In this issue:
- Dead Airlines—Airline bankruptcies are incredibly unsurprising: the structure of the industry gives new entrants surprisingly good economics for a while, but ultimately forces them to run a race against a worse cost structure, and to take bigger risks from fuel.
- Synergy—The GameStop/eBay offer makes perfect sense from GameStop's perspective, but not for reasons that are exciting to eBay.
- Indexes—In the end, the job of an index is to be whatever people assume it is.
- Warehousing—Yes, sophisticated traders try to exit trades by selling to more naive market participants. But also, sometimes the affinity between an industry and a capital source is historically contingent and not the natural way for things to work.
- Capacity—Unfortunately, the EU needs to apply a higher cost of capital to their AI infrastructure investments than a profit-seeking company would.
- Defense—Satisficing and optimizing.
Talk to this post on Read.Haus.
Dead Airlines
It's kind of convenient for anyone writing fiction set in the last half of the twentieth century that there's so much turnover in airlines. It means that there are evocative ways to set the scene, in terms of when it's happening and what social class the characters are, entirely by having them take a flight on a particular airline—a college student flying PEOPLExpress in 1983 is not too well-off but maybe upwardly mobile; someone flying Pan Am in 1965 is part of what was then called "the jet set." Flying TWA meant one thing in the 50s and another entirely by the late 90s.[1] Airline stocks are also priced as if an airline isn't a permanent business; the good ones tend to trade at high single-digit P/E ratios mid-cycle, which implies that the market thinks they're good for maybe one more cycle on average.[2]
Airlines are volatile for some of the reasons other cyclical companies are: planes last a long time (and have long lead times), travel demand fluctuates, and that means that travel recessions happen at peak capacity. That's true of plenty of other industries with slow feedback: the biggest chip fabs, cruise ships, oil platforms, midtown office leases, etc., all happen at the peak. But it has a few other annoying features: airlines are unionized, and the unions are aware that the industry is cyclical and try to be senior creditors.[3] So it's an industry where many factors conspire to make shareholders the most junior claimants on some very volatile cash flows.
The annoying thing about the airline business is that it's not uncommon to be able to pencil in some plausible cost estimates and find that it would be a good idea to start a discount airline. Low-cost carriers have a smaller minimum size than airlines that run a network and make money on connecting flights (there is a very long tail of connecting flights here, where the same traits that make airlines tough—high fixed costs, with the incremental cost of a ticket being, literally, peanuts—actually come in handy. It's very plausible for the cheapest way to get from one mid-sized city to another is to pass through CLT, ATL, DFW, etc. But if you can develop a route map that connects people outside of hubs to the kinds of destinations they'd like—Florida and the Caribbean for Spirit, Vegas for Allegiant, many places technically close to popular destinations for Ryanair—you can get a nice little business going. Part of the way that business works is that unions push for seniority-based pay, and the faster a company grows, the lower the average tenure of employees. So there's basically an automatic cost advantage, in terms of paying for labor-hours. But you're not just paying for labor-hours. You're also paying for maintenance, paying an opportunity cost for idle planes, paying for backups, paying for passenger refunds when you turned out not to have enough backups, etc. All of these problems get relatively easier at scale, even if they're still harder in an absolute sense—a younger fleet is easier to maintain, for example, so an airline that's growing fast is also working with newer, better-functioning, more fuel efficient planes.
So the structure of the industry is constantly pushing for the creation or reinvention of budget carriers that grow fast. They tend to emerge in regions that were growing particularly quickly at that time that they were founded, they grow for a while, and, in many cases, they ultimately implode. Spirit, as of Saturday, has joined that grim parade. The irony of high-growth, low-cost carriers' early cost advantage is that they have one key disadvantage: since their other costs are so much lower, fuel is a larger proportion of their cost structure (26% of revenue for Spirit last year, compared to 17-20% for Delta, American and United.[4] So when fuel costs go up, their tickets get relatively more expensive, which is the last thing a company competing solely on cost can afford to underwrite. Airline collapses are inconvenient for travelers who were halfway through a round trip, and obviously bad news for employees and investors. But they don’t have much of an effect on long-term travel demand: the planes, pilots, mechanics, flight attendants, airports, etc. all still exist, and the demand Spirit was servicing is still there, too. There will be some frictional costs associated with reshuffling those assets, but they’ll also end up slight better-aligned with where demand for air travel is.
The general rule might be that discount carriers are an asset like the cash flows from a specific movie, song, oil well, or video game: they can be positive expected value, but their cash flows will go to zero in a few years. It's just a feature of the industry that new entrants can get attractive economics for a while, but end up more exposed to oil prices and more dependent on continued injections of capital to maintain their margins. Essentially, a growing airline can toggle between operating leverage and financial leverage, and grow at whatever pace they think has the best risk/reward. For a trade like that, at any given time the winners will include a healthy cohort of companies that were overbetting for a while and finally had a streak of bad luck.
Enough time has passed that I can even apply this to my own story: nothing screams "transplant" quite as loudly as moving to New York by taking a Southwest flight to Long Island MacArthur Airport and only then realizing just how long Long Island is. ↩︎
"Mid-cycle" P/E is a slightly fuzzy term, but it's useful. The usual dynamic with cyclical stocks is that there's money to be made shorting them at 5x earnings or owning them with negative earnings, because there's a point in the cycle where demand is rolling over or costs are spiking but the last four quarters looked decent. But their P/E at that point is basically meaningless—they generally won't earn that kind of money until the same point in the next cycle, but by that time the industry will be a bit different. Right now, the industry is going through exactly that process of rapid multiple compression ahead of lower earnings. ↩︎
The most unionized industries tend to be the ones with high fixed costs, because those are the industries where work interruptions are costly. Pilot's unions are a special case: there aren't many jobs with above-average pay, high union membership, and workers who vote pretty Republican. So they tend to be more insulated from some of the politics around labor. ↩︎
United and American are 30bps apart. Delta's number is complicated because they own a refinery, which changes the economics and also adds some jet fuel revenue to the numerator. So their reported number is 15.5%. You can back into a more comparable number by looking at Delta's airline-only revenue, and adding back the fuel expense saved by the refinery, which gets you to 17.1%. ↩︎
You're on the free list for The Diff. Last week, paying subscribers read about how importing random libraries used to be a wonderful part of tech's gift culture but is now a bit more dangerous ($), the possibility of a supply crunch in forward-deployed engineers ($), and another lens on why CEO succession is hard ($). Upgrade today for full access.
Diff Jobs
Companies in the Diff network are actively looking for talent. See a sampling of current open roles below:
- Series A, ex-Navy defense firm building AI-enabled drone defense systems is looking for an electrical engineer with range: schematic design, electrical simulation, printed circuit board (PCB) design, and firmware development in high performance languages (C++, Rust, etc.) If you’re interested in power and control systems for mission critical technology, this is for you. (Austin, TX)
- A startup building a new financial market within a multi-trillion dollar asset class is looking for a data scientist with commercial financial experience. (if you’ve been an investor, banker, etc. but are newer to the data side, that’s great too.) (NYC)
- Lightspeed-backed team building the engineering services firm of the future is looking for founding members of technical staff excited about working alongside civil engineers to translate their domain expertise into the operating system that powers the next era of great American infrastructure. If you’re an engineer with strong product intuition, who's energized by access to users, and excited by the prospect of transforming how we design and construct our built world with frontier AI, this is for you. (NYC, SF or Remote)
- AI Transformation firm with an ambition to build an economic world model to run swathes of the private, unstructured economy is looking for FDEs, Platform Engineers, and business generalists who understand how to solve problems. (NYC, SF, Remote)
- Well-funded, frontier AI neolab working on video pretraining and computer action models as the path to general intelligence is looking for researchers who are excited about creating machines that learn from experience, not text. Ideally you have zero-to-one pre-training experience and/or are a high-slope generalist who’s frustrated that the big labs aren't doing this. (SF)
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.
And: we're now actively deploying capital into early-stage companies through Anomaly. Our focus is on defense, logistics, robotics, and energy. If you'd like to chat, please reach out.
Elsewhere
Synergy
In one sense, GameStop and eBay are in the same business, of operating a retail platform that also lets people sell used products. In another sense, a merger between them makes more sense from a financial-stunt perspective than from a synergy perspective. The fundamental pitch is that GameStop's current management was able to cut their overhead enough to flip game stop from a $381m loss to $418m in profit. On the other hand, GameStop in 2021 was a structurally challenged company that wasn't adjusting to their industry's limited growth prospects. But eBay just reported its biggest YoY gain in gross merchandise volume since 2021. It's a company that's found a way to get growth from an aging brand name, something GameStop hasn't managed. GameStop has done a good job on capital allocation, in that their shares are overpriced and they're happy to supply investors with more of them. (Disclosure: short a little GME as a funding short). The way to maximize that advantage is to use their stock as currency to buy something. That's a straightforward enough pitch, one that fits with the premium they're offering, and that doesn't have to go overboard on projected synergies. But it’s also a pitch that isn’t compelling to the sellers: telling eBay shareholders that they’re getting adversely-selected is not going to make those shareholders excited to sell.
Indexes
When someone creates a financial index, they're making a claim about what's worth tracking and how to track it. But over time, there's a feedback loop, where an index starts to represent a bet on some abstract set of companies: you know exactly which index is Big Companies, or Tech Companies, or A Few Big and Relatively Old Companies, are. And once that's how some market participants treat them, they're doing their customers a service when they bend their original rules to reflect what people would expect. So S&P joining the Nasdaq in relaxing index inclusion criteria makes sense. An average investor will assume that the day SpaceX, Anthropic, or OpenAI go public is roughly when they join the roster of Big Companies and Tech Companies, while institutional investors who care about indexing-driven flows will account for changes in inclusion criteria. So it actually makes the market more efficient to make the index whatever a randomly-selected investor imagines it to be.
Warehousing
Banks are looking for ways to sell some of the datacenter loans they've made ($, FT). This kind of thing always gets interpreted as banks concluding that the easy money has been made and looking for someone to take larger risks, but there's a simpler model: sometimes there's a new target for investment, and it turns out that one kind of capital is much better-suited to it than the others. So that asset class turns into an industry bet, and the industry becomes dependent on the asset class. That ends up making the whole system slightly riskier than it needs to be. It's happened with software companies (which were too much of the public equities market and not enough of the PE and debt market until PE started fixing it), and previously with high-yield bonds expanding from something only issued by financially precarious companies to being a source of growth capital. Over time, some assets do end up skewing to particular markets that can best support them, but not as extremely as they once did.
Capacity
A plan for the EU to spend €20bn on AI infrastructure is being criticized because there isn't enough domestic demand. Which feels odd given that most of the rest of the AI world is complaining about scarce compute instead. But, unlike neoclouds, the EU has some political constraints that would make it hard for them to build a giant datacenter and then lease it to an American company instead. So, this is a case where America's flexible capital markets are such a nice advantage: investors expect companies to be pretty mercenary, and if a company ends up pivoting how it monetizes an asset after it's built that asset, shareholders can either sell their shares to someone who believes more in the new model or try to run an activist campaign to stop it. Put exactly the same asset under the control of a different kind of institution, and it has fewer degrees of freedom and a harder time taking risks.
Defense
The Department of War has made deals with seven AI labs to use their products under similar terms to what Anthropic previously rejected ($, The Information). The US government is an interesting buyer of AI tools, because while they're sensitive to top quality, the tools are also complementary to some capabilities that are unique to the US, sometimes in magnitude but sometimes in kind. So they don't strictly need the best model, as long as they have better overall capabilities than adversaries who do have such a model. The scenario they have to worry about is one lab pulling ahead of all of the others, but the usual leapfrogging dynamic makes that unlikely: a lab would have to have a competitive advantage it sourcing compute, and at getting investors to underwrite extremely aggressive spending, and at retaining talent even as their equity skyrockets. The government can afford to satisfice, and expect to have access to the best model much of the time and a good enough model all the time.