关税、自由贸易、出口管制、H20与稀土禁令｜BG2 对谈 Bill Gurley & Brad Gerstner

Almost everyone in the AI space that I see with a microphone in front of them says the U.S. has to win the AI war.

Yeah. I don't know what that means. And if I guess as to what it means, I don't think it's possible. Right? It's an infinite game. It's an infinite game. No one that I talk to would argue we're going to somehow prohibit them from moving forward in AI. Let's not forget it wasn't that long ago that OpenAI—they'll say—borrowed or copied the innovation that happened at Google and DeepSeek. This is how innovation works.

Bill, great to be back in person. Good to see you, sir. That was fun last night. We have to give a little shout-out to our friends Vinnie and Bill Lee and some others who we got together with in Austin last night for some poker. I think you did all right last night. You had some winnings, and you had some more winnings. In fact, I brought you some flowers. These are for your Florida Gator. I'm bringing you your flowers.

Thank you. I'm sure the listeners are tired of hearing that I started the last 2 podcasts, one before the Sweet 16 and one before getting into the Final Four. Then, of course, everyone knows the Gators went all the way. I had a bunch of friends out. We spent the whole day, the Monday of game day, on the river. We walked outside the team hotel and saw a bunch of old friends.

I know fandom is a weird thing because you don't actually do anything; you just go along for the ride. But it creates a very special experience, and I'm super proud of that team and how they won. I had talked about it before, and it was dramatic, too. It was a great time. Just a great time.

What have you been up to?

Speaking of games, that was an exciting upset by the Warriors in Game 1 over Houston the other night. I hope our guys make a run at it this year. It certainly looks like they're a totally different team with Jimmy Butler on the team. No doubt, I think they're 25–4 or something like that since the Jimmy Butler trade.

1. Complex Systems

Completely. I tell you, man, I'm thinking a lot. In fact, this is something you really drilled into me ages ago about thinking about complex systems. When I look at all the things this new administration is trying to change simultaneously—you and I talked 3 months ago. I said I was worried about the markets: discount rates had to go up; there was more uncertainty in the world.

But when I start thinking about what's going on, what this administration is trying to accomplish at a very large scale, think about this: whatever you think about each one of these individual things, they're trying to totally transform global trade. So that's trade negotiations with 100 countries. We effectively have an embargo between us and China, 140% tariffs, export controls, and so we have to renegotiate that deal. Bessent said today it really hasn't even started. That's going to be a slog.

On top of that, we have taxes. If we don't lock in the tax bill, it doubles at the end of the year. We have this reconciliation package. We have DOGE and deficit reduction. We have questions around the 10-year and whether or not the deficit is going to actually be larger this year. We have the war in Ukraine, and lots of back-and-forth there. We have the volatile situation in the Middle East. It sounds like we were on the verge, maybe, of seeing an Israeli attack on Iran. All of these things are happening simultaneously.

One of the things that's on my mind is that if you look at the discounted rate of dealing with any of them successfully, and then you multiply those discount rates together—the probability that you're going to land this plane with a really smooth landing—I think is challenging to see. We're going to have some wins; we're going to have some losses. Today, I know we want to revisit some of these conversations around tariffs and free trade and these export controls, and kind of where we are in the market.

But I would just say we're attempting something at a very high difficulty level, and if this administration is able to land the plane on all of these, it will be quite an accomplishment.

Yeah. And you bring up complex systems. This is the reason I spend so much time down at the Santa Fe Institute, because that's all we talk about. But a couple of features of those things relate to exactly what you said. First, it's very hard to know which variables are dependent in a complex, multivariable system. There may be one you don't know about that flips, and the whole system takes on a different shape.

This makes these things very hard to predict, which is why I don't like talking about macro for that exact reason: there are so many variables. It's very hard to be accurate and correct. Even if you are, it might just be by happenstance, because you might not know the exact variable that caused it.

Unfortunately—and I'll state again, I don't like talking about macro—but we're kind of forced into it because the market, and even the destiny of many of the high-tech companies that we would prefer to focus on, seem to be caught up in this whole thing.

For sure. For sure. I just saw a random tweet before we came in here—somebody estimating that the revenue hit to Meta this year from reduced advertising by Chinese merchants on Meta, like Temu, is $7 billion. Wow. Now, think about that. That's pure profit, right? There's no incremental variable costs against serving those advertisers.

So, if that just disappears, it's a very, very high-margin business. And so, again, the downstream consequences backfill with something else that may not pay as much.

2. Tariff Negotiations & Free Trade

Yeah, perhaps. But I think what it reminds me of is that it feels like there's news every hour. It's pretty exhausting, but we've got to lock in. There's a super-busy calendar ahead.

Why don't we start with just maybe updating where we are on tariff negotiations and revisiting, I think, some important arguments about free trade? Last time we were together, we really framed this by saying, on the one hand, it seemed there were 2 different sides in this administration.

On the one side, we had a more tactical, narrow view of tariffs—maybe the best consensus, I called it—which was a fully loaded gun, rarely discharged, with targeted tariffs to re-onshore critical national industries. On the other side of this, we had what I described as more the nuclear Navarro approach. This is where we generate trillions in tariff revenue: we tariff everybody. High structural tariffs. We replace the Internal Revenue Service with the External Revenue Service. This is much more structurally different—not tactical at all.

Well, where are we now? Markets are now down 15% during this period since “Liberation Day” on April 2. The markets are clearly worried about a global trade war. We now have export bans. China implemented an export ban on rare earths on April 5. We retaliated—or at least implemented an export ban of our own—last week on NVIDIA chips. We'll get into it, and all of these bans and tariffs are quite controversial even among Republicans.

Remember, the Republican Party on Capitol Hill has historically been against tariffs. I think there's increasing concern on Capitol Hill about where all of this is headed and what it means for the economy. Many people are saying this is feeling quite chaotic. They're nervous about the 10-year. They're nervous about the impact on the dollar. Bessent's saying it's going to be a slog and this is going to take a while. But the president says on Friday he thinks we'll have it all resolved in the next 3 to 4 weeks.

So what is your reaction to all of this? What I want to do is take it up a level, because I think there's an important conversation. Republicans have historically been a party of free trade and lower taxes, right? I think of Reagan or Milton Friedman. And yet today there seems to be this growing consensus around tariffs.

So let's just go back: make the case for us. Steelman the case for why this may be a misadventure in the first place.

Brad, when we were talking at the beginning of this podcast journey about skilled immigration, we played a short clip from Ronald Reagan, which was one of his last speeches that he gave in office. This week, with all this tariff discussion, there was another Reagan clip floating around that I think does just as great a job of summarizing something in a very efficient way. So, if you don't mind, let's play that clip real quick.

[Ronald Reagan clip]

And today, many economic analysts and historians argue that high tariff legislation passed back in that period, called the Smoot-Hawley Tariff, greatly deepened the Depression and prevented economic recovery. You see, at first, when someone says, “Let's impose tariffs on foreign imports,” it looks like they're doing the patriotic thing by protecting American products and jobs. And sometimes, for a short while, it works, but only for a short time.

What eventually occurs is, first, homegrown industries start relying on government protection in the form of high tariffs. They stop competing and stop making the innovative management and technological changes they need to succeed in world markets. And then, while all this is going on, something even worse occurs.

High tariffs inevitably lead to retaliation by foreign countries and the triggering of fierce trade wars. The result is more and more tariffs, higher and higher trade barriers, and less and less competition. So soon, because of the prices made artificially high by tariffs that subsidize inefficiency and poor management, people stop buying. Then the worst happens.

Markets shrink and collapse. Businesses and industries shut down, and millions of people lose their jobs. The memory of all this occurring back in the ’30s made me determined, when I came to Washington, to spare the American people the protectionist legislation that destroys prosperity.

For me, everything I learned up until this point in time is congruent with what Reagan said. Tariffs lead to increased inflation. They lead to reduced innovation. Putting a protective blanket over our companies and our country leads to them being less globally competitive. And then, fourth, which he mentions, it leads to retaliation.

I’ve just been revisiting a lot of my priors. One of my favorite economists, Ricardo Hausmann, who spends time down at Santa Fe, put out a great piece that said, “We’re not adding up everything the right way. We’re only looking at physical goods, services, intellectual property, and net income that comes from companies like Meta operating around the globe.” Niall Ferguson did an incredible hour with Bari Weiss that I would just tell everyone to go listen to.

Every one of these things takes me back to what I learned over 20 or 30 years: These things aren’t going to work out in the long run. I’m skeptical. I’m always open-minded to revisiting my priors and to someone putting out an argument for why this is all going to work. But you and I were talking earlier, and I find there’s a lot of conflation going on about different reasons for why different actions are happening: the tactical idea that we need to have these specific technologies be resilient so we’re not overly dependent on 1 country, versus this bigger idea that we’re going to replace all of tax income. I think you could do a little bit of the first. The latter just violates all my priors.

I’d have to start fresh, and I’d have to be convinced that all these great thinkers who came before us were wrong. If our friend David Sacks were here—and I think he brought you up on the All-In Pod last week—he would say, “Okay, free trade is all well and good.” Very Jeffrey Sachs. But he’d bring up John Mearsheimer, and he would say, “The reality is we’re in a great power struggle. We have a peer rival in China today, and national security must take priority over free trade.” He might make the argument, I think, that creating resilience in these industries—rare earths, chip fabs in the United States, auto manufacturing, steel, aluminum, et cetera—is essential to our national security, and so therefore it’s not inconsistent with free trade. What would you say to that?

Well, we talked about this in the past. I think, to the extent you’re doing something very tactical, it could be proposed as such. The tone and the bravado of the description of what we’re doing aren’t consistent with that.

There’s also this: If you’re using the word “resilient,” that can include other countries that are allies. It doesn’t have to be on the shores of the U.S. I know people who, in the first Trump administration, were told to diversify away from China, and they built supply chains in Vietnam. Now they’re worried about tariffs there. So there needs to be consistency in order for people to adjust to those things.

One more piece I would mention—and we’ll put links to all these things—is Thomas Friedman’s op-ed in The New York Times that says, “I Just Saw the Future. And It Was Not in America.” He takes it further. He believes that, in advanced manufacturing, we’re not just not doing it here, but we’re not on the cutting edge of what it would take to do here.

If you recall, back in a day and age when Japanese manufacturing and just-in-time were this huge thing, all of America was convinced Japan was ahead of us in auto manufacturing. He says we need to adopt that philosophy about manufacturing in China. He goes to the point of saying, “Look, we need to force them to do JVs here so we can learn from them.”

Yes. Now no one’s talking that way in the administration because there’s so much vilification. There’s no recognition of where they might be better than us. But those problems, I guess the point I’m making, are worth solving.

I’m still not convinced. I think we have a huge labor problem when trying to be competitive in manufacturing. Niall makes a great point about the argument that we shouldn’t have done this and that we regret global trade. That argument takes you to a place where you’re allowing yourself to imagine taking where we are today and competing with an imaginary version of how things evolved if we didn’t embrace global trade, and assuming that we would have manufacturing and be in a better place.

I mean, I might also argue, in response to Mearsheimer, that it was an incredible debate last year at the All-In Summit between Mearsheimer and Jeffrey Sachs. He comes out of the realist school of political international relations, and I think there’s a lot of truth to it. You can’t stick your head in the sand and pretend that we don’t operate in an international system.

In the middle of the Cold War, in the 1970s, we had all sorts of embargoes on Soviet Russia that were probably smart embargoes. A lot of people think that the historical telling is that what ultimately brought the wall down against Soviet Russia was that our economy was triumphant.

And so I look at the situation today, and I think 1 part of the problem is that it’s not clear from the White House what the goal and objective is. If the goal and objective is Navarro, then I think there’s broad opposition within not only the Democratic Party but also the Republican Party to the scale of tariffs and disruption that’s going to have on the economy.

I think it’s fairly noncontroversial that we should have tactical, narrowly tailored tariffs and tax subsidies and incentives in order to re-onshore some critical capabilities. I mentioned a few: chip fabs, some pharma, and critical rare-earth production. We don’t want to be dependent on the Chinese for whether or not we can produce an automobile or whether or not we can produce a phone.

Some of that makes a lot of sense to me. What doesn’t make a lot of sense to me is that we launched what was kind of this nuclear-style tariff war against every country on the planet. The sequencing of events here is confusing. It should not be this McKinley-style, multitrillion-dollar tariff regime if your objective is principally to re-onshore some of these critical industries.

What I think is missing from this is that all we’re talking about is tariffs. We need to be talking about what the accelerationist agenda is that we’re going to put together in order to get these chip fabs. There are a lot of things we could put together: AI funds, tax incentives, and other packages to get people to build here. I think that gets lost in the conversation.

If we think that this alternative path is all about this re-onshoring bill, is that consistent, do you think, with the Reagan and Milton Friedman view of free trade? If it was more tactical like that?

If it were more tactical, I’m probably going to push back on you and disagree. I don’t think our current democracy is set up in a way that’s really good for the country to deploy dollars to help industry. I gave this speech at the original All-In Summit about regulatory capture.

I just can’t imagine that. First of all, the AI boom in the U.S. has attracted more capital dollars than I’ve ever seen accumulated in the history of mankind, so why sprinkling a little more on would be helpful? You also get into this game of picking winners and losers.

Now, Friedman highlights—and I think this is true—that the Chinese government is just much more successful at state involvement in innovation. If you go back to the space race or the Manhattan Project, we were capable of mustering forces in that way, but I haven’t seen it done recently, at least not in the past 40 or 50 years.

In many ways, the private sector has done a more efficient job of taking on the Manhattan Project-style projects. You look at what’s going on in Abilene, Texas, and Denton, Texas, around building out incredible supercomputer clusters, all done really without government involvement. I think that’s an efficient and preferred mechanism where there’s a lot of capital available.

But there are areas where I don’t think we will move forward much without government involvement. A great example would be energy. As we talked about many episodes ago, Korea and China are delivering nuclear fission at one-fourth the price we are.

I think the Korea example—South Korea—is very important because that is a democracy. Here’s 1 country that’s authoritarian, 1 that’s a democracy, and they’re both achieving it. We’re 4× over that. That’s going to require some type of government involvement to figure out why we can’t build that better and faster.

And to be fair, there’s a lot of movement around the country, mostly at the state level, about removing red tape, something governors celebrate, which is helpful.

That may move us in this direction. If we're going to help something along, I think figuring out why our governments have become friction and mud and changing that is probably going to be more impactful than the government doing the work itself or picking winners with capital. We're quick to blame others around the world for our challenges, right? But I think we need to spend at least as much time accelerating ourselves.

Part of that acceleration is getting rid of needless regulation, looking at our own systems, and asking that question: Why does it cost 4x more to build a fission reactor today in the United States than it does in South Korea?

Yeah, totally. By the way, one thing you just spurred in my brain: Thomas Friedman also did a podcast with Ezra Klein, and I'm just going to steal his story. I don't know that it's true, but he says that when China targets a new industry these days, they encourage a ton of startups. I didn't know this. He said that when they went after solar power, they encouraged, like, 100 startups.

So rather than pick a winner, they give money to everybody, in a thousand flowers bloom kind of way, and those 100 compete their way down to 5. He said there's a lot of fraud and grift and waste among the losers, but what you get out of the 5 is something that's actually honed by market forces. I don't think anyone who thinks about authoritarian China thinks about the state using market forces to get to a higher level of efficiency and production. But if what Thomas said is correct, they're actually using this kind of competitive market force to hone the better players in these industries.

Let's talk a little bit about tariffs and export controls. I think a lot of things get conflated when we talk about tariffs, and one of the most important is export controls. An export control is effectively a license requirement that acts almost as an embargo on all trade in a particular good. On April 5, China implemented this rare-earth ban, and then we reciprocated last week with the ban on all NVIDIA H20 chips going to China.

3. Export Controls: AI War, H20 & Rare Earth Ban

When you look at the rare-earth ban, I have a list here of the materials that were banned on April 5. You have critical elements that go into magnets—samarium, terbium, and so on. About 60% of these materials are mined in China. You do have other places on the planet that are mining these rare earths, but the problem is that 90% of them are refined in China.

These are the magnets that go into EVs, phones, computers, and machines of all sorts. This was called by some in and around the White House a kill shot—a kill shot against the U.S. economy. That causes massive problems in industrial production. When I think about, again, the destabilization that comes from overdependence on something like this—if every magnet that goes into every machine in the United States is coming from China and now we've effectively banned it, even though the ban was in response to the tariffs by the United States, right?—what is the off-ramp to this?

I now see, just this afternoon while we started recording the pod, that Trump has said he doesn't expect the tariffs to stay anywhere close to where they are on China. He doesn't think they're going to be zero, but they won't be anywhere close to where they are. We see a kind of unilateral walking down of these tariffs, but these export bans are really dangerous, Bill. What are your thoughts about an export ban, perhaps just compared to the tariff itself?

Well, I mean, I go back to the Reagan video, right? He says this will eventually lead to retaliation and escalation, and the fact that this video was recorded in, whatever, the '80s—what is that, 45 years ago?

Yes.

This isn't a new perspective. You could argue, "Oh my God, I can't believe this happened." Or you could say, "Well, if you studied history, you should have expected this to happen."

In fact, there's this Milton Friedman video going around, and he talks about retaliatory tariffs. He says, "Okay, let's say that somebody shoots a hole in the bottom of your boat, right?" It's taking on water. A retaliatory tariff would be like you shooting another hole in the bottom of the boat, and then they shoot a hole in the bottom. It just sinks faster.

Yeah.

Right. And he's saying—but in fact, what we did last week is the U.S. did ban H20s going to China. This was a pretty controversial move. I was looking at some of the analysis that was done this week, and the newly launched Huawei CloudMatrix 384 is China's competitor to the GB200. SemiAnalysis is out with an assessment of this and many others.

Basically, what they concluded is that Huawei is already on the frontier, or near the frontier, with respect to chips. In fact, one of the analysts on CNBC this week said, "To be honest, we basically just handed the Chinese AI market to Huawei anyway." Bernstein analysts said that what it effectively does is clear the decks for Huawei to have monopoly profits in China. They no longer have to compete with CUDA and NVIDIA in China, even though we were sending them the H20.

The H20 was a nerfed chip, a chip that was 2 or 3 generations behind what we have in this country. Some people think it was in retaliation to the rare earths. Other people think it's just part of the AI strategy to try to win the AI war against China.

Well, let's just take the argument that it was a retaliatory move. How good must it feel to be NVIDIA and Jensen and the executive team there, knowing that you have become a pawn in these games that you had nothing to do with creating? In fact, you designed this product specifically to meet the export-control rule that was in place before this one, which is very similar to this Vietnam thing that we were just talking about. So you can't be happy with that.

I reposted, when this all happened, a great video that Deirdre Bosa did over at CNBC about how this will be good for Huawei. I think if you ask Jensen, he would say, "Yeah, this is great for Huawei." It's not just that you get an isolated market where every participant in the AI market in China may have been working with NVIDIA. Now they start working with this, and they all collectively help improve it over time.

I read one article that said that China's getting some of its very first commercial planes off the ground. One of the reasons is that they became reliant on Boeing and Airbus, and so by denying your exports to a market, you increase the incentive they have to develop their own technology.

Well, particularly when they have the capability that China has. I mean, the fact of the matter is, if they had no capability to develop it, it would be one thing, right? I think more like the race around nuclear arsenals, if you will. There are some countries that will never have the capacity to build a nuclear arsenal, but the fact of the matter is China is so close in this regard with respect to the H20s.

From my perspective, it's a very close call. I don't think we should be sending GB200s or cutting-edge chips to China. I think there's a legitimate national interest in, quote unquote, staying out in front on AI. I think there's a debate to be had about that. That's where I come down on that side of the issue.

But banning the H20s seemed to me a very self-defeating move. I think the competition with CUDA in China distracted the competition and denied them monopoly profits within China for these Chinese internet companies. There was some insinuation last week that Jensen was intentionally skirting all of these rules around China and that maybe 30% or 40% of all their sales were going into China.

As I look at this, I think we have to be really careful. These tariffs, all of a sudden, can find us in a place where we're demonizing NVIDIA. I think NVIDIA and Jensen are national heroes for the work that they've done and the leadership they've given us in global AI—American success stories.

Totally agree. And on this idea that somehow they comply with the exact attributes of the H20 prescribed by the U.S. government, I want to go back to that. They sell those in China, and then after the fact we go back and somehow start demonizing them. I think that's a very dangerous thing.

So when I gave that speech a while back on regulatory capture, I made the argument that when, inside of America, we want to accomplish something through policy, we gather a bunch of people, get a bunch of ideas, and then write some complex piece of legislation. What I highlighted is that in many cases it fails, and in some cases it spectacularly fails.

So what does that mean? You get the exact opposite outcome from what the policy intended to create. The fact that this happens all the time inside of America is an interesting fact because, in some ways, export control is external regulation. We're trying to implement a policy now not just in a place where we control most of the pieces, but in a world where we don't control many of the pieces. And then we expect that policy to work perfectly.

The problem is, borders are porous and money finds a home. I always tell people who try to build internet businesses: Internet arbitrage is undefeated. Global markets kind of have the same thing.

We're just coming off a remarkable—I would use the word "spectacular"—failure of Russian economic sanctions, right? So why do we think, "Oh, that didn't work, but I'm going to go do this and make it work"?

I heard people who said we have to ban H20 saying, "Oh, well, the Biden rules didn't work. They failed." So you mean global export controls didn't work, so we're going to redouble them and see if they work again?

And I just think if there were a piece of technology, like, say, a fighter jet that you're giving to maybe 1 or 2 countries that are super-deep allies, that's 1 thing. If you're going to sell something in 120 countries around the world and try to tell 1 country they can't have it, there's no effing way that's going to work. Especially when these goods have digital attributes. They're just going to work around it, and I think it's going to get worse before it gets better.

I've seen hints that the same people who push the H20 want to put even more restrictions on ASML, which is a Dutch company, or fine TSMC for building Huawei chips. That's a company that's headquartered in Taiwan. I'm sure Jensen's miffed that his company's become a pawn in this game.

How do you think these companies respond? We talked about ASML in the past, where the management there has already expressed some reluctance to be our whipping post or whatever. I don't know that we're just allowed to start grabbing pieces of the chessboard that we don't even own and bringing them into this competition. And you said earlier, the way we implemented tariffs, we've alienated many of our allies and partners. I won't be surprised at all if we wake up one day and one of these companies tells us no.

Yeah. Again, it comes back to defining the objectives narrowly enough that you don't end up in this big retaliatory and escalatory posture. Again, I think it's a reasonably close call around H20s. My sense is that NVIDIA understands and is on board with not selling leading-edge chips to China, right? They didn't fight the fact that they weren't selling GB200s to China.

And the H20 is—if you just looked at the facts on the table, the Huawei Ascend 920s, the number of these that they're now producing, and the way that they're interconnecting these things together—yes, they may be slightly less efficient. It may take 2 or 3 times the number of chips and amount of power in order to end up at the exact same place. But we just talked about how there are 100 nuclear power plants under construction in China. Power is not their limit.

Well, and they produce those plants at 1/4 the price we do. So if it comes down to power, and many people in the AI world like to say it's going to come down to power, well, I guess they're going to win.

Well, I mean, so there's an interesting question here. If you think about the off-ramp on these for President Trump, again, there's a lot of suggestions over the weekend that if you think we're going to land the plane in 3 to 4 weeks, Bill, the only way you do that is basically through unilateral concessions that you define as victory. So all of these countries are calling us. They want to do deals.

And so we're going to walk the Chinese tariffs back to 30% or 34%, and we're going to suspend, maybe for 90 days, the H20 ban in the hopes that they will reciprocate by suspending the rare-earth ban or something like that. So I don't think we've seen the end of this when it comes to this NVIDIA ban.

But the flip side of this is NVIDIA just had a big announcement with the White House about a $500 billion investment, in combination with a bunch of other companies—Foxconn, TSMC, et cetera—building fabs in the United States. I know the first GB200 has rolled off the assembly line out of those fabs in Arizona.

To me, again, we need to rebalance and focus on accelerating and re-onshoring these critical industries in the United States. I think the distraction of trying to keep China away from frontier-level AI will be a losing strategy in my mind. I don't think we need to go out of our way to make it easy on them. But I think the obsession, the time spent, and the distraction for U.S. CEOs to try to figure out how they comply with all of this is not particularly great.

I'll just give you where maybe the next frontier is going. I've heard increasing chatter that somehow we're going to try to ban DeepSeek models.

Yeah, right. And somehow they want to prevent U.S. hyperscalers or cloud companies from integrating DeepSeek models into their platforms. I heard the same thing. So talk to me about that and the slippery slope associated with that.

Before I go deep in on DeepSeek, which I want to do, I want to mention a couple of things from a very broad level. One is, back to this Friedman post, he mentions a quote from a book: “How we do anything means everything.” And the quote is, “Interdependence is no longer our choice; it's our condition. Our only choice is whether we forge healthy interdependencies and rise together, or maintain unhealthy interdependencies and fall together.”

There's a book that I've been rereading that the Collisons talked about once called Finite and Infinite Games by James Carse. There's a reality that a lot of the strategy that we think about, and how we compete in the world, comes from finite games. A finite game is a game that has a beginning and an end and a winner. So, a football game, a basketball game, that kind of thing.

Infinite games are different. They just go on forever. The stock market and most large companies are infinite games. They're not—there's no clock that ends. There's no one that declares a winner.

I fear that a lot of the attitude, the vitriol, the language, and the tone that's used between us and China is borrowed from finite-game thinking. It's like we're supposed to win the AI war. What does that mean? Is there someone that's—is the clock going to end, like the Warriors versus the Rockets, and they say, “Oh, we won”? There's no chance of that.

And most of the people that are even the biggest China hawks, I say to them, “Do you think you're going to prohibit China's long-term AI development?” And they say, “No.” I just think there's a lot of zero-sum thinking going on—win-lose. I totally believe in this interdependence.

I think there's zero chance that you're going to stop China or somehow increase the gap. There's a lot of arguments that the AI gap has been shrinking, not rising. Eric Schmidt had a big interview this week where he said it was shrinking. We'll put show notes on that. A lot of people are saying they're catching up. That SemiAnalysis thing says Huawei is getting there.

And so this argument that, “Oh, we've got to do everything we can. We have to act now to win the AI war,” is almost this belief that we're going to increase the gap and that it's a finite game that's just going to end one day. I don't believe in either of those things. And so I think the tone's wrong. I think you've got to get out of this enemy-threat framing. All those words allow you to hate.

Yeah, one of my favorite quotes ever.

The whole piece of this Friedman article, which I think everyone has to read—he just went over there and spent time—is that it's not what you think. They're way ahead in a lot of different areas. They're super smart. I think a lot of people thought, “Oh, we have a democracy and capitalism,” and people get confused between those 2. They think we've got the perfect American exceptionalism. We're the only place where you can have innovation.

And because we talk about, “Oh, they steal, they steal,” we don't ever spend the time to go see if innovation's happening over there. It is happening over there. These people are educated in the same schools and universities. The thought that they couldn't be educated or can't innovate is going to lead you to a lot of really bad policies.

And worse yet, if you believe they're on the frontier and they're going to have a frontier compute stack with frontier models, whether we help them or not—which I think even those people who are more China hawkish agree with that statement—and if you wanted the rest of the world to run on U.S. compute and U.S. models, what's the number 1 thing you would not do? You would not anger the rest of the world by launching a tariff assault on them simultaneous with your tariff assault on China.

And so it is a mysterious policy. When you look at all of these headlines that are coming out, the EU is now negotiating directly with China. We now have Japan, South Korea, and Vietnam negotiating with China. You have the Middle East entering into deals with China. I'm not sure, again, that we're advancing our cause by making it harder on everybody else in the world and telling them they are the cause for the demise of the U.S. middle class.

Well, and I'll go even further. The tone they're adopting is more statesmanlike than us. If you read the tone in those discussions, they highlight that they haven't invaded anybody in a very long time. The U.S. has been involved in a lot more wars around the globe.

So our ability—and I brought this up a few podcasts ago—I think you're right. You mentioned DeepSeek. So we all believe that there's an imminent DeepSeek ban, the level of which I don't think we fully know right now. But my immediate reaction is, “Oh, I guess Europe and South America will run on DeepSeek,” because the minute we start vilifying even that, it is, I believe, right now, the best-performing, most-open model. If you had a quadrant, it's the one that's on the upper right on performance and openness. And if we ban it, boy, it's just going to incentivize even more people working with Huawei chips and DeepSeek, trying to optimize it.

But two, in this world we've created where there's so much animosity, I think people would be glad to use it in these other countries.

4. New AI Cold War & Zero Sum Game

Well, the first thing it dawns on me is that I ultimately don't think—somebody asked me on CNBC, and I said your decision as to whether to be in the market comes down to whether you think the president is more of a tactical free trader, in the spirit of a fully loaded gun, rarely discharged, or whether you really think he wants to reorient the world in the way Navarro discusses it.

I still believe the president is fundamentally a free trader, that this really is about making the US more resilient. He has to bring us back there, because if we head in the direction of Navarro, it's quite clear to me that we're going to so anger the rest of the world that the chance the rest of the world is going to run on US compute is greatly reduced. I think what we do in that instance is move everybody in the direction of China. It backfires. We don't widen the gap with China with respect to AI or global trade or anything else. In fact, we shrink the gap.

[Speaker?]

By the way, there's one hilarious irony with the Navarro point of view combined with export controls: if we're upset that there's a trade imbalance, the worst thing, the last thing you'd want to do is start penalizing our best creators, our best manufacturers, and our best companies. How are you going to solve the trade imbalance if we tell people they can't buy our best stuff? We're going to make them buy our shitty stuff. Just as an example, I think NVIDIA sold $12 or $15 billion worth of H20s. If you ban those, then your trade deficit with China goes up by $15 billion.

My very point. Let's not wash over the fact that I think NVIDIA is such a high-performing company, worth so much money, with such a large market cap, so much cash, and no liquidity issues, that this $5 billion write-off they're going to take is kind of NBD. Move on. It didn't affect the stock that much, but a $5 billion write-off—they're just going to throw a bunch of inventory in the trash can.

I mean, that's remarkable. Think about that, Bill, right? Some of the people who are arguing—including some US AI companies—that they shouldn't be able to sell the H20s to China. If Jensen or the team at NVIDIA turned around and said, “Great, you buy them,” nobody would buy them because they're so underperforming, right? And yet it makes the point for them.

But think about this: $15 billion in sales of H20s probably results in $8 or $9 billion in earnings back to NVIDIA. You tax that at a corporate rate, and that's $2 or $3 billion to the US Treasury. We are effectively taxing the Chinese on a nerfed product to provide profits to NVIDIA, to further distance the lead against Huawei, and to put money in the US Treasury. And we just unilaterally disarmed around that in a way that doesn't even advantage us against Huawei, because their chips are already surpassing the H20.

And that leaves out—and I agree with 100% of that—there's even more because of escalation. So, once again, I go back to this finite-game mindset. I think people are evaluating a decision on an export control, a tariff, or a ban totally in isolation, as that one decision and what its impact will be. But any one of these decisions, and certainly the combination of them, could lead to escalation that could go as far as leading to a hot war. You have to accept responsibility for making any one of these decisions that there are butterfly effects and escalatory effects that you may wade into by doing that.

You can think about it this way: Some people define artificial intelligence as existential, that it's absolutely essential to your national security and your national economic security. Now, if you are firing a shot at somebody that says, “I'm going to deprive you of being able to build your national security or your national economic security,” it's a serious thing.

I want to go on record as saying almost everyone in the AI space that I see with a microphone in front of them says the US has to win the AI war.

Yeah.

I don't know what that means. If I guess as to what it means, I don't think it's possible, right? Because it's an infinite game. It's an infinite game. No one that I talk to would argue that we're going to somehow prohibit them from moving forward in AI.

Let's not forget, it wasn't that long ago that OpenAI—they'll say—borrowed or copied the innovation that happened at Google and DeepSeek. This is how innovation worked. Ideas spread like the wind. They move fast. People study what other people do. It happens all the time. Only when we apply these labels of theft, copy, or steal do we start vilifying it, but it happens all the time in our own ecosystem. It's how these things evolve.

So, yeah, I don't think there's a finite game to win. You know, it's funny: I did some research on something because I was very curious. I went back and studied the space race. When the space race was underway, if you go back and look at the quotes from all the senators, President JFK, and everybody else, one of the reasons—maybe the primary reason—they said we had to win the space race is that they all believed the entire globe would be covered in this mesh of rockets and satellites that would be used to control the military of the globe, and whoever got to space first was going to have this control.

Now, that never played out. But I see a very similar vein in AI, where people—especially people who believe in AGI and ASI and magical AI—think it's just going to explode, and then whoever controls this uncontrollable force will control the globe in a zero-sum way. I just don't buy it. I don't think that's going to happen. If that does happen, we're going to have much bigger problems.

I think you and I would maybe find a point of agreement here. What is the north star that should guide us? I think the point of agreement, if we were giving advice, would be that the re-onshoring of critical national industries through some tactical means is an important objective.

I think you and I would also agree that we need to shift the focus radically toward accelerating our own race, and refocus from spending all of our time trying to slow down everybody else. The objective should be—the KPI should be—do we widen the gap? Are American companies as successful in AI as they were on the internet?

I think anybody who looks back at the last 20 years would say that the United States has won both in terms of free trade and in terms of the global internet competition—won as defined by improving the standard of living, driving the growth of the economy in the United States, in a way that was successful for the United States.

You know, in one argument with someone, I used the simple metaphor that comes from swimming. Anyone who becomes a competitive swimmer, within a year or two, a coach tells them, “Don't look to the side, because that's wasted energy and someone might pass you. Just swim as fast as you can. Just look ahead.” I feel like it's almost worse than just looking to the side. We're trying to intentionally inhibit the other player. Going back to that, since I mentioned you're trying to deflate the football—

Yeah. Yeah, exactly. Imagine if you were just watching a finite game and one team insisted that the other team have a technical disadvantage. What would your reaction be? How would you think about that player? You would probably assume they're scared, that they're unsure of themselves.

Also, if you told your team at the start of the season, “Don't worry. You don't have to train very hard because I've handicapped the other team. They don't even know what I've done to them, but I've tainted their system, so they're not going to be very good,” as opposed to just focusing on your own training and winning the game, that would be a very different mindset.

I might end by going back to this argument from the Friedman piece: that it's an interconnected world, and there are a lot of very smart, successful, capable, innovative engineers, researchers, and founders in China. If you think you're going to keep them down or permanently prohibit them from playing in AI, I think that mindset is going to lead to some spectacularly problematic policy.

Yeah, I agree. Well, maybe we can talk about Tesla and wrap up. Tesla reported tonight, and the markets are clearly struggling—down one day 4%, up the next day 4%, clearly struggling to get their hands around where we are.

Since we were here last time, the S&P is down 10%, and the Nasdaq, I think as of yesterday, was down 20% peak to trough. The question people ask me all the time is: Is this all priced in? How many units of risk do you have in the market? We have Bessent doing a talk tomorrow morning on the financial system. The VIX is still over 30.

5. TSLA & Market Check

There are fears and rumors out there about our 10-year—do we have demand for it on a global basis? The dollar is under assault here. Many of these conversations weren't happening 5 months ago, right? Exactly. We were just talking about AI and Nvidia.

I go back to this Stanley Druckenmiller quote. After the president's election, he said, “This is the biggest shift from an anti-business administration to a pro-business administration in my 50 years of being in the business.” That is how CEOs generally felt at the beginning of January, moving into this new presidency. It's really mind-boggling to think how, in a few short weeks, we went from that level of optimism to this level of concern.

And I was just looking at a chart: Apple's down 21%, Nvidia's down 25%, Google's down 20%, and Tesla's down 40%. Right? We have major moves in the market.

Well, both Tesla and Apple have a huge amount of revenue in China. And when you sense that this thing may be escalating, or that there are ripple effects going back and forth, one plausible thing that could happen is either one of them gets kicked out of the country.

Right. Right. That could happen.

Yes. And so, of course, you have to discount this. And the question now, of course, is: Is enough discounted, given where we are? And so, you know, I said at the end of January, early February, when you and I did the pod, I said, "We've taken down our units of risk because there's just more uncertainty in the world. Discount rates need to come up; multiples are coming down." And we did.

I would say that over the course of the past several weeks, we've had very little risk-on directional risk in the market. You can like Nvidia, but if you wake up one morning and $12 billion of its sales are gone because it can no longer sell chips in China, that's something, again, that is the macro that is hard for you to forecast on going into Nvidia.

I've heard rumors that they want to do a customer audit. I don't know what that is or involves. And, by the way, not to sound like a victory lap, but I said a few episodes ago that the biggest risk to Nvidia was the government. And that turned out to be true here. They may not be done.

Well, they were suggesting it last week, I think, even on our friends' pod, where maybe you should look into all these shipments or all these sales that are going to Singapore or whatever. And, again, I welcome all of that. I think all companies should have to comply with U.S. rules and regulations. I suspect that they do.

But, again, if I just step back here, gold's outperforming the S&P so far this year by about 40%. Right? What does that tell you? People are moving into safe havens. They don't know where this is all going to land.

Here's one of the things that really concerns me, Bill. If we look at consensus earnings expectations for the S&P so far this year, we started the year expecting 15% growth. We expected the S&P to do $273 a share. Now the consensus is 12% growth, so $265 a share. So consensus earnings expectations for the S&P have barely moved down.

When I think about the chaos of the last 8 weeks, every CEO that I talk to says, "We're on hold. We're tightening our belts. We're taking a more cautious approach to the year. We can't make decisions. We don't know where tariffs are going to be. We don't know where this is going to be." It's hard for me to think that you can literally wipe 2 months out of the year in terms of where decision-making is going on in these companies and only have a 3% adjustment to S&P earnings.

Now, what are we seeing over the course of the last 2 weeks? Scott Kirby from United Airlines comes out. He gives 2 different guides, and he says, "If there's a recession, which we may have, then earnings are going to be $7 to $8. And if there's not a recession, earnings will be $12." So you have companies that are doing highly unusual things in terms of saying, "We don't know the future." We're either not going to give you a guide at all, or we're going to give you this wide dispersion in a guide.

I think for most managers, people are looking at this and they're at max caution. From an Altimeter perspective, we were pretty cautious, and now we're looking at leaning back into markets on some of these down days where it's big risk-off, because I think ultimately the president is going to make deals. Bessent is going to get deals done.

I think we're probably going to land somewhere in this kind of 5% to 10% tariffs for the rest of the world. If you add that up, that's going to be somewhere in the order of magnitude of $200 billion of tariffs. And you're not going to end up at 140% on China. We're probably going to get rid of these export bans and embargoes and land the plane somewhere around 25% or 30%.

Mind you, that's still a 3- to 4-times increase over the tariffs from last year. But I think what the world needs is predictability, and they need it soon. I think if we're still having this conversation in 8, 12, or 16 weeks with this level of uncertainty around tariffs, it's going to have a much bigger headwind to those S&P earnings for the year. And as those earnings come down, we know that the market's going to have to come down with it.

So the S&P, as of the end of the day today, is only down about 10% for the year. And if you think about all the things that have changed, that feels to me like just the air coming out of the balloon in terms of the certainty and the uncertainty that we have in the market.

I think the next move down, if we have it, is going to be around growth. And, yeah, you really haven't got a lot of growth earnings. Tesla's out tonight, and even though earnings came in a lot lower than expected, I think people are looking through that as kind of one-time. But by the time we get another month or 2 into the next earnings season, if we continue to see this level of uncertainty and drifting down, I think we're going to have further problems in the market.

I admire your hopeful optimism and the data points that I see in the past 24 hours about how soon something might come together. I think there was an announcement that India had finalized the terms of reference. I don't know what that means, but it doesn't sound like you're very close to being done.

That was JD Vance today, in meetings with Modi. He came out of the meetings and said, "We had a big breakthrough. We came to terms on the terms of reference for what the trade deal will be," but it will still take months to hammer out.

So months—I think months is too long. I think reflexivity already starts to take place. Everything you just described, once it does, some of it is certainly self-reinforcing. You start decommitting capex. You start decommitting basic capacity on different manufacturing lines, planes, whatever. And it will become self-reinforcing. You raise prices when you decommit because you have fewer lanes. That's inflationary. Everything starts—the fear of it starts to become the reality, right?

I guess a thought experiment: If I described everything to you that has happened thus far this year and I asked you, "Should the market be higher or lower from the all-time highs where we started the year?" you would say, "Of course it should be lower," right? And I don't think that being down 10% on the S&P sounds irrational to me, given the level of uncertainty that we've injected in a very short period of time.

So, to me, if you're thinking about units of risk, it still feels to me like you're in that bottom third of whatever your normal exposures are. You're in that bottom third. Everybody wants to buy the dip and go all in, and I think we need a lot more certainty as to where this is going.

I do think that it would be very helpful, and it sounds like the president is starting to make some of these comments just this afternoon, to disabuse ourselves of this Navarro-style idea to replace the Internal Revenue Service. If we can just get a 4-year forecast that's about re-onshoring a few industries, and land the plane on China, or on Japan and India, and then on China, I think that gives us what we need to plan. But there are a lot of other things that have to come together in order to keep the market moving forward.

We didn't even get to talk about the impact this is going to have on startups and on what we're seeing in the startup ecosystem. We'll come back and do that next time.

But I think that, listen, I'm kind of tired of talking about tariffs, but it is the most important thing, and it is impacting everything that we're looking at, both on the public and on the venture side of the business today. So it's unavoidable.

We have this really important divergence, I think, in terms of points of view. And what's interesting to me is that the free-trade side of this argument has largely been drowned out. So I appreciate you steel-manning that side of it.

I mean, look, I assume most people that have at least had a finance class have studied comparative advantage. But it's very mathematical, it's very deterministic: doing stuff you're good at and letting other people do stuff they're good at is a win-win-win. You start backing that up, and you're going to get lose-lose-lose. I'm certain of it.

Yeah. Hear, hear. Let's land the plane and get back to building America.

Okay. Take care, man.