Trump V2: Tariffs, American Dynamism, Higher Ed | Byrne Hobart

Hey, Upstream listeners. Today, we're sharing my conversation with Byrne Hobart, recorded earlier this week. We explore the economic consequences of Trump's tariffs, examine OpenAI's latest innovations and their implications for AI's future, and discuss the potential transformation of white-collar work. Please enjoy.

Last week we didn't get to discuss the market and the economy shifting rapidly as a result of Trump's tariffs, even though we discussed them before at a high level. What do you make of the last couple weeks of market activity and what, if anything, should we learn from it in terms of what it means going forward?

Yeah, so I think that part of what we learned is that there's this weird— I think one of the first ways to look at this is to compare the first and second Trump administrations, because one of the things that was going on in the first Trump administration that between non-Trumpy Republicans and just Democrats, it was really hard for Trump to do a lot of the things that he kind of impulsively wanted to do. And there was also just not, not a ton of long-term planning from the Trumpy Republicans on exactly what they wanted to do. So there was just a natural level of ambient chaos throughout that first, first administration. And my assumption was that we'd see pretty much the same thing the second time, but it turns out that a lot of people spent the years from about 2016 2021 to early 2025, figuring out exactly what it is that they'd want to do. And so you had this spree of executive orders right at the beginning, and a lot of them were just hitting the high notes on things that Trump had always said he wanted to do and hadn't really gotten around to doing. And it kind of looked like, just from the neutral standpoint of, does this administration have its act together? Do they have coherent goals that they're actually trying to pursue in a way that makes sense? It kind of looked like they'd really scaled things up. And then the tariffs were one of the first cases where it was just this kind of last-minute sort of spontaneous decision. And one of the reasons we know that is that there were a couple blow-by-blow articles in different publications, New York Times, Wall Street Journal. I think Politico also had one that just talks through, here's how this process actually came about, that Trump had other options. He went with the most aggressive and kind of bluntest one. And You know, one of the things that stood out there was just this administration is suddenly as leaky as the first Trump administration. So in some ways it is like this high watermark for state capacity Trumpism in that he did actually implement this tariff thing that he'd been talking about for decades. And at another level, it was the last gasp of state capacity Trumpism because the way that some of the earlier Trump moves seemed to work was that some very smart heterodox right-wing person from the think tank world was coming up with exactly the plan that they would use on immigration or on, you know, how, how higher education is funded or whatever. And then they just implement it day one and it's a shock and awe thing. And then with tariffs, presumably there were such plans. They all apparently got torn up and discarded. And now, now the administration is trying to negotiate other deals, but what it looks like is that Trump does react when the market goes down, if it goes down enough, and he has enough people in his orbit who tell him this is actually a disaster and that the only reason the market's not down more is that everyone thinks that he'll back down. And one of the problems is this actually creates a really strong incentive for China in particular to do things that make the US equity market go down. So that's not good. We now have this situation where one country that is much more centralized, much more top-down, and has far weaker checks and balances is able to deploy its economic power in order to make things difficult for the US. And the US just doesn't have the same capacity to reciprocate, although there is a lot that we can do. So yeah, that, that kind of setup is just not good. And then, yeah, you look at the market and you kind of see roughly what you'd expect to see, given that there was general bad news expected and the specific bad news was worse than expected, where stocks go down, volatility goes up. And one of the things that you saw at that time, which there are a couple explanations for, is that Treasury rates went up. And so in general, since about 1998, it has been very reliable that when stocks are down, treasuries outperform. And that if you want to have a more balanced, a kind of safe and not perfectly cyclical portfolio, you have a mix of stocks and bonds. And usually one of those trades is winning when the other one isn't, which as we were kind of reminded of, like this was always theoretically true, but it was good to get a bracing reminder of this in 2022. That only works if inflation is low, because when inflation is low, that means that when the economy slows down, central banks just want to inject a lot of liquidity to speed things up. And that tends to push rates down. But when inflation is high, they can't automatically always do that. They have to consider whether they're actually pushing price levels up more than they're pushing growth up. So it, that weakens the correlation. And there is a level at which tariffs are just inflationary. They are a tax. They're a tax that shows up in terms of higher prices paid by consumers. So straightforwardly inflationary. If you offset that tax with a tax cut, you now have higher, higher consumer prices and a little bit less revenue for the federal government because of the deadweight loss. And so to keep U.S. standards of living flat, you actually do need to do some inflationary things. But then the other piece is that there's a narrative that the U.S. dollar will no longer be the reserve currency. I think that to the extent that that's true, it is just, it takes way, way longer than frankly than one president's term to actually put that into effect, but also that That doesn't seem to be the big driver of treasuries moving. We'll have to see as different data from different countries comes out, but it doesn't seem like there are that many countries that are just mass panic liquidating treasuries because again, they seem to think, and they're probably right, that a lot of this tariff stuff will be reversed and that they shouldn't make these long-term changes to their plans just based on short-term tariff considerations. But one of the things to consider is that if Trump gets his way, Then countries that were running a trade surplus with the U.S. will have to buy American-made goods to offset that. And what they bought instead was treasuries. And so even if, even if some, if Trump does actually negotiate the deal of his dreams with all of these countries and they're all buying American gas turbines and medical equipment and planes and weapons and LNG and soybeans and whatever else, that means they're not buying our treasury bonds. So in that case, treasuries should go down. And also if Trump just does, does a tariff and offsets some of the impact with a tax, that's also inflationary. So treasuries go down. It's only like the scenario where treasuries actually outperform right now is one where tariffs are mostly or completely unwound. We do mostly return to the status quo, but we return to the status quo having gone through this very disruptive period where markets went down and a lot of capital expenditure plans, both in the corporate sector and the private sector, were deferred. And everyone's a bit more nervous, so they're saving a bit more money. In other words, if there's a recession. But it's kind of a, it's a sort of separation of powers recession where one way it would happen, because I don't think it's that likely that Trump just completely blinks and says, yeah, I, I accidentally made a series of gigantic mistakes over the last several weeks and they're all my fault. The only, the main way that that happens is through Congress or the courts, which means that in the end, treasuries are actually this really interesting bet on the, the future constitutional structure of the U.S. So if you are a believer in a strong executive branch, for better or for worse, Then you think that Treasury yields go up. And if you are a believer in separation of powers, then you probably expect a recession, but it's kind of a dodged a bullet recession and you expect Treasury yields to go down. So that's where we are on that side. Another piece of the Treasury market though is that one of the big trades that a number of large hedge funds do is the Treasury basis trade, which is basically betting on the gap between Treasury futures And the actual bonds themselves. And in some ways, this is just the simplest possible trade because you can look at the value of treasuries implied by futures. You look at the value of treasuries that you actually buy. You look at how much you have to pay to finance them, how much capital you have to put up, and you can calculate very straightforwardly what your return is on buy the treasury, sell the future, and then either exit that trade at some point if the prices converge or just when the futures reach their expiration date. You have either rolled onto the next set of futures or you just exit the trade. Like you deliver your futures to close out the short future, or you deliver your bond to close out the short futures position. And this trade, it is kind of like a, at one level, really, really simple trade. At another level, you can do some more complicated things with it because not all bonds are created equal and there are different times when you might adjust your position size, et cetera. So it's like there's room to do interesting stuff there, but The main way that it seems to work is that if you are a gigantic hedge fund and you do a lot of business with every major bank and you are constantly negotiating the terms on which you borrow from them, what you would often want to do is be able to borrow larger amounts of money than you really need and on better terms than you really need, you know, have a lot of essentially dry powder. And if you have that dry powder, why not put it into a very, a trade that can be very straightforward? like the basis trade and just take advantage of that. So some of what could be happening is that some of the funds that have been doing that, they have other ways they can use their liquidity. So they are gradually reducing their exposure to that trade. I don't know for sure, but it is a plausible explanation for how rates are moving right now. So that is another piece. And then we can finally move to what everyone means when they talk about the market, which is the stock market. So stocks had some of their, some some of their worst 3-day stretches in a very long time, their worst 1-day performances in a really long time. And I thought the most interesting thing there was, you know, anytime any kind of asset moves a lot, people will make jokes about how the market's not efficient. And these jokes have varying levels of seriousness. And I think pretty much anyone who's discussing market efficiency in a serious way is discussing it as this sort of directionally true thing, but not as this completely ironclad principle where every price is always and everywhere the exact value of the asset in question, because, you know, you can always just create any fake trade and immediately prove that that's not true. Like you can log into your brokerage account and look at some thinly traded company, wait till the market is closed. So it has a wide bid-ask spread. And then you could, if you felt like it, buy 100 shares at some price that is wildly off the market. And then obviously that price is not the efficient market price like that. The valuation for that company based on your 100-share trade is not the true worth of that company. So it's never strictly true, but I think it is actually worth talking about what market efficiency means in different dimensions when you look at a really volatile period like this. And especially when you look at things like market rallying on the rumor of a 90-day pause and then market ripping on the fact of a 90-day pause and then market being choppy as that pause kind of gets squished around a little bit in terms of what it actually means and to whom it applies. And as people also digest the possibility that the US will not in fact be able to negotiate dozens of bilateral trade agreements in a 3-month period, given that those agreements typically take many years of negotiation to actually get right instead. And so the way that I look at that volatility, and especially the volatility around things like there's a rumor, stocks rip, you know, the value of the S&P rises by several trillion dollars and immediately comes back down. The way I'd look at that is that when the market is volatile, the price of liquidity is higher. And one of the measures of the price of liquidity is how much do stocks move when you buy a given amount? So in a market that is efficient with respect to prices, you would not actually expect these huge swings in volatility, or you'd expect smaller swings in volatility because the world is more uncertain, but it's not, it's not, you know, trillions of dollars of intraday fluctuations more uncertain, just more uncertain than usual. But if you think of volatility as something for which there's a market and that volatility, or if you think of, sorry, liquidity as something for which there is a market and that liquidity is expensive when the situation's really uncertain and when whoever takes the other side of your trade knows that there could be some huge news story that they got a fraction of a second too late and that they turned out to have made a bad trade. In a case like that, stocks should actually move pretty massively on fairly small, fairly insignificant news flow. So I wrote a couple pieces kind of touching on that from different angles over the last week or two, including just one look at kind of why do markets crash a lot faster than they rise? Because if you go back and look at the best one-day update, the best single-day updates in the S&P 500's history, they are not good times. They're not times that you associate with it was a great time to own stocks. Like a lot of the best days in the S&P's history are in the 1930s. Early 1930s when everything was falling apart. Some of them are in spring of 2020. Again, things were falling apart somewhere in 2008. And then you have one, I think, from 1987, a couple days after the crash. So pretty much anytime someone is fully invested on one of the best days in S&P history, it's because they were fully invested on one of the worst days in S&P history too. And that's just, that is another reflection of volatility gets really, really high. And that means that whereas like If you look at periods where the market just generally did well, like the 1950s, you don't have a lot of crazy days, in part because just the flow of news during good, during a good economy is things are always just slightly like things were always looking good and they're always looking slightly better. Like the, the big positive news in 1955 will be something like Ford made a slightly larger, slightly more expensive car. And fortunately gas prices are extremely stable and therefore people will buy this slightly more expensive car. So this is slightly good news, but it's It's not world-changingly good. The world-changing news is almost always really bad news, especially for large publicly traded companies. So if you look at something like AI, it is good news for just the economy and for human flourishing in general that we can offload a lot of intelligence to machines. It's also really, really bad news for any company where they have like this implicit balance sheet position of we have a lot of people together, we have built all these tools for coordinating them and getting them to do fairly low value added white collar work. And now that work can be done orders of magnitude more cheaply and more quickly, and we can check the work more easily, et cetera. So it often means like it's good news for the world. And then the profits of the public companies that are a large proportion of the market are sometimes impaired. We're actually just very fortunate that this is a case where there's a labor saving technology that is primarily being built and deployed by very large cap companies that are very profitable and that are able to convince their investors that this will be profitable. You could totally imagine a case where there is an AI-driven bear market because we just have a slightly different configuration of what companies are large-cap companies and what companies aren't.

We'll continue our interview in a moment after a word from our sponsors. There's a growing expense eating into your company's profits. It's your cloud computing bill. You may have gotten a deal to start, but now the spend is sky-high and increasing every year. What if you could cut your cloud bill in half and improve performance at the same time? Well, if you act by May 31st, Oracle Cloud Infrastructure can help you do just that. OCI is the next-generation cloud designed for every workload, where you can run any application, including any AI projects, faster and more securely for less. In fact, Oracle has a special promotion where you can cut your cloud bill in half when you switch to OCI. The savings are real. On average, OCI costs 50% less for compute, 70% less for storage, and 80% less for networking. Join Modal, Skydance Animation, and today's innovative AI tech companies who upgraded to OCI and saved. Offer only for new US customers with a minimum financial commitment. See if you qualify for half off at oracle.com/turpentine. That's oracle.com/turpentine.

It's fascinating. There seems to be some, you know, as a euphemism, you know, confused messaging over whether the goal is to improve our negotiating position in order to have more free trade. Actually, you know, tariffs in support of having other countries block their preexisting tariffs, which is a different argument than saying, no, we actually need these tariffs to last because we either need the revenue or we need to reindustrialize. And so supporters of the sort of policies are basically explaining both arguments, even though they seem to be somewhat contradictory. And that's perhaps— I'm assuming that there are people in Trump's camp, and you alluded to your link to a Tanner Greer post about sort of different factions in Trump's camp who think differently.

Yeah. So I think one piece of this is just that to the extent that tariffs are actually a negotiating tool, It is just going to, like, anytime you do that, you open yourself up to some level of embarrassment because you, if you are a good negotiator, you do not get what you said you wanted. Like, that is the whole point of being a good negotiator is that you got what you did secretly want, but the other side feels like you compromised and that you actually met them halfway. So yeah, to the extent that Trump is actually doing a really, really clever 5-dimensional chess thing here. Then then yeah, he's he's not going to get all the things he says he wanted. And I think that some of the messaging just it like it it is in Trump's interest regardless of what the end goal is for the stated for the implied end goal to be that the U.S. is manufacturing a lot more stuff and that China in particular is manufacturing a lot less. Now there are plenty of configurations in the world where that either doesn't happen or happens to only a very small degree, or the U.S. shifts what it manufactures but doesn't actually increase manufacturing output that much in the aggregate. And, and the U.S. still ends up better off. But some of that, you know, some of these goals, it is again, like, if I'm sure that there were really sophisticated plans for particular parts of particular supply chains that could realistically be moved to the U.S. and other parts that could be moved to U.S.-aligned companies and other things that we could do to sort of box China in where we find different countries or collections of countries that can, in the aggregate, mimic some of China's comparative advantages. And in some cases we just won't, like it is, you just can't really mimic the density of the, the social network that is Shenzhen in addition to all of the, all of the equipment and all of the people and all of the infrastructure, like that part just has to get built over time. And if not everything is co-located in the same city or the same set of cities, then it is going to run less efficiently. And in that case, what we're doing is basically buying insurance. So if we take a supply chain that was all China and move it to it's 80% not China, and, you know, you have 4 or 5 different countries that each have a big chunk of it, but nobody has a majority of it, then that's just a world where a lot more intermediate goods spend a lot more time on ships. And there are a lot more cases where people are trying to coordinate something, but they all speak different languages. And so there are miscommunications and things like there, there is a lot of noise and friction in that. And one difference between that world and the, the continued reliance on China status quo world is that in the latter, you can have situations where the U.S. can no longer buy drones or the U.S. can no longer buy GPUs. There are, there are a lot of things, you know, the GPU supply chain, like, you know, we don't directly buy them from China, but we buy them from places that are very close to China and places that are, if you look at a map in China, that are part of China. So definitely you, There is definitely a case for saying that if the US were spending a little bit more on some of its strategically important imports, but also knew that those imports would still be available if there were a conflict with China, that the US would be better off. And also a conflict with China would be less likely in a scenario where China can't actually, can't immediately remove the US's access to some critical inputs. But yeah, we'll, we'll see how the, how the tariff regime and how the negotiations evolve to get us either In that direction, or maybe back to the status quo, or maybe to some world where there are more barriers to trade and the US out of necessity is manufacturing more. But I think in that world, there's also the question of which companies would actually want to invest in the US. And we, we do know that there is a set of policies that can get companies to invest in the US because private sector spending on manufacturing or on, on fixed capital did actually rocket over the last couple of years, mostly because like Partly, partly because of AI-induced demand, but also because of this targeted subsidies for particular strategic industries, some more strategic than others, but it did work. And I think that it's actually harder to credibly engage in a strategy like that. Like if you wanted something like the, the IRA or CHIPS Act, but just targeting, say, defense and thing and dual-use technologies that, that could be applied to defense. I think people would be telling you like, we want cash upfront before we build our drone factory in Texas. We, we don't want the promise of future subsidies because you had just promised, you know, this, this great trade rearrangement and then you unpromised it and then repromised parts of it and so on. So we just can't underwrite, we can't really underwrite a long-term investment in the U.S. on the basis of a strategy that changes day to day. But, you know, there's, there's a, there's a level of subsidies where you can underwrite that, but that's also a level of subsidies where you are subsidizing grift. So if you pay people upfront enough money that it is worthwhile to build a drone factory in the U.S., there are people who are going to take the money and do other things with it. And this is actually a problem that China had too. They, they had this massive scandal with a chip fab in Wuhan, actually, where a lot of the money seems to have just disappeared, as well as some of the people involved. And it's, it's unclear if they disappeared themselves or got disappeared, but either way, when, when there is aggressive industrial policy that tries to pay upfront because there's some policy expectation and stability, it does select for some fraud.

All In this past week had a fascinating debate between Sacks and Chamath on the one hand and Ezra Klein and Larry Summers on the other hand. It was kind of sort of a representative of a broader debate. It seems like most people can agree that this has been handled not super well from an execution perspective, but people on the right are saying, hey, directionally something like this, or some people had to happen because the current situation was untenable. And this is sort of like short-term pain for long-term gain, either as it relates to our sort of need to reindustrialize or they make some economic arguments. And my guess is that the geopolitical foreign policy arguments are stronger than the than the economic arguments, but I'm curious to hear your reaction. I'm curious to hear your basically take on the idea of did something like this directionally need to happen, or did Larry Summers' position of, hey, free trade is actually this sort of era that we had was actually very successful for us. He would dispute the Zeihan thing of the Zeihan idea that we were sort of subsidizing an order that didn't make sense for us anymore.

Well, I think that there are definitely parts of the free trade system that, like, absolutely, the free trade system worked really well for the US in terms of GDP growth, and it worked really, really well for a lot of the developing world also in terms of GDP growth. And the question is more about what is, like, is there a long-term goal where everything kind of equilibrates and equilibrates at US standards of living or equilibrates at some other standard of living? And is the US willing to defer consumption in order to preserve its, in order to preserve industries that are high value, high margin, have a high upfront cost and a lot of uncertainty, and that probably just don't come back if we lose them. And I think that's, that's the piece where it gets trickier, where if you have a planned economy, but with some capitalist characteristics and the, the level of capitalism, the economy can fluctuate opportunistically, but the state is still able to direct a lot of resources towards the goals that it cares about. Then they can, they can do catch-up growth very straightforwardly. And that was the story for a lot of the East Asian economic miracle countries like Japan and then later Korea, much less Taiwan, but they were, they were kind of attached to Japan in a way that, that let them sort of free ride on, on some of what Japan was doing. But part of the, the approach in both Japan and Korea was we're going to subsidize industries that lower the, lower the cost and increase the pace of iteration in higher value-added industries. So we know that if we dump enough resources into steel, We'll have really cheap steel. We'll be able to make cars. We'll be able to make low, low cost export goods, you know, utensils and just like steel, you know, simple steel components and simple, you know, simple industrial components, but we can also move up the value chain once we've built that. And anytime they're doing that, you can, you can have this period where a U.S. manufacturer who used to produce goods for many different levels of that value chain, you know, some really low margin commoditized stuff, some really high margin, very specialized stuff. As they're able to outsource some of that, or as they stop producing some of the low margin stuff because there is overseas competition, they actually end up with a higher margin business that has more competitive defensibility on its own. But then they can end up in a case where it turns out that it was actually valuable for them to have some of this lower margin, lower value added stuff. Specifically because it just increased their overall economic footprint, meant that more people were working with the company. The company was working with more suppliers. It had more scale advantages, was able to promote more talent internally. And it just, it is, it is really hard to have this fluid way to promote a lot of really smart people within a smart and effective people within a given organization. If that organization's main job is to slowly shrink into its highest margin activities rather than continuously growing at, at all levels. So you can kind of look at, you can look at Amazon as a sort of like taking Japanese industrial policy and just doing it on a private sector basis within one company where they do think very long term. They are very happy to find something where the only limitation is their willingness to put a lot of capital into something and lose money for a while. And then they just continuously scale it and try to ramp up the quality to whatever the acceptable level is and then ramp up the monetization. Continuously beyond that. And as long as what they're doing after that is reinvesting that in more of the same, then they're able to sustain that growth for a very long time. But as they do that, they end up having a very strong negotiating position with respect to many of the companies they do business with. And it's just, you know, over the last 20 years, it's always been better in the long run to be Amazon than to be someone who was doing business with Amazon. Um, in any case where Amazon was the main dependency and the bigger they get, the more they are the main dependency for other, other companies. So I think that that's, you know, that in that framework, you basically want to have some way for either companies or countries to subsidize some low value added complements to the high value added stuff that they can export competitively globally. And if you lose that, it is just much more likely that you end up eventually losing the high margin, high return on equity things that were just that last layer because someone else has done the entire process that you did to get up to that point. And they also know what that point looks like. And, you know, they also know what the prices for those products are. And as they get closer to being able to build them, they also know more about the margins. So I think for, you know, for free trade to be this really, for people to be very long-term optimistic on free trade with sort of protectionist characteristics on the part of other countries, You have to be optimistic that the U.S. will just have some natural advantage in doing, you know, in, in doing higher margin things and that we don't actually need a lot of expertise in the lower margin stuff in order to achieve that. And I think there's, you know, there, there's a kind of coherent model of the world where that is actually true and where, as the U.S. does less manufacturing, we get more and more specialized in other things. And because we're more specialized in those things and we have critical mass, It's hard for anyone to compete with us. And it does seem to be true for both consumer-facing software and business-facing software. It is just really hard for other countries to build and scale companies the way that you can in the U.S., but that also means that those, those same companies are vulnerable to a world where they're tightly integrated hardware and software bundles. And at that point, the companies that are good at the capital-intensive, operationally messy world of hardware, They, they already know how to do that and U.S. companies would have to relearn it.

Let's get deeper into that. This is a, you know, big bet on American dynamism. Say more about what you foresee as to our opportunities and challenges in sort of reindustrializing. And what are you predicting or expecting in terms of how it's going to play out?

Yeah.

So I think the U.S., we, we have like our, our general disadvantage in all of this is very high labor costs and There's an ambiguous disadvantage in terms of regulatory uncertainty where the, the old regulatory uncertainty was micro level, which is you want to build something, you have no idea how long it will take. You have no idea how much money you will spend dealing with every single holdout. And we have a lot of, we give a lot of people veto power and the power to delay things. And we, we have some, some rules that I think were very well intended, but are open themselves up to being manipulated. So I'm just a piece in the New York Times a couple of months back talking about endangered species and There is this kind of dark model where if you take a species that's not endangered and then you figure out, you find 2 subpopulations, you split it up into subpopulations based on some unique trait, and you're able to get each one of those classified as a species. Well, now you have 2 species and they both have smaller populations. And so if you divide them finely enough, you can eventually get your way, you can define your way to an endangered species and then someone can't build the factory or the dam or whatever they want to build. And there, there seems to be some appeal to doing that. So the U.S. had that, that kind of, that kind of obstacle to reindustrializing that actually seems to be going away under Trump. And it has been seamlessly replaced with the macro uncertainty of, we just have no idea how U.S. products will be priced relative to products made elsewhere in the world because we just don't know what the tariff situation will look like. But I think there are, there are some categories like in drones, it is Just kind of weird that for, for, so for, for military purposes, the US does, does do drones, but for, for kind of for law enforcement purposes, there, there's a lot less, a lot less of a US presence. And that seems like a straightforward place where we can look at what's going on in, in Ukraine and say that drones are going to be a pretty important part of the world. We can look at just the, the general history of military equipment and say that it is good to have dual use producers. It is good to have companies that can, from year to year, make a lot of money selling passenger planes and can occasionally scale up their production of military aircraft. Or we look at the U.S. auto industry, which got huge over the course of the 1920s, and say that it was actually, from, from the perspective of a country that needed to produce a lot of tanks and jeeps, it was really, really convenient that we had a lot of people who were in the business of putting, putting tires on things that move and getting them off the assembly line as quickly and cheaply as possible. So I think that something like drones, drones makes a lot of sense. I think that going through, going through the list of inputs for like going through the list of raw materials inputs is probably more of like a check the box thing. And it is still, it is annoying to me that China is able to disrupt supply of rare earths because they are, and I think this is more widely known than it was a few years ago, like they're not actually rare. That's more of a branding thing. What is rare is the facilities to process them. And China has invested in that. And That seems like this is actually a case study in why controlling your supply chain is strategic, is that China can just tell the U.S. that you don't get any rare earths for a while. And we've, there, there've been a couple inputs like that that got disrupted during Russia's invasion of Ukraine. And it turned out that a lot of the companies that depended on those inputs were actually reading the headlines and knew that this was a risk and had done some stockpiling beforehand and Did think of alternative sources, but on a country level, it actually seems like a very useful thing to do. So I think that looking at it from the perspective of we take that, we take our currently globally competitive products. So planes, medical equipment, LNG, and gas turbines and some other kind of industrial capital goods. And we start iterating through the supply chain for those and look at one, what are the cases where we depend a lot on imports from one country, which is usually going to be China, and then look at what, where else we could depend on imports from or whether or not we could onshore that. And if it is the case that onshoring that, you know, requires some subsidies, but not immense subsidies, and it's, it is, you want to subsidize the factory so that the return on investment goes from 5% to 10% and not, you want to subsidize it, the return goes from negative 50% to 10%. Like if we find cases where it is that incremental subsidy that actually makes it worth doing, Then we can look at what else is downstream from that input and then look again at, is there, is there a case where it is almost profitable for the U.S. to build this? And if so, we might want to subsidize it so we can build it. And I think that exercise actually gets continuously easier over time as you, as you iterate over it because you get more used to underwriting this kind of thing and you also just get more visibility into the overall supply chain as more of it is located within the U.S. or at least being carefully looked at from kind of a national security standpoint. So I would say that I think another piece of this is just the U.S. annoyingly does actually have an advantage in just natural resource extraction. We, whether it is drilling oil out of the ground or growing corn out of the ground, things that come out of the ground, we just, we have good ground for that. So there's a different set of supply chains there where oil, oil is a very complicated industry with a lot of very specialized equipment. Some of that is manufactured domestically, some of that is not. And again, knowing what the dependencies are, knowing which countries could be, which countries, if they are blockaded, mean that the US is impaired in its ability to supply a lot of oil and natural gas. It's useful stuff to know. On the shipping side, Trump has talked about bringing back the US shipbuilding industry. That would be very hard to do. China makes a lot of ships and it takes a long time to scale such an industry. And the US does have allies that also produce ships. So I think if I were, if I were building out high-level industrial policy, I would probably say we should continue outsourcing that to countries that are actually capable of doing it. And there may be a category of ships, like maybe you're, maybe a fully autonomous ship is just such a different design question from a ship that is crewed that you could actually just build that supply chain from the beginning in the US. And I think that is a case where we should be looking at what is the next step that hasn't scaled yet and trying to scale that domestically because it's just a whole lot easier to maintain high market share in something than to get the market share from zero to non-zero, especially if someone else has already as part of their industrial policy subsidized the early stages of it.

Other thing people have been debating about related to this are who's, who's negatively affected more, the U.S. or China, or who has more leverage here. Some people say that the country with the deficits is the one that has more of the leverage. Who, who can kind of withstand this, this sort of where we're at right now for, for longer? How do you think about this?

So I think that just in terms of the accounting, it does look like China has the advantage in the sense of they are not any worse off if they sell, if they build the same amount of stuff and instead of buying treasury bonds, they just don't, you know, they just dump it in the ocean or something like they, they still have the same level of employment. They still have the same scaling for their manufacturing base. And it's kind of dumb to do that because the Treasury bonds are a claim on future output, future dollar-denominated economic output. But it is something where it would not day-to-day affect anyone's life in China if they were not actually selling these goods to the US and were just building them and throwing them away. On the other hand, longer term, the US does have more population growth ahead of it. The US does, you know, you can, you can debate a lot of the nuances on the military stuff and certainly the alliances that the US has are more frayed than before, but the US seems at least positioned to survive a military conflict with China long enough to rebuild the parts of its industrial base that it needs to, to actually be competitive. So, you know, in the, in the really extreme worst case scenarios, I think the US does okay. It's just a really dicey situation for a while. And China is, is, you know, there was always that debate over will they get rich before they get old and they are getting old and they're no longer getting rich at nearly the same pace. So in some ways, China does need to wring what it can out of the current system before that system just doesn't really serve China's interests. And they, they did take their time in opening up their financial system enough that they could kind of globalize their, their own efforts to buy claims on foreign assets in order to pay for the, the livelihoods of people who are living in China and are at retirement age and thus no longer working, no longer contributing directly to the economy. And there are, there are a lot of those people. So I think that China's, China's, it kind of has a wobbly sort of dominance in that sense where they, they in the short term rely less on the U.S. than, than the U.S. relies on them in terms of standard of living. But over longer periods, I think they are just more dependent on there being an end market for the kinds of goods that a middle-income country could produce because they're not a high-income country and it doesn't seem like they're really on the path to being a high-income country. I think that is another component of US industrial policy is that we can basically make a list of things that are done by Chinese companies where if you didn't know better, you would assume that an American or European company would be dominant in that sector. So I think things like very popular consumer-facing social media apps that are used outside of China or telecom equipment, For example, and I think in cases like that, having a more explicitly defensive industrial policy and saying that these these are things where we a want to make sure that there are domestic producers and b want to limit the ability of China to to sell these globally. I think it does actually make make sense.

That that does all make sense.

We'll continue our interview in a moment after a word from our sponsors. Hey everyone, Eric here. In this environment, founders need to become profitable faster and do more with smaller teams. Especially when it comes to engineering. That's why Sean Lanahan started Squad, a specialized global talent firm for top engineers that will seamlessly integrate with your org. Squad offers rigorously vetted top 1% talent that will actually work hard for you every day. Their engineers work in your time zone, follow your processes, and use your tools. Squad has frontend engineers excelling in TypeScript, React, and Next.js ready to onboard to your team today. For backend, Squad engineers are experts at Node.js, Python, Java, and a range of other languages and frameworks. While it may cost more than the freelancer on Upwork billing you for 40 hours but working only 2, Squad offers premium quality at a fraction of the typical cost, without the headache of assessing for skills and culture fit. Squad takes care of sourcing, legal compliance, and local HR for global talent. Increase your velocity without amping up burn. Visit chewsquad.com and mention Turpentine to skip the waitlist.

One thing we've been talking a lot about tariffs. I want to make sure we get to a couple other topics here, though I'm sure we'll return back to it.

OpenAI.

There've been some developments with the company. You've written about them. You've written about the company a few times in the past couple weeks. One of them was that Sam Altman has hinted that they are in fact going to go open source or release an open source version. Any commentary on, on, on this or, or just broader sort of OpenAI developments? Of course they had their, their fundraise. You wrote about their, their, their cap table. What else should we, should we talk about as it relates to OpenAI? Yeah.

So there's an interesting, you know, you can look at the economics of open source software in a couple of different ways. And as, as with a lot of things, you sort of have to, if you vary your level of cynicism, you get to very different answers because one version of the open source model is we're going to build something really, really great. And we will understand, we will build something everyone wants to use. We will understand it really well. And therefore we will make our money from helping other people implement our wonderful open source thing. And that was kind of a popular model for people who did want to build software and share it and also pay their rent and pay their mortgage. But it turned out to be a model where you are limiting a lot of your economic upside and you are, you're also running what ends up being a, a very headcount-centric kind of organization. So if you are in the business of helping companies use Linux and Linux is all free and open source, so It's hard to directly charge for it, but you can charge for developer hours, installing Linux and maintaining it and so on, then you tend to be running a company that just has a lot of people and only grows by hiring more of them. And I think for some open source people, this was actually kind of utopia, which is, it was basically, we, we have a guaranteed job for all of our favorite developers who also love the software we've built. And we get to spend all of our time thinking about how to make Linux better and how it can solve more people's problems. Et cetera. So that, that kind of model did work. But another model that is a somewhat more cynical one, but a lot more effective is that you build a free product that anyone can use. And then you build a very similar product that is paid, that has better features. And you continuously advance the quality of the free product such that it continues to be widely deployed. You also continue to advance the features of the paid product. And you always, you, because you can see, because you just have a lot of information on how the product is used, you can kind of see, or at least infer what parts of it you could charge for. And that model seems to, seems to work pretty well. And if more interactions between people are mediated by AI, then very cheap open models where you are basically paying a tiny markup on just the cost of the hardware required to run it, those models will end up getting used very, very widely, especially in low value added tasks like customer service, but also scams. And once they are, if there is a slightly better version of that same model and a model that is built around, built in such a way that it understands some of the limitations and deficiencies of the open model, then the, the, the proliferation of that open model ends up being a subsidy to the closed model. And I think this, this probably works a lot better in cases where you, in cases where it's a domain-specific model, because you just have a lot more control over how it gets used and how it gets deployed. So if you imagine like someone creates an open weight HaggleBot and HaggleBot's job is that you are, if you are selling some product on Facebook Marketplace, HaggleBot will handle the negotiation for you and help you get the best price possible. Well, in that case, if someone buys HaggleBot Pro and HaggleBot Pro is specifically designed to out-haggle HaggleBot, then the fact that everyone uses HaggleBot means that HaggleBot Pro has this built-in locked-in market It would be pretty ambitious to do that in a totally open, you know, general purpose model kind of way. But it does, it does seem like that's kind of the shape of things where it is just a lot easier to produce a huge stream of tokens that are all customized for the recipient. And that means there's a much higher premium on being able to ingest and summarize a lot of data and to do so in a way that's helpful to one specific person. So it kind of ties into their other recent feature, which is Improving memory. And this is just a great thing for lock-in because once you, because whichever LLM you've, you have the most user hours or the most user tokens exchanged with, that LLM knows you better. And so it's actually able to give better answers to your specific questions than other LLMs. So, for example, let's say you sometimes use LLMs to figure out what restaurant to go to when you're visiting a new city. If you previously mentioned to ChatGPT that you're vegan or you're, you're carnivore or you're kosher or whatever. Then, then it will always just know that and you won't be explaining these details. Or if the, if you have other idiosyncratic preferences that it's slowly picked up on, you won't be repeatedly explaining those details. So you end up defaulting to whichever LLM you've used the most. And then you, you are, of course, when you default to that, you're giving it more data. And if the data is actually improving it, then you probably end up using it more than you otherwise counterfactually would have. And then I forget who pointed this out on Twitter, but someone pointed out recently that it is It's possible that this can be tied in with a login with OpenAI button. And at that point, what that means is that it can port a lot of your preferences over to whatever new app you're using. So if we, if we have that example of you've used OpenAI, you've used ChatGPT a lot of times to figure out where to eat when you're visiting a particular city. If you log into TripAdvisor with OpenAI, then maybe either OpenAI shares a bunch of your preferences with TripAdvisor, or maybe that's a privacy violation. What actually happens is that when you interact with TripAdvisor, you're actually interacting with an OpenAI layer on top of TripAdvisor, where TripAdvisor is just the kind of dumb API play and it's piping in a bunch of data. The LLM is analyzing that data for you, and then it's giving you your ranked list based on your preferences. And TripAdvisor does not actually even know what the ranking was and what the final output was. They just provided the data. So it does become a way to make LLMs more of an interface and more of a default way for interacting with other services. And I think that for a lot of companies, that's just, that's something they have to think about is that for now, if they have some service where the idea is they can answer complicated questions in some specific domain because they have a lot of the data, which does describe, you know, describes TripAdvisor, that's also LinkedIn, that's also GitHub, et cetera. They do want to have a natural language interface because that is a way that people are now looking for that information. But they also probably want to think about what their business model looks like if they are purely providing outputs via API as their default, and that the actual front end for the site is basically a demo version that you would use to learn what TripAdvisor is so that you can now use the data that they have collected and filter through your LLM of choice. So I think that that will, I'm sure some companies will just make that transition pretty seamlessly to pure, purely providing data or almost purely providing data. To LLMs rather than providing data to end users. And then other companies will probably just get run over and they will, they will die, but all the data they gathered will live on forever in the weights of all the models that come after this.

You wrote a blog post I wanted to get into asking if white collar workers will experience a blue collar economic recovery. Why don't you unpack that? Yeah.

So there was this period in the early 20th century where The U.S. economy was growing and it was growing because it was industrializing. And then there's this period in roughly the mid-20th century where the economy is growing and manufacturing is not actually growing with it. And then there's this period starting pretty much 1990 or so is about when the inflection happens on the chart. So whatever the real forces behind that were happened earlier. And that's the period where the economy is growing and manufacturing, at least in terms of manufacturing employment, is shrinking. And that is, you know, actual manufacturing output. Still looks pretty good. There are some weird price level adjustments that, that end up being made on that, that make those numbers a little bit hard to compare over long periods. But, you know, the U.S. does still make stuff, still makes a lot of stuff. And China, China has surpassed us, but it's a fairly recent phenomenon. But actual U.S. manufacturing employment has not really recovered from that peak and does not seem likely to ever recover from that peak. It will, there, there will probably be bounces up and down, but the U.S. The US has more of a comparative advantage in the kind of manufacturing where you have a lot of really expensive equipment. You have very specialized people who have some combination of, you know, theoretical knowledge that you learn from a textbook and practical knowledge that you get from using the machine according to its instruction manual, realizing the instruction manual has missed some nuances that you have been able to pick up, and then, you know, slowly, slowly accumulating that kind of expertise. So like we, we probably will just have this increasingly small, more, more skilled up manufacturing labor But one of my questions on this was, what happens, what if white-collar labor actually goes through the same kind of transition? Because one way to look at this industrial cycle and how it affected employment was when manufacturers were doing well, they would upgrade a lot of their equipment. And when they do that, that means that they are basically increasing the amount of capital versus the amount of labor as inputs into what they do. And then when there's a downturn, they can lay some people off, but they still have this equipment. They can rehire some people, but they, you know, if they are still profitable operating at less than full capacity, they, they probably operate at less than full capacity. So they can, they can kind of take their time on hiring people. They have these efficiency gains from having enough equipment to service peak demand. And then when they're at sub-peak demand, they just don't have as many people employed. And for a long time, it's been hard to automate white-collar workers in the same way, but that is getting easier. So. You can imagine a lot of companies where they maybe made a decision to spend a lot on an AI customer service tool or on building a model out of their building a model trained on their internal source code or whatever the thing was, and that they underwrote this based on the level of demand for their product that they saw a few months ago. That demand is no longer here, but they have actually made the investment already. So now their output per software engineer, their output per customer service agent is a lot higher, and they may just find that they don't need to. To staff up that much in the next round. And also, as they're building more of these specialized in-house tools, you have to know more about how an individual company operates in order to function effectively there. So it actually becomes more expensive to hire people. There's a higher opportunity cost to interviewing them and training them. And then there's also a higher cost to letting them leave. So you actually end up with this more static workforce after a while. And it seems like in some, in some cases, what the blue collar workforce did was they retired earlier than they otherwise would have with a lower standard of living, but they did okay, just not as well as they might have. And then in other cases, you have people who downshifted from, they had a union job at the steel plant. Now there's no union and no steel plant. So they spend the last couple of years before retirement delivering pizzas or driving a truck or something. And that is definitely not what they wanted, but it is, you know, they still need to work and it's better than nothing. And I think for, for a lot of workers, they just end up in kind of a less structured, less formal employment environment where it is a lot more gig-based, a lot more variable, et cetera. And also the cost of that kind of gig work has gone down because of LLMs too. So you can have a 24/7 live support chatbot that you just pay for that speaks in natural language, can understand natural language queries in a variety of natural languages. And you can, you can get some sleep. You don't actually have to scale your business to the point where you can have full-time 24/7 customer service. So you can have a lot more just kind of gig work in the white collar space. And if the transaction costs for working with random, a more random assortment of partner companies goes down again because of LLMs, then, then maybe there is, there is kind of an afterlife for steady white collar jobs that exist in that, that more informal sector. And I'm still, I'm not, I didn't write that because I was convinced it would happen. I wrote that because I was kind of speculating that it, it is more likely than it was years ago. Then I still think there's a completely coherent view of the world where the higher productivity of software engineers who are using LLMs means that we write a whole lot more software. A lot of that software interacts with other software and there's just not indefinitely forever more demand, but there is more demand for a very long time. But it's not the only outcome that we could see.

That makes sense as an overview. You also wrote about Harvard and you wrote about them borrowing. And I'm curious to just ask you the broader question of where are the Ivy League schools, the state of them right now with what Trump is planning to do with their own sort of economic position separate from Trump and sort of lost popularity for a bunch of reasons, both structural and political and sheer What value are they offering prospective? What is the state of elite higher education?

That's one where I think I should actually punt on it because what I would love to know and don't have handy is what is their actual annual operating expense? What is their operating income inclusive of and exclusive of federal subsidies? And then how big is the endowment and what's the expected return on that? So my understanding, the annoying thing about talking about Harvard is that Harvard is like, it is a great way to refer to, it's a great metonym for just high status, high prestige school. But actually a lot of the rules that apply to those schools, a lot of the generalizations that apply to them do not apply as strong at Harvard because it is just very far ahead of many of its peers in terms of things like having a very large and very well-run historically endowment fund and having enough of a brand name that if people are choosing among, if someone has their choice of elite schools, They will often choose Harvard above the other options. So Harvard actually can save itself from some of the difficult trade-offs that other places have, but that there's, there's no free lunch. You always just have a different set of trade-offs because now you are much more deeply symbolic. And you also just have a bigger responsibility to, to actually pick the right students, get them well-educated, get them well-networked, get them into important jobs, and then make sure that they actually do good things in those jobs. So that's, that's one that I want to look into. It does seem like Harvard just is probably, you know, they are fine without federal funding. They would obviously prefer to have more money than less, but they will do okay. And I think that's part of the purpose of an endowment fund is that you want to be in a position where you don't have to suddenly change your priorities because somebody else did. And if you're Harvard, you can just keep on going. And then, you know, there are these other interesting open questions on just what is the social utility of having various investment vehicles that have indefinite time horizon and are attached to institutions that are literally older than the United States of America itself. And I think it is probably pro-social. Like, it's, it's probably useful to have at least some organization where it literally makes sense for someone to ask, if we had an investment where we think that the payoff happens in a century, what is the price at which it's worth doing that today? Because pretty much any other organization just doesn't have that kind of timeframe and shouldn't have that timeframe. Like, it's a silly exercise for someone at KKR to ask about the 100-year return on anything unless it is something where it actually has cash flow between now and that point 100 years in the future. But at Harvard, they at least can put a discount rate on it and ask if it's worth doing because they probably will be around once the payoff actually hits. And then the specific thing that I wrote about them was just that partly in response to that shakier fundraising, the shakier funding situation, they were borrowing some money and they announced this plan At a time when markets were pretty choppy and they would've gotten much better terms if they had decided to borrow a few weeks earlier. And there was some speculation online that, hey, maybe this means Harvard is financially distressed. But I think what it actually means is that Harvard is a relative safe haven and they're just very, very unlikely to default on their debts. But they're also a large investor. They try to invest in riskier assets. They're trying to earn positive carry if they borrow money. So they are lending to people who are going to pay them more than they're paying. And that probably the rates that those people are paying have gone up more than the rates Harvard is paying. So in the event that Harvard does need to move more money from its endowment to the operating budget, because that budget is not going to be covered by the federal government, it does actually make sense for them to borrow money to do that because the alternative is sell things that are, sell things at a more distressed price and forego some long-term gains. So again, if you, if you had a shorter outlook than indefinitely, maybe you would choose something different. But if Harvard is expecting to be around forever, and if the endowment fund makes it very likely that they will in fact be around forever, then I think they are in the privileged position to take those long-term contrarian bets when they have the opportunity.

Speaking of privileged position, we'll close on this. Some people are saying, hey, is there going to be mass capital flight? And we alluded to this earlier, but you wrote a piece basically suggesting how, or talking about how sticky the reserve currency is.

Yeah. And one of the elements of that is just that anything that actually weakens the US's status as the issuer of reserve currency is so bad macroeconomically for the rest of the world that people get really cautious. They go into low-risk assets or just they try to get liquidity in order to pay off liabilities that they might otherwise have difficulty paying off. And that often means that it's a flight to dollars. So when the US has a financial crisis, dollar goes up because people have borrowed in dollars and they're worried that they won't be able to source dollars by selling goods and services to America. So they have to source dollars by selling assets for dollars instead. So that means that reserve currencies have this weird self-healing characteristic where anything that's bad for the issuer is actually good for the reserve currency in the short term. And we also, because one of the things that's associated with having a reserve currency is having a large financial sector, and those financial sectors are really good at reallocating capital, but they're always kind of built around whatever the previous status quo was, and that means they are incredibly good at snapping back to that previous status quo. So financial markets are just unbelievably good at pretending that the last big market break did not happen and that things that worked a year ago and temporarily seemed to stop working, but then slightly recovered, they're very good at pretending that those things will just work just fine indefinitely. So you do have anything that helps the US, anything that hurts the US's status as a reserve currency, that reserve status tends to direct things back to normal. And then anything that helps recover that status, the US financial Markets tend to really push strongly in that exact direction. So that, that plus just, it is, it's hard to overestimate how much the world just thinks in terms of dollars and how much of a global default dollars are for a lot of transactions. And this is something that, you know, I look for evidence that this is not true. And I will definitely start paying attention if large multinationals have debt that was denominated in dollars, it's rolling over, and now they're issuing euro or yen bonds because those are the better deal. But they view those as the better deal long term. But my suspicion is that a lot of dollar-denominated debt will be rolled over in dollars and that the financial system will remain dollarized by default. And I think what a lot of people are calculating is that whatever Trump does, it is very hard to completely undo the dollar as a global default and to completely unwind that network effect over the space of 4 years. And that the closer he gets to doing that the less likely it is that anything even remotely Trump adjacent will be a viable policy direction for the US. And so, you know, you're sort of, I think if you're thinking about reserve status, you have to ask yourself not just what could be done to the US's reserve currency issuer status over long periods, but literally what could get done between now and the midterms, because midterms will not be friendly to such strategies if those strategies start to actually hurt US standards of living.

We're— that makes sense. We're at time, so I want to be mindful. But, Bernd, as always, this is a great, great wide-ranging discussion.

Until next time.

Alrighty. Next time.

Upstream with Erik Torenberg is a show from Turpentine, the podcast network behind Moment of Zen and Cognitive Revolution. If you liked the episode, please leave a review in the Apple Store. Hey everyone, Erik here. At Turpentine, we're building the first media outlet for tech people by tech people. We're the network behind the show you're listening to right now. We have a slate of hit shows across a range of topics and industries, from our AI and investing cluster of podcasts to shows that drive the conversation in tech with the most interesting things founders, investors, and influencers like Econ 102 with Noah Smith. We're launching new shows every week and we're looking for industry-leading sponsors. If you think that might be you and your company, email me at erik@turpentine.co. That's E-R-I-K at turpentine.co, and let's partner together.