"Capital Light" is a Trait of the Cycle as well as the Industry
Many investors who take a checklist approach to public and private investments will put capital efficiency near the top of the list: all else being equal, it's better to own a company that produces cash rather than consumes it, and one that can produce more cash over time without consuming a lot to get there. This is great for investors, and it's also great for founders, because companies that don’t produce cash need to be funded by issuing equity (diluting their share of the upside) or debt (diluting their control in downside scenarios).
The love of capital-light businesses is one reason so much investor attention moved to software companies in the early 2000s: the relentless decline in hardware costs and the rise of open source solutions meant that the minimum cash outlay for starting a company kept declining. And that decline has, at least in theory, continued; if you're moderately familiar with AWS and at least one major software stack, and are willing to use ChatGPT or code completion tools, spinning up some specific niche software business ("Market research tools for indie creators!" or "A mobile-first personal finance app for gig economy workers!" or "CRM for missionaries!") has a time cost measured in hours and a cash cost measured in pennies.
But obviously, if anyone can create a new product, the creation of products will become commoditized. If you Google "CRM for Missionaries," you'll find plenty of companies advertising in roughly this category (usually it's a broader CRM for nonprofits or faith-based organizations, but since these products are sold to organizations rather than individuals, that makes sense). Notably, several of these results are ads. Even in extreme niches, to convert interesested hypothetical users into free trials or customers, you're going to need to pay upfront.
So sure, actually building the product isn't necessarily capital-intensive, but it’s only a stable situation if getting users becomes the cost bottleneck. And there are plenty of software companies whose core product wasn’t especially hard to develop and whose ongoing sales and marketing spend is the bulk of the cost.
But, in fact, there are many more products where R&D spending as a share of revenue stays stable, or doesn't decrease much, over long periods. What's happening there? Often for enterprise products, the core use case was easy to build; it's not especially hard to make basic clone of Docusign, and there are many other companies with lightweight products that offer the basic feature of allowing someone to digitally sign a PDF and then email a copy to all the signatories. But that turns out to be a subset of what the product really is; there's a cost to continuously adding new integrations—where the documents get automatically stored, where a contract can be generated, who gets alerted to the existence and status of these documents, and how. At first, a document-signing company might think it just needs a handful of these integrations. Maybe a Slack tool for tracking status and pinging salespeople to get their contracts signed, a Dropbox integration for getting them saved and backed up, and a connection to Salesforce so they're generated inline and automatically shared with the appropriate person.
But it turns out that none of those integrations integrate with a product that has 100% market share. If you only work with Slack, you'll struggle with Teams-based companies; if you assume everyone has a chat product, the email-centric companies will be annoyed. Meanwhile, not everyone with a CRM uses the one from Salesforce. Given enough time, even the CRM-for-missionaries gets an integration.
All of that effort—again, all of which is in the fantastically high-margin, low-capital-intensity world of software—creates three new ongoing costs:
- There are so many features, and so many interactions between them, that the size of documentation balloons. There's also a trickier marketing question, because the marginal new user of the product is not someone looking for the original core feature, but someone trying to integrate that feature into specific workflows. As their needs get more specific, broad advertising works worse, and hyper-targeted search ads whose keywords indicate an immediate propensity to spend get expensive.
- The cost of user support also rises, and the amount of information support staff need goes up. They're no longer helping customers use a single product; they're helping them debug the surprising interactions between different products.
- Even worse, these other products are constantly evolving; the price of maintaining a single piece of software that doesn't plug into anything else can be low, but the cost of making sure it continues to cooperate with everything else is not.
In accounting terms, this mostly shows up as operating expenses, with a little bit of cost-of-goods-sold sprinkled in. But really, the economics look more like this: building a complex product that integrates with other complex products, and then acquiring customers and keeping them happy, creates an intangible asset. That asset, like any physical asset, has a finite useful life. And, like a piece of factory equipment or a plane, the cost is reflected in a) the cash outflow from paying for maintenance and spare parts, and b) the non-cash cost of depreciation, which, on average, must be met with a cash expenditure to continuously replace these depreciating assets before they, and thus the business built on them, gets depreciated down to zero.
Software can still be a wonderful business with favorable economics, but capital will always flow into places where it can get a high return, and out of places where returns are low. There are traits of software companies that make them great, like high margins, the ability to pivot early, and the possibility of scale benefits like network effects and lock-in. But these traits must, over time, be offset by the competition to either start similar companies or to acquire customers for them. Absent excellent management, government protection, or extraordinary good luck, the industry's economic balance sheet, as opposed to its accounting balance sheet, must end up looking like that of a capital-intensive industry like airlines. And, in fact, there are airlines with good management, government protectionism, good luck, or a combination thereof that have produced high long-term returns for investors.
Understanding an industry in the early days means assessing whether or not it can grow, what impact it will have, and what its particular capital cycle looks like. But in the long run, understanding an industry is about understanding why it's different today in order to understand why it will end up looking similar to every other industry given enough time.
There are some investors who are ruthless about making a list of criteria and only investing in a company that ticks every box. For example: must be #1 in their category, must have high margins, must earn a 15% return on equity and be able to deploy more at similar returns, must have exceptional management, etc., etc., etc. Run a long enough list of these and you end up with a few dozen stocks, all of which trade at 30x earnings except the one whose economics are falling apart and the one that's actually a fraud. A patient, concentrated investor can make money with this kind of checklist, by waiting for one of the companies to either run into a rough patch or have a long period where growth keeps happening while the stock goes sideways, but this has a cost in terms of patience and, if they're managing outside money, explaining why they should collect management fees in exchange for mostly sitting on their hands. It can be more fruitful to have a long checklist of ideal companies and a sense of what discount they deserve for missing a given item on the checklist. ↩︎
In fact, one big difference is that a physical capital asset is much easier to finance with debt—if an airline goes under, the lessor can seize the planes. If you lend to a software company, good recovering your investment by selling the source code at a bankruptcy auction. Of course there are other parts of the business that can be lent against. As Robert Smith pointed out, software companies are, in practice, among the most senior creditors to their customers. This creates a bizarre situation where a company that sells enterprise software to big businesses is, through many layers of indirection, raising equity in order to acquire a portfolio of long-term interest-only investment-grade corporate bonds. ↩︎
Companies in the Diff network are actively looking for talent. A sampling of current open roles:
- A company building the new pension of the 21st century and building universal basic capital is looking for a GTM / growth lead. (NYC)
- A crypto proprietary trading firm is actively seeking systematic-oriented traders with crypto experience—ideally someone with experience across a variety of exchanges and tokens. (Remote)
- A vertically integrated PE-backed cannabis company is looking for a data analyst with banking experience. Excel wizards encouraged to reach out. (Remote)
- The leading provider of advanced options analytics — “the ASML of options trading” — is growing rapidly, very profitable, and looking for a generalist who can excel in chief of staff and business development functions. A trading, quant, or similarly technical background is a big plus. (Connecticut, NYC)
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.
Individual Bets and Risk Premia
The WSJ has a piece on Pentwater, a fund that bets on the outcomes of mergers and has done well betting that some recent FTC lawsuits will fail ($, WSJ). Merger arbitrage is a fun space because it's a human problem that turns into a math problem. Calculating the odds of a merger going through is a process of thinking about the incentives for all sides, and how each party will respond as things change. If a company has launched an acquisition, for example, it already has a sunk cost in time and money, and probably has some expectation that the deal might have to be modified in order to get through. A company might end up accepting conditions for a merger that are bad enough that the merger wasn't worth doing, as long as some of the costs have been paid upfront.
The other subtext here is a point The Diff raised a few weeks ago ($): a portfolio of stocks that are all the targets of pending mergers will have a worse risk-adjusted return if there's some risk that hits all of them at once, instead of idiosyncratic deal-specific risks. When the FTC was getting more aggressive, this lowered the value of all merger targets (and rumored merger targets), which is especially painful for a fund that buys these stocks with leverage. But when the FTC doesn't succeed in blocking a given transaction, 1) the market's implied odds of all pending deals go up, so the stocks go up, and 2) the correlation between those stocks afterward goes down since "FTC Risk" is more company-specific. So a firm that times the market in antitrust enforcement wins even more.
Tinder's Price Discrimination
Tinder may introduce a $500/month ultra-premium tier. On the one hand, choosing a partner is probably the most important decision people make, and while they spend a lot of time on it, they don't necessarily spend the optimal amount of money. On the other hand, it's hard to convert money directly into better partner-options, or at least hard to do it without being unseemly and facing adverse selection. Tinder is probably not the ideal product for this particular task anyway. The real lesson might be that some users of the app have a basically unbounded appetite for converting money into more or better matches—but another reason is that a product with a ludicrous price point makes Gold ($8.33/month) and Platinum ($10/month) look positively cheap by comparison.
In 2017, there were over 700 smartphone brands. Now there are around 250. Both the peak number and the current trend are a good case study in how hard it is to build a business selling durable goods, especially when the replacement cycle slows. The biggest winners can handle that, but smaller, undercapitalized ones can get wiped out by one bad demand cycle or (as in 2020) one tough period of supply shortages. Meanwhile, the phone market has shifted to a phone-and-subscription-software-bundle market, as part of the long process of moving the recurring revenue from phones away from telcos and towards handset companies instead. Since software has a high fixed cost relative to its marginal cost, software economics favor big winners—and one result of that is many small losers.
The Partner-With-AI Arms Race
One lively market with a persistently wide bid/ask spread is the market in trading some level of exclusive access to foundation models for some amount of preferential distribution of those models. A handful of companies have the model and the distribution (Meta, for example, may announce some AI-powered chatbots soon), and Google has been able to use Bard alongside search. But in other cases, there are gains from trade in allowing companies to tie themselves together. So, Amazon has agreed to invest up to $4bn in Anthropic in exchange for Anthropic using AWS ($, The Information). It's hard for this deal to be truly exclusive given that there was a similar investment-and-cloud-spending deal with Google earlier this year. As it turns out, one of the big corporate finance decisions AI research companies face is how much exclusivity to give up when, and by strategically selling this at different times, they can get a valuation premium as a strategic investment multiple times from multiple investors.
Disclosure: Long Meta, Amazon.
Abraham Thomas has an important, brief piece on metrics: when a company tests something, the worst possible outcome is a slightly positive but unimpressive result. A huge success is great, because whatever they were testing clearly works. A huge failure is actually nice, too, because it indicates that they can move on to something else. But a mediocre success has minimal information content, and there are many explanations for mediocrity that it takes time to work through. One way to frame this is that for early-stage companies, information is uniquely scarce: you know that your particular business did not exist before and are on a quest to find out whether that's because it was a good idea no one had thought of or executed correctly, or if, as should be your prior, it was a bad idea that didn't exist because everyone who had previously considered it either rejected it or quietly failed. The earlier a company is in its life, the more it feeds on novel information rather than revenue—because before a company is profitable, its first-order goal is to take actions that add compelling slides to the next fundraising deck, with revenue and profits as a second-order concern whose main initial value is that they lead to compelling up-and-to-the-right charts rather than the business impact itself.