In this issue:
- Will The EU Choose to Compete?—Mario Draghi has produced a detailed report that is unsparing in its assessment of the EU's policy mistakes, and surprisingly pragmatic about solutions. Some of them will take time to implement, but in a few cases there's potential for immediate shifts.
- Unit Economics—The rise of the three-mile train.
- CoreWeave and Capital Structure Arbitrage—CoreWeave as a well-structured trade.
- AI and the Distribution Problem—Some of the best use cases for AI only work if they're implemented by the biggest platforms.
- Markets and Narratives—An easily-explained move in Chinese bond prices leads to conspiratorial thinking.
- Convergence—PE as an accidental way to build a better bank.
Today's issue of The Diff is brought to you by our sponsor, 2 Hour Learning.
Will The EU Choose to Compete?
A popular topic in American tech circles is Europe's inability to compete. There are some cheap shots here, of which I'm guilty, but there are some policy decisions European countries have made that, whether as an unintended consequence or as an accepted policy tradeoff, have made Europe a worse place for growth companies. And that's contributed to making Europe a slower-growing economy: the current EU 27 had roughly the same share of global output as the US in 2011, at 21.4% vs the US's 21.1%; last year, the US was 25.8% of global GDP and the EU 27 countries were 17.6%. Last week, the European Commission released a report, authored by Mario Draghi, analyzing the EU's competitive situation and outlining steps to fix it. This report is shockingly pragmatic—on energy, for example, it makes a point of leading with a discussion of natural gas, deferring long-term plans for renewables to a later chapter. And it makes a good contribution to both understanding European competitiveness and finding ways to improve it.
A good place to start is to quickly review some long-term policy choices, by the EU and other countries, that got us close to the recent status quo:
First, rich countries other than the US tend to consume more leisure as their GDP rises. That takes the form of shorter hours, more vacations, and earlier retirements. The US is an outlier in that respect (as of a few years ago, the entirety of the gap between Japanese and US GDP per capita growth could be explained by work hours and labor force participation—in both countries, output per hour grew 1.7% annualized, but the US has a younger population and now works longer hours ($). Making the transition to working fewer hours and retiring earlier will hurt GDP per capita, even if it improves quality of life.
But in a steady state, the impact of those two is different: lower work hours will reduce output and almost certainly reduce growth (at least some of individuals' productivity growth comes from deliberate practice; increase your work hours 10% and you're getting 10% more practice at whatever you do, and even if your skill peaks, it peaks earlier—and you have more time in your career to transfer skills to more junior people). But the impact of earlier retirement is actually mixed. If people in your country retire at 60 rather than 65, then at a given level of output per worker they'll consume less per year, both while they work (because they're saving) and when they retire (because the consumption patterns post-retirement are anchored to their existing consumption). But that output still exists, and if it isn't being consumed then it's being saved instead; over their working life they’ll save more money, and over the course of their retirement their average savings balance will be higher because they’re starting with more and spending it at a slower pace. If there are high-return investment opportunities, earlier retirements can actually lead to more growth, since they mean that more output is invested rather than consumed.[1] What you'd get from this, over time, is a more capital- than labor-intensive economy, which is also a good description of the world's most productive economies.
A second issue, more specific to Europe, is trying to provide labor market stability at the transaction level rather than in the aggregate. There's a natural case for making it hard to fire workers; a company with 500 people is not going to be put out of business if it suddenly has to work with 499 instead, but a worker who loses their job can end up in a tough spot. Similarly, a landlord who jacks up rents can afford the occasional vacancy better than their tenant can handle being suddenly homeless. But the equilibrium from codifying this in law is that it's a tax on providing jobs or property, and that tax isn't always offset by subsidies elsewhere. The US has more or less decided that this kind of friction is okay.[2]
And this collides with a third problem: while the US government's immigration policies are a mess, the US's growth industries are very good at putting highly-educated immigrants to work ($, The Diff); Stripe and Tesla each have market values in excess of half of the aggregate market cap of their founders' home countries. This tends to skim off a lot of talent that would be building domestic companies and is instead building US-headquartered multinationals, which may or may not do much business with the origin country.
Meanwhile, European tech protectionism often takes the form of punishing outside companies retroactively for their success instead of proactively encouraging local ones to grow. And this is true even of superficially neutral rules like GDPR: a fine levied based on global revenue is an implicit tax on companies that do business outside of Europe, relative to the ones that do all of their business locally. Draghi's report actually makes some cautiously negative remarks about such rules ("while the ambitions of the EU’s GDPR and AI Act are commendable..."). The report indirectly contrasts this with China's approach of subsidizing local companies while making it hard for foreign ones to do business there, especially for certain strategic inputs: China doesn't have dominant share in extracting many metals and minerals, but they have over 50% share of processing facilities for cobalt, lithium, and rare earth metals, and for graphite their processing share is 100%.
This points to one of the most insightful frameworks from Draghi's report. Part of the reason the US has grown so much faster than Europe is a story of volatility exposure in two places, first in energy and second in strategically important technologies. In the early 2000s, the US was basically short energy volatility, which meant being short geopolitical volatility. If the world got crazier, to the upside or the downside, US consumers suffered: a crazier-upside world was what we mostly got, with China's growth putting constant upward pressure on the price of commodities, most importantly oil. That meant that globalization was a drag on US consumers' spending power, which was offset for most of that period by continuous appreciation in assets, especially real estate, and advances in the financial technology that turned housing appreciation in Phoenix into demand for imports and trips to Disney World. This was obviously unsustainable, but, fortunately, it flipped around due to fracking.[3] There are few countries that are truly long geopolitical volatility, since it's a net destructive force—Iran and North Korea might be the only contenders—but by being a net producer of oil and the world's main military power, the US is at least better-hedged. When capital globalizes, it seeks out good risk-adjusted returns, and even if US economic fundamentals don't improve that much on average, the US now is a much lower-volatility bet than it was a decade and a half ago.
Volatility matters, but levels matter, too; natural gas is much more expensive to transport in liquified form than through pipelines. For a while, this was fine for Europe, since they were getting plenty of it from Russia. That's no longer the case. And as a result, the report notes that gas prices in Europe are 3-5x the US level. And Europe has some energy-intensive industries, like steel, cement, and chemicals. Europe also charges higher carbon taxes than other places, which would be noble if they had otherwise lower costs, but in effect means that they're paying for the frictional cost of moving emissions elsewhere.
The report proposes a few solutions to this (yes, like American wonks they would strongly prefer a more streamlined permitting process for new investments in renewables and electricity transmission; there's also a suggestion that environmental directives be weakened when they interfere with renewables). But they also propose some bolder moves. One plan is to use some of Ukraine's natural gas storage capacity to have bigger reserves, blunting the impact of price swings. (The report proposes "counter-guarantees to de-risk gas storage in Ukraine," though it's unclear what these counter-guarantees might be.) It also suggests a sort of Organization of Natural Gas Importing Countries, where the EU would collectively bargain for long-term access to natural gas at more fixed prices. A collective buying organization would avoid some bidding wars between countries, but it wouldn't change the fact that natural gas prices are intrinsically volatile, and there's a reason long-term agreements are usually benchmarkedto market prices rather than fixed. What they could do is buy LNG at the usual variable cost structure, and sell it to domestic firms with longer-term fixed-price contracts, and issue EU-wide bonds (a precedent they set during the pandemic). The economic impact of this isn't that it makes energy inputs any cheaper. In fact, by telling energy users not to worry about volatility, it makes them more expensive. But it shifts the near-term impact to the EU economy as a whole rather than to particular sectors, and if they're worried about losing those sectors, that can be a net win.
One of the surprises in the report is on telecoms. This is partly a surprise because they don't seem that indispensable to long-term growth. Yes, we all use more bandwidth as we get richer, and sure, smart cars and smart industrial devices will need a lot more. But the other surprise is how the report analyzes them, which has the flavor of a sort of global macro activist banging the drum that Europe would get a great IRR from restructuring its telecom industry. There's literally a chart on page 71 showing aggregate return on capital employed compared to weighted average cost of capital, showing that the industry has been destroying investor capital for the last decade, and the problem is generally getting worse.
The reason for this is another policy choice: the EU encourages competition in telecommunication, and tends to have more small firms offering it, especially small companies serving small countries. This keeps rates low, but it also keeps investment low, and eventually means that Europe is the single best place to get mediocre service at a great price. If there's a ceiling to the economic utility of providing more bandwidth, that's a sound policy aim, but it's also a bet against progress. (Somewhat surprisingly, this report doesn't seem to have moved the share prices of many tiny European telcos—of which, again, there are just far too many. In the aggregate, European telcos with enterprise values under $5bn trade at 6.2x EBITDA, below any of the big US carriers, and there are plenty with a low single-digit multiple. Later this week, paying subscribers will get a more detailed look at some of the opportunities here.) It's a very Kalecki-flavored argument: returns for businesses are sometimes constrained by access to inputs and by comparative advantage, but within those limits the profits are partly a policy choice. If Europe wants a more dense telecommunications network, it needs to be a continent where telecom is not a sector where money goes to die, and that means transferring some money from households to the corporate sector.
Of all of the proposals in the report, loosening up M&A restrictions on telecommunications sector mergers sounds like the easiest to implement. Draghi's proposals have a sensible, technocratic flavor to them: if the EU is one market, then competition among mobile carriers should be considered EU-wide, and it's fine for tiny countries to have fewer of them. If anticompetitive mergers are a problem, regulators should deal with that after the merger has that effect, rather than beforehand. And if a merger reduces consumer choice, but also gives the buyer an incentive to invest in better infrastructure and in R&D, maybe giving consumers a smaller number of better choices is the right move.
The report highlights some bright spots, like space, where European firms are doing well, and defense, where Europe's problem has been demand-side rather than supply-side—they're making weapons, but exporting almost 40% of them. Pharma is another clear case where European companies are globally competitive, albeit with room for improvement.
Overall, there are good questions about whether or not any of this will be implemented. Defense has momentum right now, and space has secular drivers, but those are both areas with a long lead time. Energy costs are salient, but pragmatic solutions involving a slow move off of hydrocarbons just aren't as exciting as renewables, even if the path to a fully renewable grid is long and expensive. The telco M&A plan sounds like the area where it's most plausible to see immediate action, because all the regulators really need to do is make it clear that they'll step aside and let big telecommunications companies and their investors get to work consolidating the market. Which wouldn't represent a sea change in the EU economy by any means, but which would be a sign that the Draghi report is being taken seriously. Draghi himself has ample political capital; more than one media outlet has referred to him with the shorthand of "the man who saved the Euro." He's 77 years old, and presumably won't be involved in government all that much longer, so putting his name on a report like this is a way for him to spend that political capital while it still exists (much in the same way that the Volcker Rule has outlived Volcker.)
Every country that's trying to grow its economy has to decide where the shock absorbers for the inevitable volatility will be. A good reading of this report is that the EU is trying to act as if that volatility simply doesn't exist: Russia will always be there to sell natural gas; Chinese electric vehicles will never compete with European internal combustion engines; and if every European is viewing the digital world through a lens provided by an American smartphone OS, an American social network, and an American search engine, it's not the sort of problem that periodic random fines can't somehow fix. The US's implicit system is that the two big shock absorbers are the Federal government and the net worths of the very rich—both categories a) have vast amounts of wealth, and b) tend to see disproportionate drawdowns during recessions, and individual impacts when industries get disrupted. This report is a call for the EU to recognize that volatility and uncertainty exist, but they're not a good excuse for inaction.
The magnitude of this effect depends on many factors—for example, savers in Japan tend to put their money into safe assets, and are implicitly funding the country's deficits through their spending. Japan does automate, but its ability to do so isn't really constrained by household savings. But as long as 1) savers' money either goes directly into risk assets or at least gets channeled into those assets, and 2) there's some level of home-country bias—a German retiree would rather own shares of Siemens than of GE Vernova or Hitachi—then these savings ultimately lower the cost of capital for domestic growth. ↩︎
Some developing countries make this decision by omission, and in their context it has the side benefit of encouraging workers to save as much as possible because there won't be a safety net waiting for them. That crushes near-term consumption, but funds a lot of investment. ↩︎
In light of the talent-importing point above, one of the key players in fracking was George P. Mitchell, whose father was a first-generation immigrant from Greece named Savvas Paraskevopoulos, but changed his name when "a paymaster got tired of writing his long name and threatened to fire him." ↩︎
A Word From Our Sponsors
Can this really be true?
Unit Economics
Transportation is often a more-is-better business: when planes, ships, and trains get bigger, the unit cost per passenger- or ton-mile goes down. Some of this is just physics: you're expanding the volume of what's being moved faster than you're increasing the area that has to overcome air resistance, and some of the cost of friction both within the engine and between the train and the tracks are lower. The result of this is trains extending for three miles or more, blocking crossings for over seven minutes, and potentially longer if they need to get decoupled ($, WSJ). This is partly an information problem: one of the reasons for the US's rail renaissance is that railroads got better at scheduling, so it's more feasible to know where they're going and have traffic route around them, especially if that traffic increasingly consists of cars whose drivers use Google Maps, or are in fact autonomous. There's also room for Coasian solutions, and in one sense, longer trains are a tradeoff rather than a pure loss: if the same amount of cargo is carried on fewer, longer trains, then crossings will be blocked about the same percentage of the time, but in longer, less frequent chunks. Railroads get the benefits of cheaper transportation in the form of rising margins, but if that transportation makes everything else a bit less convenient, they'll do more of it than is optimal.
CoreWeave and Capital Structure Arbitrage
CoreWeave, which is in the business of buying GPUs that other users can't get their hands on and then renting them out to those users, is raising another round at a $23 billion valuation, which may include $400-500m for employees. CoreWeave's fundamental insight is that there's a slice of the AI capital structure that 1) generates immediate cash flow, 2) isn't something Nvidia feels the need to do right now, and 3) benefits Nvidia by keeping GPU ownership, and thus AI, fragmented ($, The Diff). They've basically converted a few elements of Nvidia's strategy into a stream of cash flows, and since those cash flows are backed by assets for which there's still high demand, they make great collateral for loans. It's always possible that this will break, but you have to model an immediate and severe decline in the demand for training or inference in order to imagine a world where there isn't substantial residual value from their assets.
AI and the Distribution Problem
AI tools are great at handling tasks that can be described in simple natural language, occupying the niche where something is too trivial to delegate but annoying to just do. This makes them great as glue between apps—but most chatbots don't, by default, have access to all of the apps that they need to handle a task like "put all of the dates from this document into my calendar" or "Make me a list of all the people I frequently communicate with and are located in New York so I can figure out who to meet with when I visit in a few weeks." These tasks can be done if the AI tool is something integrated with an OS, or with a suite of frequently-used apps that are all controlled by the same company. So, at least so far, the rollout of AI continues to benefit big companies ($, WSJ). There was always something vaguely strategic about Google having an office suite or Microsoft owning LinkedIn, even if the direct connection between those products and the company's core business was more tenuous. But now that there are tools that are good at exactly the kind of busywork that involves cross-referencing between multiple apps, the companies that own a suite of them have more things they can do for customers—and more ways to lock those customers in.
Markets and Narratives
The Diff has written a few times about the decline in Chinese government bond yields, which is partly a function of their low supply relative to other big economies' sovereign debt, and partly a result of Chinese investors having few other enticing options. This piece covering the phenomenon is interesting for just how conspiratorial everyone insists on being: "Experts attribute the rapid yield decline not only to macroeconomic imbalances but to strategic market manipulation by certain financial institutions... Speculative trading by smaller financial institutions as well as individual investors has also played a role, with some treating bonds like stocks..." You don't need an elaborate conspiracy for why a country working through a credit-fueled property bubble would expect a flat yield curve and low rates on government bonds. But if that country is China, where there's an interventionist state that can make life very difficult for financial institutions, and that state still derives its legitimacy from growth, what you end up hearing is that this move is mosty inexplicable but probably somehow suspicious.
Convergence
The classic private equity model is to borrow a lot in order to buy a company, fix it up and/or wait for the market to turn, and then flip it. Over time, this model has evolved, as PE firms concluded that 1) they can do more trades than that, and 2) there are times when they want to be selling equity and buying debt rather than the reverse. As they've moved in this direction, some of them have locked up long-term capital that gives their funding similar duration to their lending, in the form of life insurance companies. And that leads to another natural expansion: Apollo is increasingly participating in investment-grade lending ($, FT). One of the earliest fundamental insights in PE was that a so-so level of capital appreciation could, with sufficient leverage, lead to extraordinary returns. The same is true on the credit side: if your cost of capital is set by the returns that life insurance or annuity holders demand, then finding a way to deploy lots of capital safely matters more optimizing for the highest possible gross return on that investment.
Diff Jobs
Companies in the Diff network are actively looking for talent. See a sampling of current open roles below:
- A well funded early stage startup founded by two SpaceX engineers is building the software stack for hardware companies. They're looking for a data engineer with 7+ years of experience across the data engineering stack, ideally with experience building distributed systems in Python or Go. (LA)
- Growing edtech company with successes online and offline is looking for an influencer who can reach the parents of gifted and talented kids. (Remote, full- or part-time).
- An AI startup building unique tools for compliance-focused companies is looking for a director of information security and compliance with 5+ years of info sec related experience. Experience with HIPAA, FedRAMP a plus. (NYC)
- Growing team at a multi-strategy firm is looking for senior quantitative researchers with direct experience or exposure to a relevant field (statistics, optimization, ML) 250k-400k base + bonus (NYC)
- The policy development arm of an established stablecoin is looking for an analyst to help define US crypto policy from first principles. Some background in crypto or policy preferred, but not required. (Remote)
Even if you don't see an exact match for your skills and interests right now, we're happy to talk early so we can let you know if a good opportunity comes up.
If you’re at a company that's looking for talent, we should talk! Diff Jobs works with companies across fintech, hard tech, consumer software, enterprise software, and other areas—any company where finding unusually effective people is a top priority.