The Modern Private Equity Business was Invented in Beverly Hills in the 70s

In this issue:

• The Modern Private Equity Business was Invented in Beverly Hills in the 70s—The PE business has evolved from making directional bets on equity to being in the business of providing capital more generally. This evolution ends up recreating some of the economic benefits of the original high-yield bond market, where a well-capitalized and exceptionally well-informed central intermediary kept the rest of the industry running smoothly.
• Product Convergence—TikTok experiments with longer-but-not-too-long videos.
• Musk v. Altman—Tech history is contingent.
• The Bundle—Be careful offering consumers too good a deal.
• Financing a Boom—AI offers opportunities on the credit and equity side.
• Investor Protectionism—Industries go to wherever they have the biggest advantage, and for capital-intensive technology companies, that's the US.

The Diff November 18th 2024 17:21 • Click to play audio

The Modern Private Equity Business was Invented in Beverly Hills in the 70s

Finance combines abstraction and regulation in a way that leads to messy semantic drift: a term can keep its literal denotation while completely changing its denotation. A "hedge fund" in the mid-80s meant a highly levered investment partnership whose manager placed big bets on currencies, commodities, and interest rates. There are still such funds today, but a growing share of what hedge funds do is almost the exact opposite—quantitatively rigorous portfolio construction designed to isolate idiosyncratic return without any unnecessary exposure to easily-hedged risks. What's remained consistent is that it's structured as XYZ Capital LP, whose limited partnership is XYZ Capital Management LLC. Everything else is pretty different.

Private equity has also gone through an evolution of its own: the earliest deals were, by modern standards, comically leveraged, with 10:1 being a pretty common number. They were often for fairly mediocre companies that would be sold for parts, but these early deals were also for incredibly cheap companies; it's perfectly fine to buy a marginally-profitable company in a declining industry with a lot of leverage if you can get back most of the capital you put into the deal in a busy year of liquidations.

That model worked well for a while. There were some poorly-thought-out conglomerates that had been assembled in the 60s and 70s and needed a few amputations to stay viable in the 80s. Assets were cheap, though capital was expensive, but rates declined in the 80s, making those assets more valuable and making it easier for PE firms to refinance debt. Over time, though, the world ran short of undiscovered companies, and PE had to switch to a different model: find reasonably good companies that could be great companies, figure out the right model for them, own the decent ones for a while and the great ones for as long as possible. They also refined their revenue model: over time, private equity firms develop expertise in credit, both because they need a good deal when they're making the initial acquisition and because they need to deal with messy recapitalizations when things don't work out. So these firms branched out, and started doing lending and underwriting. They acquired more stable sources of capital, particularly life insurance capital that was a perfect match for long-term, illiquid fixed income products.

So the PE model today looks like this:

1. Identify a set of companies where you can add value through both management advice and rearranging the capital structure.
2. Identify a broad financial ecosystem that will be a good buyer for the parts you don't want to keep, with varying proximity between these capital providers and the PE firm.
3. If something goes wrong and a company turns out to be over-levered, don't fret&dmash;recapitalize! As long as the company is fundamentally okay and just going through temporary issues, the current owner has an advantage in figuring out and funding whatever the revised capital structure is.
4. Have enough of a balance sheet to keep whatever slice of the business is most attractive. Run multiple pools of capital, align people's compensation as much as possible with their own ability to produce, but lock up key employees' capital in deals for a long time so they're not tempted to quit.

This is the direction the industry has drifted over time, for many reasons: changes in regulation after the financial crisis made banks worse at warehousing risk; those same changes meant that banks paid less and more of their best people moved to PE; economic cycles create plenty of entry and exit opportunities for funds that have access to capital over a decade; and the growth of SaaS means that there's a larger set of companies that have predictable cash flows that can be levered up while the underlying business continues to grow.

But this business isn't just a productof circumstances, because a fairly similar one existed in a completely different set of circumstances: this high-level model is a decent description of how Drexel Burnham Lambert dominated the high-yield bond market in the 70s and 80s. All of the pieces were there, as were some of the players (Apollo Global was founded shortly after Drexel's collapse, by Drexel alumni) and cultural norms (from Apollo's investor day last month: "We have woken the team up at 4:30 in the morning for meetings to prove to them that we need to do a wake-up call and something different").

The Drexel story starts with making a market. And, in this case, really making one. Corporate bonds trade less than equities, for a variety of reasons, and prior to Drexel, bonds rated lower than investment grade traded even less. The typical bond buyer was aiming to get some slight excess return while preserving their capital, and any time a company's bonds were downgraded to the point that the ratings agencies thought that 100 cents on the dollar was not the most likely outcome, they weren't serving their purpose. So a holder could mentally write those bonds down to zero, and hope that they bounced back, or sell them and perhaps recover a bit of the initial investment. But who, in the bond world, would be interested in buying something so sensitive to credit—an asset that was legally a bond but that behaved more like a stock?

There weren't very many people willing to do this, but Michael Milken was one of them. Interestingly enough, he'd learned about high yield bonds from an early version of what today's investors would call a quant, economist W. Braddock Hickman, who compiled an exhaustive study of bond performance since the turn of the twentieth century, and concluded that low-rated bonds paid investors far more than they needed to earn in order to be compensated for risk. Any one issue could be a problem, and would certainly provide a less certain outcome than the average bond investor expected. But a broad cross-section of the low-rated market could do well. Specifically, not only did buyers of low-rated bonds earn returns in the teens when their bonds didn't default, they actually earned more than investment-grade bond investors when bonds did default; the higher interest payments more than offset the loss from defaults.^[1]

Milken would have been happy just buying these bonds, but he was a small part of a mid-sized firm, so when he started out he had all of $500k in capital, and a deal that the department's bonus pool was 35% of whatever he could make from it. He might have done fine buying and holding, but could do a bit better trading: there were a handful of investors who were open to buying high-yield bonds, and Milken spent substantial time early in his career evangelizing them. (Later on, he didn't need to do nearly as much of this, and instead would have companies come to Drexel and take a meeting before the real workday started. Since Drexel was based on the West Coast but operated on Eastern Time, this meant flying into LA and showing up for a morning meeting at 4am or so. But by this time it also meant a good chance of finding out, at that meeting, that not only could Drexel raise the hundreds of millions you wanted, but they wanted you to raise even more.)

The early high-yield market consisted of formerly investment-grade bonds that had fallen from grace, and of bonds that had been created for conglomerates' exchange offers.^[2] Natural buyers didn't really exist, but there was enough generational turnover in the financial industry to pitch the idea. And the early results spoke for themselves: there were lots of cheap bonds out there, a few willing buyers, and one central node connecting all of them.

Part of the appeal of bonds is that they're a specific, quantified promise: this much every six months, this much at maturity. So junk bonds are a great space for contrarians: no matter how wrong the rest of the world thinks you are, bonds naturally drift towards their fair value, and there's a date on which you know that either you or your counterparty gets vindicated.^[3] That dynamic puts a premium on research. A market-maker trading in an illiquid category will sometimes end up with inventory because there simply isn't a buyer, and the better they know their inventory the more comfort they have in both holding it and pitching it externally.

So that's Drexel 1.0: central node in a growing network of bond traders who are willing to touch the junkier products, but need someone to provide liquidity. Over time, Drexel started tracking positions closely; their high market share meant that they knew roughly who owned what, and had a sense of when they'd be willing to buy more or sell some. That advantage compounds: a market-maker who has a buyer in mind can offer better pricing, so they get the deal and collect both a spread and a bit more data to inform the next deal.

Things got more interesting once Drexel started doing new junk bond issues, too. These were typically small, had high interest rates, and often had weird terms. But they were a business necessity, because Milken had solved the supply overhang problem so well that he'd created a shortage of junk bonds. This would have struck investors just a few years earlier as a bizarre possibility, like a global shortage of pocket lint, but it happened: the bond funds that bought junk did better than the rest, and as rates declined and the typical bond no longer offered a double-digit return by default, investors reached for yield instead of accepting lower returns.

When Drexel did these deals, it negotiated some onerous terms: underwriting fees were multiples of what they were for investment-grade debt, and the firm often asked for warrants in order to place the debt. These warrants were ostensibly something Drexel had to offer customers—if they're taking the risk, they need some upside to compensate. But in practice the company sometimes kept them.

Or, to be more specific: Milken's group kept them, either on behalf of the firm or for various partnerships that Milken or close associates operated, which managed money on behalf of him, other Drexel employees, and favored outside clients.^[4]

Who was issuing these bonds? Sometimes, it was exactly who you'd expect: a company that used to issue investment-grade bonds, had seen a deterioration in its business, and needed more capital to keep going even if it had to pay high rates. There were also growth companies, like MCI and Golden Nugget; paying 16% for debt is painful, but there were times when equity valuation wasn't too far off from that; when every kind of capital is painfully expensive, it's just a question of how much of the upside you want to keep if your business works out.

One practice Drexel came up with was to first convince companies to raise more than they felt that they needed, and then to get them to invest the surplus in other high-yield bonds. This gives the companies some flexibility, but also means that they're paying yet another data dividend to Drexel, in addition to paying those 3-4% underwriting fees on a bigger deal.

Put all of this together, and you have a suspiciously familiar financial ecosystem: there's a central node that 1) has a good fundamental understanding of the businesses it's involved in, 2) has essentially unlimited creativity for crafting new financial solutions if the old ones don't work out, and 3) has enough of its own capital to own the highest-upside piece of every deal and enough captive outside capital to place all the paper that gets sold.

This is a very profitable way to dominate a market, though the cost of this is that you have to choose the right market to get obsessed with, and sometimes that means having the correctly quirky reading habits while you're in school. It does run risks, because there can be imbalances on the supply or demand sides, and because that central liquidity provider can quickly transmit risks throughout the system. But it has some general advantages, too: it creates an economic incentive to deeply understand companies, because distressed credit performs more like equity than like debt, and because these companies will keep coming back for more financing. And there's an incentive to understand how other investors think, to know what's going to keep them involved in the market and what's going to make them run away screaming. It's a subsidy for operational excellence, because the speed at which the dominant intermediary operates sets the pace for the entire industry. Combining all of these makes markets more efficient, and it pays extraordinarily well.

It's also a model with natural conflicts. You have one party who's well-informed, and who exercises more discretion than everyone else: clients do the best trades they can, but the very best ones are trades they won't even see. Negotiating complex deals as both a principal and agent can leave plenty of profit for all sides when bid/ask spreads are wide, but the more asymmetric the information is, the longer those valuation gaps take to close. It's a fantastically profitable approach, but it requires inputs: not just capital and talent, but an accumulated stock of either trust or owed favors that can be tapped at will. Counterparties know that they're dealing with someone well-informed who aims to make a good profit from the transaction, and they're willing to do this if they trust their counterparty to make the right moves. But this kind of leverage is hard to measure, and like other varieties it cuts both ways.

Diff Jobs

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

• A growing pod at a multi-manager platform is looking for new quantitative researchers, no prior finance experience necessary, 250k+ (NYC)
• An AI startup building tools to help automate compliance for companies in highly regulated industries is looking for a director of information security and compliance with 5+ years of info sec related experience at a software company. Experience with HIPAA, FedRAMP a plus. (NYC)
• YC-backed, post-revenue AI company that’s turning body cam footage into complete police reports seeks a senior founding engineer/tech lead who can build scalable backend systems and maintain best practices for the engineering org. (SF)
• A Google Ventures backed startup founded by SpaceX engineers building data infrastructure and tooling for hardware companies is looking for a staff product manager with 5+ years experience, ideally with AI and data intensive products. (LA, Hybrid)
• A well-funded startup that’s building the universal electronic cash system by taking stablecoins from edge cases to the mainstream is looking for a senior full-stack engineer. Experience with information dense front-ends is a strong plus. (NYC, London, Singapore)

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.

Elsewhere

Product Convergence

TikTok is testing out a compromise between short clips and full-length videos: Quibi-style shorts ($, The Information). Quibi's high-profile failure was a victim of many forces: they raised enormous funding and did a splashy, heavily-marketed launch, which was implicitly a bet that they'd found a great format that needed to scale; they made some unforced errors in their app, particularly around sharing, which meant that low-quality Quibi clips were going viral on competitors' platforms; and the reason those clips went viral was that the shows were not very good. (One benefit of starting small as a media enterprise is that the poor quality of your initial content doesn't get seen by many people; when something takes off, that's the first impression new users get.) The format itself might make sense, and given how badly Quibi failed, it's a format that might be underpriced. There are some stories that can't be edited down to thirty seconds, but also aren't worth a full sixty minutes of screen time. And even if most of them are boring to most potential audience members, TikTok is no slouch at figuring out what users want and then delivering as much of it as they can handle.

Musk v. Altman

The internal OpenAI emails revealed in the OpenA suit are presented here in a clean, readable format. What stands out is how contingent all this history is. Cerebras could have been an OpenAI subsidiary, OpenAI could have merged with Tesla, an OpenAI cryptocurrency could have been a real thing rather than an the monetization angle for a repeated series of embarrassing Twitter account hacks, a lot of bad blood could have been avoided if the cost of training frontier models had been more apparent in advance (or would OpenAI simply not have existed if they'd know what the future economics looked like but been unaware of the future funding environment?). These decisions all got made fast, mostly by people who had other day jobs to worry about. That tends to be a surprisingly common feature of projects like this: they're marginal enough that people make quick decisions without worrying too much about the long-term issues, or optimizing for the one long-term goal of creating AI that isn't controlled by Google. And as they get more important, these awkward structuring details get cleaned up. OpenAI scaled too fast, both technologically and as a business, and didn't have time to fix such issues before they became bigger problems.

The Bundle

Meta has been fined almost €800m by the EU for bundling marketplace with its Facebook product ($, FT). What's true about this complaint is that Facebook Marketplace benefits from its placement in the Facebook app, and that its monetization model partly stems from giving low-engagement users one more excuse to be weekly actives even if they no longer necessarily log in daily. It's a testament to the quality of Meta's ad targeting that random newsfeed ads monetize well enough that it makes sense for them to under-monetize a classified ads site, given that classifieds have been such a durable moneymaker online. All of this is bad news for pure-play classified ads providers, who can't hope to compete with that on price. But it's net good for consumers. In fact, the wealth transfer is that marketplace produces a consumer surplus, some of which is captured by time spent on Facebook, which allows Meta to capture more ad revenue. So Meta is taking a commission in exchange for using its advertisers to subsidize a bit of peer-to-peer commerce.

Disclosure: Long META.

Financing a Boom

The WSJ has a look at how banks and PE firms are thinking about AI. Obviously the first thing they're thinking is: there are lots of fees to be earned. But a secondary consideration is that financing AI bets requires some creativity. The long-term economics are unknown, and, if you're optimistic, unknowable, but in the short term there's some structuring to do: GPUs are an asset with a market price and a yield that can be roughly estimated, so there's a layer of the stack that's best financed with debt. And that means that even though the industry is capital-intensive, it's possible to structure deals so that the equity has more of a venture-style payoff.

Investor Protectionism

French telecom billionaire turned venture investor Xavier Niel is asking European AI companies not to sell to US competitors ($, FT). That's relatively easy to say when these AI companies are small, because Europe has enough of a venture ecosystem to fund some early-stage deals, albeit on terms that would strike a Silicon Valley founder as onerous. But at later stages, it's just hard to provide enough capital to keep firms scaling, whereas US companies have enough capital to do that and make the occasional strategic acquisition as well. If AI ends up being vertically fragmented—company A's product powered by company B's model trained in company C's datacenter running on company E's chips—then this doesn't preclude the existence of European AI champions. But to the extent that there are benefits from vertical integration, the country that has the most capital and can ingest the most talent will win, and that country remains the US.