge. Maybe I only need 2 over time, but we don't know. It's too early.

But why keep them in the mix? If there's clearly a difference in operations and philosophy and how they want to run their business, and there are other models, is their model particularly good at particular applications that make it important to keep it in the mix, given that there are 3 or 4 other alternatives here?

Anthropic, you mean?

Yeah.

The number 1 reason we were having this conversation at all was because they were deeply embedded. So now I have to disentangle them, and the other companies have not gone as heavy on enterprise sales, forward-deployed engineers, or government business. They have to catch up, not necessarily on the capability of the model, but just on how you serve the government. Anthropic is just way ahead on that.

Right. But the models themselves, you don't think, are uniquely advantaged, or do you have a view on that at this point?

I don't have a view on that. I don't think they're—I mean, certainly Claude Code was innovative and ahead. That's true. But I do believe that in 12 months, Codex is not going to be far behind.

I think you're right. There's an asymptoting that's happening. If you just look at the confidence interval on how overperforming or underperforming some of the leading models are, the error bars are shrinking, right? The confidence intervals—these things are all kind of becoming the same eventually.

They're all getting access to enough power, enough compute, and they're generating similar results, which I think you would expect. So it's even more important that you have a collection of models.

The other thing, Emil, I don't know if you saw this, but they posted about the revenue ramp of Anthropic. I have a small software company called 80/20, and I asked the team, “Let's go look at our OPEX.” I posted it because I was so shocked at these numbers.

Our costs have more than tripled since November of 2025. Between the inference cost that we pay AWS, which is ginormous, and our costs with Cursor and Anthropic, we are just spending millions.

So now more per unit and more in aggregate?

Both. But the problem is that my costs are going up 3× every 3 months. My revenues are not.

Token use is very addicting.

Yeah. And by the way, because everybody has gotten infatuated with what we call these Ralph Wiggum loops—just send the thing off and it'll figure something out. A, it never figures anything out, and B, you just get this ginormous bill from Cursor.

One of the things we had to do was say, “Guys, you have to deprecate Cursor because you're just wrapping Claude Code and charging us way too much for these tokens.”

I don't know if you're seeing any of this, where the tool usage is so great to use these tools. Let's be honest, it's super fun. It's like you feel like a genius, but then the ROI of these tools is really important. I'm not sure that's as much of an issue for you or not—

It will be. It will be, for sure.

As people find more and more use cases, the use cases get more sophisticated. So the next marginal thing you have it do is likely to be harder and therefore be more consumptive. Right?

Right. Right.

Let me just ask you the important question that I think triggered a lot of the news this week: Why then designate them a supply-chain risk? Why not just abandon them, move on, and use the other vendors? Why take this kind of punitive action?

I don't view it as punitive, and I'll tell you why. If their model has this policy bias, let's call it, based on their constitution, their culture, their people, and so on, I don't want Lockheed Martin using their model to design weapons for me.

I don't want the people who are designing the things that go into the componentry to come to me, because if you believe in the risk of a poisoning threat, it can enter into any part of the defense enterprise.

But it's just the defense enterprise. So Boeing wants to use Anthropic to build commercial jets, have at it. Boeing wants to use it to build fighter jets, I can't have that, because I don't trust what the outputs may be, because they're so wedded to their own policy preferences.

I guess a dovetail to that is, why couldn't this have been handled quietly? Is this Anthropic who made this a public spat, or was it the administration that made it a public spat? It takes 2 to tango.

They have a very good, sophisticated press operation, and they're really good at painting us as doing mass surveillance, where their issue was some commercial database thing that someone else could buy. They didn't want us to buy it to use it, which I'm not even sure we buy from them, except to do recruiting for soldiers.

We run schools and hospitals. We do a lot of things at DoD. We don't just fight wars. The way they were able to characterize these 2 things, which are genuinely scary to people but were not the real issues, was really the concern. I worried about them shutting off our system at a moment of need or messing with our system in theater.

What came to mind is, if they're selling you batteries and you need to use the batteries or the laptops—however you need to use them lawfully—okay, that should be enough for them, unless they're peaceniks and they don't want to be involved in selling weapons.

By the way, that was Google's position for many years: They just didn't want to be involved in it, because to your point, they want to recruit talent that is also aligned with that. So it just seems like maybe this isn't the right partner for the Department of War.

Yeah. If you don't want your stuff to be used for Department of War stuff, you shouldn't be selling to the Department of War.

[Laughter] Pretty sure it’s in the name. It’s in the name.

Well, and then also, I have to say, when you said, “We don’t know how we’re going to use this thing,” what immediately came to mind was 9/11. You have to go check with them if you find out there’s another unique, 9/11-like black-swan event that’s going to occur and you have to clear it with them. That’s—

That was literally the comment. So, I was in a room of 20 people. This is not deniable if anyone—if Dario—wants to deny it.

I was giving these Golden Dome scenarios and so on, and he’s like, “Just call me if you need another exception.” I’m like, but what if the balloon’s going up at that moment and it’s a decisive action we have to take? I’m not going to call you to do something. It’s not rational.

Yes, that was another holy-cow moment of how they think about it.

That just means that what he wants to be is the secretary of war.

That’s right.

He wants to be the god-king there, I guess.

You can’t do that. The thing that shocks me, Emil—I don’t know, maybe you can’t say anything, but you guys can comment on this—is that it’s clear Anthropic just lost all the Republicans. But I think that if they think they have the Democrats, that’s fleeting as well, because I think progressive Democrats fundamentally just hate Silicon Valley and technology. There’s no way they’re going to let some god-king over here that they don’t control either.

In both ways, I think they accidentally may have pissed off every constituent. The longer-term fallout among them and progressives will come home to roost, because as the progressives want more control and these guys push back on them, they’re just going to fall into the same situation.

Yeah, I mean, it’s an interesting perspective. I think if you don’t want to be involved in war, you’re right. I think you mentioned this three times, Chamath: don’t sell bullets if you don’t want to be in war. But you can’t call Smith & Wesson and say, “Can I—”

The other thing is, what the hell were the senior management and the board talking about over these last few days? To me, it would have sounded insane. So then the question is, were people just so breathless to buy this revenue curve? What is the board doing? What is the senior management really doing?

What would you change, guys? What do you think you would tell them if you were sitting on the board of Anthropic? If you’re an investor and you’re on the board, what do you say to Dario when he says, “Hey, I need to dictate to Emil and Anthropic how they use my tool,” and everybody else is just saying lawful use as the standard? What’s your coaching advice?

Well, it’s also a very unusual circumstance, because I don’t think any business in history has grown as fast as they have in the last 90 days. They’ve added—what was it?—$6 billion of ARR.

Yeah, in a month or something.

I mean, that’s absurd. It’s a great product. OpenClaw has driven a lot of this.

If you’re on the board—

You’re closing your eyes. Yeah.

You’re shutting up. You’re just shutting up because something’s working.

You’re actually—[Laughter]

I think he’s off doing his thing, and they’re going to let him do it. I don’t think that company’s worth $350 billion anymore. God knows what it’s worth.

Oh, that’s interesting. If you get a block of stock right now, where do you put a bid in? I’ll tell you where I—

Oh, my God. I had this conversation at dinner two nights ago. You have to pick between OpenAI at its current mark, Anthropic at its current mark, or Google. It’s either multiple from here or net market-value creation from here, because those are actually 2 very different conversations.

Explain the difference.

I think the net market valuation is different because Google’s already worth $3 trillion. So, if they double, they’ve added $3 trillion. But I think Google is the bet. I think Google is the market-value-creator bet. Anthropic is the multiple bet. I think Anthropic is a $1.5 trillion market cap at the end of the day.

Unless this blows them up, you’re still buying the 5x versus the 3x kind of thing.

You’d buy the 5x instead of the 2x.

But if you get to buy a block of stock now, do you buy it at the last post, do you buy it at a discount, or do you just say, “I’ll buy it at the last post”?

Anthropic is worth a lot more than $350 billion. That’s for sure.

I think it’s undervalued compared to ChatGPT.

They just added $6 billion in the last month. I’ll tell you anecdotally: everyone I talk to is on Cowork. Everyone has gone deep on this. Everyone’s amazed and shocked and actively using it. Everyone’s saying the same thing, which is that Anthropic may actually be fulfilling the promise of AI.

I’ll also say that it’s only going to take 90 days for Google to flip on a virtual version of Cowork. Once Google integrates this with G Suite and you have a virtual, hosted version of Cowork, it sweeps the market with this same competitor. But right now, Cowork is such an incredible product, and everyone’s saying the same thing.

It's like giving Elon giving truth to AI. Elon said something with respect to Grok: he expects it to exceed all of these coding models, probably in the May sprint, but for sure by June.

So, to your point, Emil, what do you guys do when all the models asymptote? Let’s just say, for the thought exercise, by October of this year. Let’s say I can guarantee you that by October, all the models are the same. Do you just take a complexion of them all and say, “Great, we’re going to build some governance layer around it,” and now we’re indifferent? What do you do?

I would love to be indifferent, because then I could compete on price, right? Then I’d have 1 main and 1 redundant, or 2 mains, and I’d need at least 2.

Yeah.

Anthropic’s not going to be 1 of them if they continue with their posture. So then it would be 3. If 1 gets wobbly from a policy scenario, too—because they all, except for Elon’s, are based in San Francisco and have that vibe to them—you kind of want to have 2 or 3 at any given time. Then you price-compete them.

I do think Google has a long-term strategic advantage, not only because of their consumer business, but because they have their own cloud. Google doesn’t have the margin on top of the cloud that Anthropic will have to pass on, so it’s an interesting economic proposition for them.

And just to build on your point, Friedberg, after you finish your insightful comments here, pull this up, Nick. Almost on cue, Friedberg. You’re such an oracle. Here is the announcement from Google: Google Workspace is now integrated for agents, and 40 agent skills were included today.

Emil, you’ve been great today—super honest. I’m going to give you some fastballs here. Dario says, “The real reason the Pentagon and Trump administration do not like us is that we haven’t donated to Trump, while OpenAI and Grok have donated a lot.” Here’s Claude’s answer to that claim.

Here are 9 companies and their activities with the administration, from the inauguration to attending the inauguration, to the White House CEO dinner, to the Melania documentary. If you go through and look at these 9 companies—Microsoft, Apple, Tim Cook, NVIDIA, Amazon—they have all participated. There’s 1 company that hasn’t participated, and that’s Anthropic.

Is Anthropic being singled out because they are not genuflecting and because they’re not paying the cover charge? People say this administration is pay-for-play. That’s the accusation he’s making. I’d say maybe there’s a cover charge. Nobody likes to pay it, but the other companies have. What do you think here?

I mean, it’s literally 1 of the dumbest things I’ve ever heard. I’m in the Department of War. I need to win wars. If you help me win wars and I don’t have to waste time transitioning you out, that makes me thrilled.

It’s a criticism of me, because it’s not like President Trump dipped in and said, “Hey, Emil, by the way, those guys didn’t get any money. You can’t use them anymore.” Obviously, it’s an invention of his own mind. I don’t know if people sleep at night if those thoughts get in there.

I was trying to work with them. Why would I spend 3 months trying to negotiate with them to get to a simple standard if I would have just said, “Okay, guys, you’re out. Bye”? I think it’s just some internal psychosis. That’s the only way I can explain it.

Okay. It could be on Dario that he’s antagonistic to the administration, both with respect to how he operates commercially and in the fact that he doesn’t want to support the administration.

I have a different theory. I think that they have a massive instance of Cowork internally that helps them come up with business strategy. I bet there’s some element of AI that says, “Yeah, you should do it. Do it. It just makes sense. Zig where they zag and get more press.”

Now there’s some Claude bot telling them to basically tell the Department of War to pound sand. It’s going to turn out to be the stupidest decision.

Listen, if I were chairman of the board of that company, I’d pull Dario aside and I’d say, “Listen, you’re obviously a genius. We obviously have the best tool in town. This is not a battle you can win, and it makes no sense. You're going to come across as not being patriotic. Tim Cook is showing up for the Melania premiere. Would it kill you to support the president? Would it kill you to show up?

Look what happened when Biden excluded Elon—that rankled him. Show up for the president. Show up for America and be a patriot. You don't have to donate, but be a patriot and show up for the dinners.

That's terrible advice. Here's my advice.

Okay, here's your advice. Okay.

Hey, Dario, call Emil back right now and say, “You know what? Sorry, we screwed up. We're going to own this and put out a press release that says we support our customers' use of our models to do everything and anything that's lawful. Number 1, our terms of service are written in stone, and you can expect stability and reliability from us. This was just a misstep.”

I would say that's what I've always wanted. I need a reliable, steady partner that gives me something that'll work with me on autonomy, because someday it'll be real, and we're starting to see earlier versions of that. I need someone who's not going to wig out in the middle. We're just at the early stages, and it's rational.

But then, you called President Trump in your 5,000-word essay on Friday a wannabe dictator.

You're going to have to apologize to more people than just me.

Yeah, maybe it's time to rewrite the position here. Let's just say kumbaya, everybody. Kumbaya, we solved the problem. And look who's on the line. Surprise guest. Dario's here. I thought I would surprise everybody. Nick, pull Dario up. No, he's not here.

What's your view on how the industrial supply chain for hardware components and systems is coming along in the United States? My understanding is that we're trying to reduce our dependency on Chinese-manufactured components. Where are we with respect to where we need to get to in the U.S. manufacturing supply chain?

We are in the early days. Critical minerals—you've seen the action around that. You'll start to see more of it. I have the Office of Strategic Capital, which has $200 billion in lending authority. What we're trying to do is offer a Treasury-plus-100-bps loan to companies, showing them that the department needs their solid rocket motors, their batteries, their fiberglass—all the things that we're heavily dependent on for our defense industrial base, which are completely outsourced to China—and domesticate them here.

We've got a bunch of great people running it, but it's early days. It's going to take the rest of the term to get there. I think we'll get critical minerals done before the rest of the term, where we have access to what we need from the U.S. or allied countries.

Batteries are the next problem I'm trying to solve. For example, batteries are totally outsourced to China, both technologically and in terms of lithium. There are, call it, 20 critical things. If I could get to all of them at some level, then it would take a few years for them to build plants and do that stuff. But it's very important.

I hope whatever administration comes next continues it, because I'm all free market, but we outsourced so much that it crippled the assembly part of putting all these things together.

Do we have a munitions risk right now, given the conflict that we're involved in?

We don't have a munitions risk, but we do need to plus up because the Europeans are taking a long time to contribute to Ukraine. Russia has consumed a lot of munitions from all over the world, and then, obviously, there are these conflicts we've been in. We need to have the next generation.

To a large degree, we're still fighting with 1980s Cold War weapons, right, and not modern weapons, so we need to plus up those things to regenerate them. Our nuclear missiles are 50 years old, and some of the planes are 40 years old, so all of it has to be renewed.

Do you think—just speak to the venture capitalists in the audience—are we in the early stages of this kind of defense-tech boom? Is defense tech well-funded at this point, or is it too hyped and bubbly? Is that not really the issue? It's not about funding the companies; it's about funding some of the further-upstream issues that we're facing. What's your view on where we are there?

There's more defense-tech venture capital than ever—3 times more than last year. It's growing. What I need to do, and what the department needs to do, is have some of these companies win big contracts quickly, whether it's Anduril, Shield AI, Saronic, or a bunch of these companies, so that more money flows in, more entrepreneurs do it, and I can buy more.

Genuinely, I do think warfare is going from big aircraft carriers that cost $20 billion and a decade and a half to build to mass-attritable, low-cost things. That's what these new entrants can do. We need those to succeed so that the flywheel goes with venture-capital money, entrepreneurs, and capabilities.

In that sense, what I've heard as the explainer for this is that we're moving from the old primes to the new primes, that there's going to be a small set of big winners and then, obviously, lots of second-tier companies and subs and whatnot. Is that really how this market's going to evolve? Are we going to end up with Anduril, Palantir, and maybe 3 or 4 others, and that's where most of the value is going to accrue from a market perspective?

Anduril and Palantir want that, and I joke with them all the time about it, but I definitely want at least a second layer that's innovative and trying to disrupt the first layer all the time. I met a mom-and-pop, wholly owned company that makes these missiles called RAMs that we really sell and send to Ukraine, and they do it with 30 people. They can do 1,000 a year because they've designed and manufactured them, and it's awesome.

I want companies like that to continue innovating. Maybe BAE buys them. One of the reasons the primes are such a small number—not the only reason, but one—is that they learned how to contract with the government. They learned how to go through the bureaucracy, and that became a competitive advantage. I'm trying to take that competitive advantage away.

That's a really important point. How do you disassemble all that bureaucracy so that product innovation can actually get to you?

Part of it comes down to requirements reform. What used to happen is people would say, “Oh, we need a new fighter jet.” So the Army, Navy, and Air Force put in the requirements, and it would be, “We need it to be stealthy, to hold a missile, to hold 4 humans,” and it became this unbuildable thing.

But the contractor didn't care because they're getting paid cost-plus. “Sure, I'll fulfill your requirements.” 2 years from now, you're like, “That was never engineered properly. It'll be another few years late and a couple more billion dollars.”

We're trying to change that to: I tell you my common operational problem. I need a bunch of missiles that go 500 miles or more and have this kind of blast. Come to me with solutions, with as few requirements as possible on that side.

On the contract piece, we're trying to get to as close to commercial contracts as possible. This is going to take—and this is where the startups are so good—they'll do fixed-cost pricing. They'll do, “You don't pay me as much if I deliver late. You pay me more if I deliver early.”

It's very disruptive to the existing system.

Super disruptive. But that's what I'm waking up every day trying to do.

So you could put out something saying, “Hey, the Strait of Hormuz is super important. We need to keep it open. We need these types of devices to keep it open.” But come to us with your ideas and let them be creative entrepreneurs, as opposed to just trying to goose the profits. It's really brilliant.

Yeah.

Emil, you also oversee DARPA.

Yeah.

DARPA is the father of the modern internet, and it's created a lot of really critical technologies. Can you talk about what's going on in there? Are there interesting things that you think our audience should know about that you're trying to push forward?

It's probably my favorite part of my office, because it's still a very honored profession to be part of DARPA. Being in government service for a long time has been reduced in stature since the Manhattan Project, because now, if you're a great engineer who wants to do rockets and stuff, you go to SpaceX.

DARPA still has the best of the best, and the most creative ideas happen there. One of the things they're working on that's public is using biology to synthesize critical minerals. How can you pull them out of the ground using biology, so you don't need to do all this crazy, messy, dirty refining? That would change the game big-time in our ability to get the critical minerals we need faster and leapfrog the Chinese in terms of technology.

They're doing a lot of that kind of stuff. They're deep in cyber. Cyberattacks are the next huge threat with AI, right? What we saw with the Chinese creating all these agents to attack systems that Anthropic happened to uncover. They're working on that.

There's not a ton I can talk about with DARPA because it's so classified, but those are a couple of examples for you.

All right. Speaking of classified, just 2 quick questions before we wrap here. Are there aliens, and what are you going to tell us? Number 2, in all seriousness, I'm curious: What have you learned about China, where they're at, the threat there, and our ability to counter it?

Give us some idea of where we're at as a country, because we hear a lot of hyperbolic stuff: They're building this incredible, mobile, small navy. They've got hypersonics. They're just way ahead of us. We hear these things, but realistically, are we competitive?

Well, I'll answer your first question. I fought for the alien portfolio. I didn't get it.

Work to do. More work to do. All the guys on my team were like, “Dude, you've got to get this for us. Please talk to the secretary. We want to do this.”

But I was like, “As long as I had 100% access to everything, I would do it,” because that would be amazing, right?

Saucers would be a game changer.

But on the second one, it is true that the Chinese have had the greatest military buildup in world history in the last 15 years. We're asleep at the wheel to some degree because we're focused on the global war on terror, so they've advanced without us thinking about the threat.

That being said, our operational expertise and our space capabilities—we have some sophisticated stuff, like our subs and our space layer. We still have the best stuff in the world, but we have to make sure that gap doesn't narrow.

Right? We can't be complacent. We should sleep well at night knowing you're there, yeah?

It's reassuring to know that President Trump is allocating money toward this and is decisive in his actions, but we cannot be complacent. I feel like this week was a true reminder of how fortunate we are to have the defense that we have for the United States.

When you look at what happened in Dubai, Doha, and Tel Aviv, and you see how people in their residential homes are getting attacked and bombed, you realize just how fortunate we are to have all of the layers of protection that we have from our government. I've actually come around to this quite a lot.

I'm a true, arguably libertarian, at heart—small government—but the one thing that I've realized is so critical for us to have the freedom to do all the things we want to do is defense. I think it's an amazing institution, very valuable to the United States. Emil, thank you for what you do.

Yeah, thank you. I really appreciate you coming on and being so candid, thoughtful, and insightful. This has been a fun, amazing episode. We'll see you next time. Bye-bye.

Love you, boys. Bye-bye. Let your winners [music] ride. We open-sourced it to the fans and they've just gone crazy with [music] it.

Love you. Queen of [singing] yours. [music] Besties are gone.

That is my dog taking your driveway. [music]

Oh man. Myasher will eat me.

We should all just get a room and just have one big huge orgy because they're all just useless. It's like this like sexual tension that you just need to release somehow.

Your feet. [laughter] We need to get merch. I'm going all in. [music] I'm going [music] all in.