The Bid/Ask View of Who Gets to Raise and What Gets Traded

In this issue:

• The Bid/Ask View of Who Gets to Raise and What Gets Traded—Surprisingly few prices are set by market activity, though many important prices are. The reason for this is that it's hard to get a product into the right shape for being traded. In fact, you can describe some investing strategies as buying something when there won't be an active market for it, then selling it when there will be one. And changes in the availability of information and information-processing naturally have an impact.
• Privacy and Insurance—We all think we're above-average drivers, unless someone's monitoring our driving to decide how much we should pay for insurance.
• AI Fragmentation—The lucrative quest for national champions in AI.
• Too Central to Fail?—The US can't afford the perception that Boeing could go under.
• Meritocracy—Elite firms targeting less-elite schools.
• Activism—A shareholder activist campaign demands that management focus more on the business than on buybacks.

This issue of The Diff is brought to you by our sponsor, Mercury. Read on to learn more.

The Bid/Ask View of Who Gets to Raise and What Gets Traded

Most assets traded on most markets exist in part because they're convenient to trade: bonds supplanted annuities as tool for financing governments because you can put a price on a bond without having to give someone a physical and still be left wondering whether they have some barely visible chronic condition that's letting them adversely-select you. It's also easy to buy economic exposure to a thousand barrels of oil, but unlike a stock or a bond, there's a lot more implementation work that goes into this: "oil" describes a category of mixes of molecules, not an element or a digital abstraction—to have a functional futures market, you have to decide what kind of oil counts, and to have a range of acceptable levels for specific hydrocarbons and other things in the mix. And, unlike a digital product, the price is a price you're paying for oil in a specific location. Turning oil from sticky black goo into something you can day-trade from your iPhone was an immense challenge of both infrastructure and inventing abstractions.

There are some assets that just don't make it onto a market. You can buy one 2.5-billionth of Nvidia's future cash flows for a specific price, $1,208.88, and if you change your mind two minutes later, your cost to exit that trade is measured in pennies. If you wanted, you might be able to find someone who'd sell you a fraction of their own future cash flows, but if you change your mind it's probably better to mentally write the asset down to zero than to somehow find another buyer.

For most of the assets you might want to buy, the market simply doesn't clear. Transaction costs are too high; members of the relevant set are too heterogeneous (you can trade pork bellies, but not hot dogs; there are too many kinds), most of the prices consumers pay are administered prices, not market prices, so there's almost never any action and when there is the market-makers get run over by informed traders.^[1] And even if a market solves all of these problems, it still needs some minimum level of liquidity: if no one ever trades, market makers don't want to spend time paying attention to a market where nothing happens, so they widen bid/ask spreads and back away. And as the bid/ask spread widens, traders decline to participate.

It would be convenient to have a more complete market, but for now it's not going to happen. This actually gives us a good description for how venture capital works: VC money goes into businesses that start out hard to understand—and are not worth understanding in great detail because the power law distribution of returns means that the expected return is unbounded, and rises with the number of investments.^[2] That same dynamic makes investors rationally insensitive to valuation: financial projections from the company are just a way to see whether or not the founders priced things out before putting together some numbers, and an actual financial model that spits out a discounted cash flow number is a huge waste of time.

That is absolutely not true in the growth stage, but it is true when the checks get bigger and the plan is to exit via IPO, at which point the price will be set by people who religiously update financial models. So at some point, the right way to evaluate a business shifts from analyzing vibes and trying to assess whether the founder is the optimal level of delusional, to thinking about incremental EBITDA margins and trends in CAC.

People tend to think of funding rounds as a way to raise enough money to get to the next funding round, but another way of looking at it is that they're trying to shrink the bid/ask spread, and make the company more legible to a cohort of investors who increasingly focus on the downside.^[3] Every company's growth is also a process of getting legible enough that a new cohort of more risk-sensitive and quantitative investors is willing to touch it. And this process continues even within the public markets: systematic traders will tend to be a larger share of the volume in bigger stocks, because those are more legible to their models, too.^[4]

This model extends to other assets, too; options used to trade over-the-counter, and part of what made them a real market was the closely-timed availability of hand-held calculators and the Black-Scholes formula. At one level, this was a disaster for options traders, because it meant that now their customers knew when they were being ripped off. But it turned out that customers were much happier to trade when they knew precisely how ripped-off they were rather than having to guess.

And it extends even further than that. If what makes a market is a low enough natural bid/ask spread that it's actually worthwhile to trade, then the main thing that makes new markets possible is the wider availability of information. There are other features, of course, and many times when markets emerge there are also a handful of people who drove them forward. This, in some cases, is a form of local legibility that has the same effect of reducing spreads: Michael Milken's success in high-yield bonds was not really a matter of better technology or of a new data source, but because he was the main participant in the market, he had a unique data source: since he was the only way to trade some issues, he always knew exactly who owned how much of what. And since he was trading all the time, he also had a sense of who the likely buyers were. That's the difference between having a position for months and having it for minutes, and this liquidity feeds on itself for a long time.

A world with more API endpoints and more tools for analyzing the data those endpoints produce is a world where there are plausibly many more markets. And the more technological end of the financial services industry is full of people who are making a good living by making markets a few basis points more efficient, but who know that they could make a lot more money being the central node in a brand new network of people trading some asset that hadn't existed before. Meanwhile, the firms those people work for have often exhausted their high-sharpe non-scalable strategies, are diversifying into more capacious but less exciting ones, and presumably have some nostalgia for the early chaotic and lucrative days.

So one long-run effect of both the digital intermediation of more transactions and the growing availability of AI tools will be that the financial sector's share of economic activity will not only grow, but grow at an accelerating pace. We are far from a world where every asset is on the optimal balance sheet and where ever risk is born by whoever is either best-equipped to evaluate it or most willing to tolerate it. That's a business problem, but it's fundamentally gated by technological limits, and those limits are rapidly going away.

---

1. A hot dog futures bet, for example, is partly a bet on Costco refusing to depeg. ↩︎

2. Note that this does not mean that spray-and-pray is optimal. Instead, it means that venture investors have to have a view on whether the hard part of finding out whether or not a company is part of the power law distribution they care about, or if the hard part is just scaling into those power law bets. Founders Fund and Y Combinator are philosophically at the opposite ends of this continuum, but they've both done very well. ↩︎

3. You can find investment vehicles that back a diversified portfolio of later-stage startups, are willing to back things even if they're not confident they won't go to zero, and count on the power law to deliver excess returns, but in this case they're very late-stage indeed, the power law is weaker, and the whole process takes longer. That investment vehicle is called "SPY." ↩︎

4. That's not the impression you might get from looking at 13Fs, or looking at the biggest holders of individual small-cap companies. But the quant strategies that lead to owning 3% of each of hundreds of small stocks are low-turnover ones that are making a diversified bet on the small-cap factor. And small-caps are very small indeed; the S&P 500 is 80% of US market cap, so pretty much any big fund that dabbles in small-cap stocks will own a lot of them. In fact, if such a fund does this, a meaningful driver of their alpha will be how good they are at minimizing how often they trade and reducing the market impact from when they do. ↩︎

A Word From Our Sponsors

The art of establishing a strong financial foundation

In the world of startups, being the first can be exhilarating. Being the first finance lead to join a startup, however, can be daunting. That’s why Mercury simplifies all your financial operations by powering them from the bank account* and shares the insights you need to succeed.

Mercury’s VP of Finance, Dan Kang, explores the five steps to establishing a strong financial foundation in your first 90 days. The article covers how to navigate existing processes, align with the CEO on key priorities, craft a financial strategy at scale, and build relationships across the company to make better decisions.

Read the full article to learn the art of simplifying complex financial operations, so you and the company can perform at your highest level. 

*Mercury is a financial technology company, not a bank. Banking services provided by Choice Financial Group and Evolve Bank & Trust, Members FDIC.

Elsewhere

Privacy and Insurance

An NYT article highlights the trend of insurance companies buying data on people's driving habits and using it to more efficiently price insurance. People mostly dislike it. That presents a fun behavioral paradox: there's a folk theory that most people think they're above-average drivers, and this turns out to be true, with more than half of Americans surveyed rating putting themselves above the 80th percentile in driving safety. And yet, the working assumption they have about insurance companies adjusting their rates based on their driving is that it would make them worse-off. It's hard to square these, and they're not a direct worry: an insurance company that successfully targets safer drivers will charge less overall while earning more, and the scenario people have to worry about is that every insurance company charges them more—but that's only a reasonable thing to worry about if they've been free-riding on artificially cheap insurance because insurers underestimate how safely they drive!

In a less competitive industry, it's fair to worry that companies will use data about you to extract more money from you, though the usual way they operationalize this is to raise prices for everyone and then discount through their apps. That doesn't make everyone better-off, but it does the next best thing: the person who sees the biggest price hike in that scenario is someone who doesn't patronize the company very much, which means it just isn't that big a deal for them if they stop buying entirely.

AI Fragmentation

One source of demand for Nvidia is that some countries have been accumulating GPUs, and now other countries want to do so defensively ($, WSJ). It's a wonderful place to be. (For Nvidia.) For some companies, a reasonable worry would be that this pulls forward demand, and that there will be a glut of GPUs if countries don't find some use for them, but the pace of improvements means that older GPUs make more sense as a way to provide subsidized inference for older open-source models than to train a new proprietary one. So that base case looks reasonable for Nvidia, while the bull case—a few foundation model providers that all offer comparable results, none of whom run away with the market—is one where Nvidia captures the upside.

Too Central to Fail?

Boeing is getting close to a ratings downgrade to junk status ($, FT). One of the quiet evolutions of corporate finance over time is that companies aim to be as close to junk-bond status without actually being there. Fixed-income investors often have a mandate to focus on investment-grade debt, either for regulatory reasons or as a risk-management policy, and an investor who is so constrained knows that their best shot at good returns is to buy the lowest-rated bonds they can. (Meanwhile, there's a break in the series for realized sharpe ratio by credit quality: barely-investment-grade does the worst, because that's what investment-grade fixed income managers crowd into; just-barely-junk does the best, because for an investor with a junk bond mandate, it's the least interesting risk to take.) So this kind of downgrade, in addition to being symbolic, also has a disproportionately big effect on Boeing's cost of capital. Boeing, while organized as a for-profit company, is also a policy tool: they make high-status exported goods in the US, and do a lot in defense. So there's a government incentive not to let things get too bad, even if that's what Boeing deserves.

Meritocracy

Large consulting firms are at least saying that they're open to hiring people who didn't attend elite schools. Part of this is a reaction to competitive pressure—consulting is often the null hypothesis for what smart, hardworking, but undirected people want to do with their lives, since it's a well-paid way to get real-world exposure to, potentially, multiple interesting industries. Tech, and to a lesser extent finance, are very competitive with this life path right now, and the economics of consulting don't support continuous pay increases of the magnitude that those industries can. But this puts consulting in a very tricky position: people coming from elite backgrounds are very sensitive to the risk that their status will peak at 22, take a hit when they get their first real job, and never quite recover. Managing around that risk-aversion can mean accepting a smaller incoming class of consultants, or running the risk that the business's status is permanently impaired. Which makes this recruiting decision all the more impressive! Not only does it mean that the low-risk career path is being run by a more risk-tolerant crowd than before, but it implies that they think their own status might be competitive with that conferred by the Ivies.

Activism

Last week, The Diff questioned Elliott's activist campaign against Texas Instruments ($): if there's a problem with TI investing too much, why not just short them and invest in their less capital-intensive peers instead? So it's a delight to see that Elliott is going after Southwest on the grounds that they need to invest more in their business, particularly in the back office, and that their model is getting old and needs to improve. A good test of being well-calibrated is that in a given category of contentious issue, you're equally likely to be on either side in specific situations. By that standard, Elliott is doing things exactly right.

Diff Jobs

Companies in the Diff network are actively looking for talent. See a sampling of current open roles below:

• A CRM-ingesting startup is scaling its LLM-powered sales software, and needs someone with data science and engineering experience to get customers on-boarded. (NYC)
• A well-funded startup is building a platform to identify compliance risks associated with both human- and AI-generated outputs. They are looking for a cloud infrastructure engineer to join their team of world-class researchers. (NYC)
• A seed-stage startup is using blockchains to enforce commitments and is in need of a fullstack developer with Solidity experience. (Remote)
• A company building the new pension of the 21st century and building universal basic capital is looking for fullstack engineers with prior experience in fintech. (NYC)
• A startup building the world’s most performant parallel-EVM is looking for a low level engineer with C++/Rust/CUDA experience who wants to deploy their GPU-EVM core to web3 projects. (Remote, EU preferred)

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.