Equity as an Option on Future Reinvestment
A good intuition pump for thinking about companies in general, and the stock market in particular, is to ask: why does the average company trade at a premium to book value? The S&P 500's total market cap is about $32tr, and it trades at 4.1x book value, so for a mere(!) $7.8tr you could replicate all the assets—buildings, factories, equipment, inventory, receivables, cash on hand, etc. And yet investors are overwhelmingly more likely to allocate their money to an existing public company at some premium to book value than they are to rebuilding the same business at a discount.
This is not an entirely silly question. Some of it comes down to accounting: the book value of $100m of last year's branding, employee training, etc. cost is $0, but obviously that kind of work still has value. And there are other accounting elements: a nuclear power plant's book value will drop over time as it depreciates, but its replacement value has gone up as new plants have gotten more expensive. Tobin's Q Ratio is an effort to look at this as a signal, by measuring the market value of companies relative to the replacement cost of their assets. Right now it's at 1.24, down from a recent peak but well above the long-term trend. (This ratio was below 1 from the 1930s through the 90s, i.e. there was a multi-generation period where corporate America, by one measure, was simply not worth the effort.)
Of course, it's easy to think of companies that are worth more than more than the money that's gone into them, even if we correctly capitalize upfront costs that have long-term payoffs. In other words, the premium to book value does make sense even in a world where capital is mobile and competition is ubiquitous, in part because the best companies have assets that are worth more than their accounting value and in part because the biggest chunk of some companies’ economic value is the fact that they can keep reinvesting in good assets. There are many ways to model this out. One fruitful option is to take the nuclear power plant comparison from before, and apply it to other kinds of investments that are cheaper to make upfront.
For example, setting a high standard for product quality, customer service, unwillingness to offer discounts, etc. has some cost. The cost of maintaining such a standard is lower. So the right time to obsess over it is from day one; when the "customer support" team is every cofounder checking a shared inbox, the cost of responsiveness is low because there simply aren't that many customers. And maintaining this—both keeping latency down and treating customers well—is partly a matter of tracking metrics and partly one of setting a good example. Being tough to negotiate with on pricing is another sort of asset whose value compounds; being stubborn early on (especially when there's room to be flexible on what gets delivered in exchange for delivering it at a particular price) establishes a norm internally and externally that can scale as revenue grows. For reputational issues in general, the "reputational replacement cost" goes up as a function of the size of a business; when a big company messes up, or is perceived to, it's a long and daunting process to get back to even. It wasn’t easy for Google to make “Google” a verb, but whatever it cost, it would be much, much more expensive to convince people to stop “Googling” and start “Binging” or “DDG-ing” instead.
There are other practices in the broader world of "get things right the first time" that can produce compounding returns. The more equity someone has in an early-stage business, the more their de facto part-time job is recruiting; this applies to VCs and to early employees. And it means that hiring and raising capital are also decisions about which network to tap into (when a company has a founder from a name-brand tech company, it doesn't just mean that the founder has been vetted by that company—it means that the hires they make through their network will include a subset of people vetted by that company and by the founder.)
And there are some kinds of expenses that are closer to a bet on the future state of the world. For some categories of product, the world needs 1 +/- 1 of them, so building it means some chance of a total loss, some chance of a monopoly, and the possibility of a cozy-or-not duopoly. Distribution networks and AI models both fit this category; the right time to build one is before it's obvious that it will get a good return, because that's also the time when it's plausible to be the only company building it, and there are plausible future worlds in which it's lucrative to be the only company to have built it.
Working backwards, the big companies must have been good at at least one of two things:
- Raising money.
- Getting an above-average return on the money they raised.
And the bigger a company gets, the more difficult it is for point #1 to remain an advantage. A company that's better at raising than at investing can do fine in private markets (and can, with a bit of luck, evolve into being good at both—some of the skills required to negotiate with VCs translate nicely into the skills required to negotiate with key hires). But as such a company gets bigger, the due diligence process gets stricter, and once it's public there's a very strong incentive for short sellers to identify flaws in its economics.
Warren Buffett has a nice quote on the matter of ROE that's worth thinking about: "If you earn high enough returns on equity and you can keep employing more of that equity at the same rate—that’s also difficult to do—you know, the world compounds very fast." That's a deep point, because there are many businesses that can achieve a high return in retrospect, but not in prospect.
Take the biotech industry: The biotech ETF, XBI, has produced an overall growth rate of 10.8% since its 2006 inception, about a point ahead of the S&P but with sharper drawdowns. Within that industry, though, we might treat each company as a business that has a 10% shot at a ~110%+ return on the fixed investment in a single drug, or a 90% chance of liquidating without achieving anything. Looking backwards, publicly-traded biotechs will have good returns on investment, but that's because they're the most successful subset of the entire industry.
Much rarer and more precious is the kind of company that can both produce a high return on its existing capital and continuously reinvest at above its cost of capital. Take Buffet’s Berkshire Hathaway as an example: the company is really a collection of two types of companies: those that will get minimal reinvestment and produce steady returns (think Buffalo News, which was acquired when it was the strongest paper in a two-paper town, and which had many years of profitable monopoly status before cable TV, AM radio, and finally Craigslist devoured its economics) and those that can take some of that harvested cash flow and reinvest it in something that will get a decent return. Berkshire's utility holdings (Berkshire Hathaway Energy) and its quasi-utilities (like the BNSF railroad) send lots of their cash flow back to the home office, but collectively they'll absorb around $12bn in incremental capital expenditures this year (Energy's capex plans are here, BNSF's here).
When companies are valued based on a discounted cash flow analysis (explainer here for readers who want a deeper dive), the valuation number hinges in part on just how long the company can reinvest and grow before its growth rate converges on something below the discount rate applied to those profits. A continuous-reinvestor has a much longer growth period, and a correspondingly higher valuation once things settle down.
Of course, part of the trick is trying to find such a company. Peter Lynch seemed to use this pattern; from Beating the Street:
I was attracted to fast-food restaurants because they were so easy to understand. A restaurant chain that succeeded in one region had an excellent chance of duplicating its success in another. I'd seen how Taco Bell had opened many outlets in California and, after proving itself there, had moved eastward, growing its earnings at 20 to 30 percent a year in the process.
Once a chain has gone from regional to working well in two regions, the way to bet is that it'll work just about everywhere. Some things have changed since Lynch's day. Unfortunately for buy-what-you-love stockpickers with a penchant for fast food, the market has gotten pretty good at slapping a high multiple on chains that can show good unit economics and can plausibly talk about reaching "the other 90% of the country."
Still, it's a strong model, and it applies to many kinds of businesses. When you're underwriting some long-term growth trajectory for a business, a good question to ask is "what is it that makes their dollars the kind that produce a return on equity of 15%? (or 20%, or 25%) when the rest of corporate America mostly earns its cost of capital?" There isn't necessarily a good answer; some businesses started with a lucky break and converged on the median performance for their overall industry; last week, paying subscribers read about Capital One's trajectory from a proto-fintech to a big bank ($), and the commensurate reduction in shareholder returns from enviably-techy to about what you'd expect from buying equity in one of America's twenty biggest banks by assets.
But these companies do still exist. And they're usually priced accordingly:
- TSM is buying equipment that other semiconductor manufacturers can buy (at least if they're not under sanctions), but their process knowledge and long-term customer relationships create persistently higher returns on equity.
- Every successful creative effort at Disney turns into an opportunity for more capital expenditures: a popular character from a movie franchise can get their own Disney+ show, or can become an attraction at a park.
- Cintas (written up for subscribers here ($)) already delivers company uniforms to over a million businesses, so adding other products to that delivery has a low marginal cost.
- Microsoft (disclosure: long) has a more abstract distribution network, but when a new business tool gets identified, like video chat or, well, chat, Microsoft can throw it into the bundle.
- Salesforce, of course, can do this too, often through acquisitions like their purchase of Slack. Nobody said this was easy.
A good company is a carefully-constructed bubble of negentropy; it's an environment in which high returns, on capital and people, are sustained over long periods, creating a company that’s worth more than the sum of its parts. But fighting entropy is hard, and only gets harder with scale; the bigger a company becomes, the more of its surface area is exposed to uniqueness-sapping, ROI-depleting problems. Some of these are internal; it's hard to maintain a distinctive company culture while adding massive headcount. Some are external: if the company does something special and profitable, competitors will try to figure out how they can do it, too, and other companies in the supply chain will try to weaken this advantage so the high profits accrue to them, instead. So the end conclusion of searching for the ability to continuously reinvest at high returns is a search for companies that have beloved brands, favorable regulations, or a relentlessly paranoid management team that's obsessed with delaying the onset of mediocrity.
This doesn't mean that short sellers have to bet that the company goes to zero; there are plenty of businesses that are merely okay, but are valued as if they're good or great. Since many short sellers are trying to balance out a long book, a trade where the thesis is "this company won't do much over the next few years" is perfectly acceptable. Especially because if such a company is raising money from public markets, the short seller solves the problem of what catalyst will lead people to sell; when the seller is the company itself, all you have to do is keep an eye on cash balances and assume that they'll tap markets when they need more funding. ↩︎
A Word From Our Sponsors
Here's a dirty secret: part of equity research consists of being one of the world's best-paid data-entry professionals. It's a pain—and a rite of passage—to build a financial model by painstakingly transcribing information from 10-Qs, 10-Ks, presentations, and transcripts. Or, at least, it was: Daloopa uses machine learning and human validation to automatically parse financial statements and other disclosures, creating a continuously-updated, detailed, and accurate model.
If you've ever fired up Excel at 8pm and realized you'll be doing ctrl-c alt-tab alt-e-es-v until well past midnight, you owe it to yourself to check this out.
The Information has a good look at Apple's early work with augmented reality ($), which they prefer to VR. Some of that preference may be a pure product judgment about which headsets customers will actually want, but—in another example of how the same investment can get a higher ROI at the right company—they also have ways to integrate AR with existing hardware:
Some of these efforts have already been rolled out for the iPhone and iPad. In 2021, Apple announced Object Capture, an AR feature that automatically built 3D models of objects based on photos taken with an iPhone or iPad. In 2022, Apple announced RoomPlan, which uses the camera and lidar sensors of newer iPhones and iPads to scan and generate virtual floor plans of rooms.
It's tempting to look at this as a sign of Apple's underlying philosophy about technology. The company tends to make tools that are good for interacting with an external world, like a better phone that taps into legacy phone networks, an iPad that's good for reading books and magazines, a watch that tracks users' heart rate and nags them to avoid being sedentary. A purely escapist device doesn't fit with that model. Whereas one of the advantages of AR, from a company's perspective, is that if it's augmenting reality, and reality goes away, it's possible to reach a world where the device never comes off. Which would make the 100+-pickups-a-day interaction level of phones seem like a comparatively low usage level.
Car Loans and Non-Stationary Distributions
When recessions hit, default rates for just about every kind of loan go up. But there are always surprises in the details. One of the unique elements of the Great Recession was that car loans held up relatively well, with delinquencies below their mid-90s peak. That's not happening this round, as car defaults are rising. One culprit is the ongoing effect of inflation, which ate into low-income borrowers' savings in the last few years even though it's abated a bit since then. And another reason is the aftereffect of another kind of inflation: since used car prices were so high during the pandemic, people who couldn't afford used cars were squeezed. A strong job market helped, but even when employment holds up in the aggregate, there's enough noise in the data as specific people get hired and laid off that default rates start ticking up.
So far, 2023 has been a great year for people who owned the kinds of stocks that gave them a tough year in 2022 ($, WSJ), as heavily-shorted shares outperform. New years sometimes signal new investor priorities, as everyone works backwards from a last-trading-day-of-the-year celebration to the themes that will get them there. And one of the things that saved long/short funds' performance last year was that even if their high-quality long positions got walloped by higher rates and growth fears, heavily-shorted stocks did that much worse as the market started asking questions like "does this company need to exist?" or "how many EV companies can grow faster than Tesla?" or "Even if they all did grow faster, how much would Tesla-but-not-run-by-Elon be worth, anyway?" It doesn't take much short-covering to move these companies, both because they're smaller than they used to be and because other short-sellers get understandably nervous when a position moves against them. In general, the more levered an investor population is, the harder it is to infer big sentiment changes from price changes; a small shift in underlying views and a properly risk-averse attitude can send already-volatile growth companies on a wild trajectory.
The total amount of working capital needed to ship raw materials around the world has risen by $300bn-$500bn since the end of 2020 ($, FT), driven by higher inflation, higher interest rates, and conflict-driven rerouting. In a low-rates environment, it's easy to forget that everything in a warehouse, on a boat, on a plane, on a truck, etc. is also on a balance sheet, and that it needs to be funded. And this funding is generally not a big component of overall costs. But higher working capital needs do require more capital to shift around, and when markets are nervous, moving that capital can be expensive enough to make a difference.
Big tech companies have been pressured to accelerate their AI release schedules as startups and pure AI companies launch increasingly impressive products. A good working model is that in any given category, the right bet is that the biggest companies have the most impressive internal prototypes, but that the most impressive product to be publicly released comes from a smaller business with less to lose. In the classic "company-as-call-option mode, a business like OpenAI is well out-of-the-money, and benefits more from volatility, whereas Google and the like are deep in-the-money and have to be sensitive to both expected value and to potential value destruction from either a botched launch or an expensive one.
But that doesn't mean those big tech companies aren't benefiting from AI. The WSJ has a good overview of Meta's recent improvements, many of which are driven by applying AI ($, WSJ):
Executives internally began pointing to signs of an imminent turnaround soon after the last earnings report. At the October internal talk, [CMO and VP of Analytics Alex] Schultz said Meta had already absorbed the worst from Apple’s tracking changes. The damage had diminished from the more-than 8% hit to revenue early in the year to just 2.5%, and likely would disappear entirely in the fourth quarter, he said.
The piece also has a notable tidbit for anyone hoping to call inflections in the datacenter and semiconductor cycle: "But even with that spending, the recent AI efforts have stretched Meta’s processing capabilities to its limits, according to the internal documents. One November memo said that Facebook engineers not working on AI would have to abstain from building computing-heavy features on company servers—and delete some of their existing work." Software businesses still have near-zero marginal costs on average. But as AI increases global demand for compute, "near-zero" doesn't round to zero.
Disclosure: Long META.
Companies in the Diff network are actively seeking talent! If you're interested in exploring growth opportunities at unique companies, please reach out. Some top current roles:
- An AI company building new services to help small companies accelerate their growth is looking for an experienced product designer who can create interfaces for products that haven't existed before. (SF)
- A fintech startup that lets investors trade any theme as if there were an ETF for it is looking for a senior backend engineer. (NYC)
- A well funded early stage startup founded by two SpaceX engineers is building the software stack for hardware companies. They're looking for a backend engineer who can build services that quickly process large amounts of data. (Los Angeles)
- A new AI company is looking for senior engineers with experience building scalable systems with Node and Typescript on AWS. Management experience is a plus. (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.