Two Games

There are two ways to build a large business. The first way is to build a valuable company in a massive market. This path is fairly well known: you must build something people want, your company must be protected against competitive arbitrage (it has a ‘moat’); typically there has to be an opening in the environment to enable the company to grow. This opening may look like any number of things: it may look like a new, overlooked market that eventually becomes big, or a change in consumer preferences, or a new technology that didn’t exist before. If markets are static and no change exists, there should not be any opportunity for a new company to grow large — incumbents will crush it. Finding the right opening demands some amount of luck, something I’ve written about elsewhere.

The second way is to build a conglomerate.

I am using the term ‘conglomerate’ loosely here. Some folks prefer ‘platform company’ or ‘rollup’ to describe specific instantiations of the conglomerate model. This distinction is not important to our discussion. Defined loosely, we’ll use ‘conglomerate’ to mean any set of businesses that are held in a structure where the capital in one business may be reallocated and used in any of the other businesses, due to central control. Berkshire Hathaway is a conglomerate, as is LVMH, and Koch Industries, and Samsung, and the Graham Holdings Company, and City Developments Limited, and Reliance Industries and so on.

If you compare these two paths:

1. The way that an individual business gets large is that it must pick the right opening, in a large market, at the right time, and then build a moat to fend off competitors.
2. The way that a conglomerate gets large is that it successfully acquires (or starts) multiple valuable businesses over a long period of time.

Many people talk about the first path. Fewer people talk about the second. It is my intention to discuss the second path in this letter; it is the path that most Asian tycoons follow when building their empires. It is the path I am most interested in following. It is the path that Postcognito is on.

If you study conglomerates for long enough, you eventually bump into three big ideas.

1. First, that conglomerates are good for owners-operators, even if they are often regarded as bad for minority investors.
2. Second, that moats at the individual business level are important.
3. Third, that moats at the conglomerate level are important.

We shall go through these points briefly, before getting to our core point.

Conglomerates Are Good for Owner-Operators

Conglomerates are seen as bad for minority investors because of the conglomerate discount. This is the phenomenon where markets value a group of unrelated businesses as less than the sum of its parts. The discount typically occurs due to a belief that the conglomerate will not be able to run the businesses as well as a focused company would; also, that most serial acquirers lack the ability to generate a positive return from their acquisitions. This is broadly true.

The discount is valid if you are a public markets investor in the West, and much of the writing on the conglomerate discount is from the perspective of such an investor. But if you are an owner operator, the discount is not as important. In exchange for a valuation haircut in certain public markets, you gain the advantage of increased business resilience. This is not a small thing! Conglomerates offer clear advantages to the businessperson: if you own a collection of businesses, you are diversified. Your business fate is no longer dependent on the secular destiny of any one particular industry or location; you are spread across many geographies and markets.

This is especially valuable once you understand that even the strongest businesses with the largest moats may fall on hard times. For multiple decades, the newspaper business was seen as having this impregnable fortress of a business model. But then the unimaginable happened. When the Washington Post declined as the result of Internet disruption, Donald Graham, heir to the Post family and son of renowned WaPo publisher Katharine Graham, gritted his teeth and sold the newspaper to Jeff Bezos of Amazon. Fortunately, in the final decade of Katharine’s tenure as leader of the WaPo, she had turned the company into a conglomerate — harvesting its considerable cash flow for disciplined acquisition of other businesses. Donald kept the rest of the companies that made up the original Washington Post Company, and renamed it the Graham Holdings Company. This body of collected businesses continue to do exceedingly well without its flagship paper.

Of course, there is much to say about conglomerate structure. It is possible to be a badly-run conglomerate; in fact there is a rich set of histories — if you know where to look — of horribly-run conglomerates from the 1960s and earlier. The best performing conglomerates today are almost always decentralised, with much redundancy in the organisation, and loose (but disciplined) control from headquarters. The vast majority of bad conglomerates — tracing all the way back to the aforementioned US conglomerate boom of the 60s — stem from bad control structures and sloppy incentive design. A serious study of good (and bad) conglomerate structures, along with the incentive design of such structures, is left as an exercise for the alert reader.

(I should note that even the notion of what ‘good’ looks like is worth discussing, since conglomerate structures in the East differ from the West. Perhaps this will be the topic of a future letter.)

Moats at the Individual Company Level Matter

The second big idea is simple: conglomerates grow either by buying other companies, or by starting new businesses of their own. In either case, they should own good businesses.

What is a good business? The technical answer is that good businesses should generate a return above every dollar of invested capital over the long term. What you are doing when you buy (or build) a new company is that you are acquiring a stream of future cash flows. To generate these cash flows, you have to put money into the business, on top of whatever you paid to buy or create it. To justify this investment, the business’s future cash flows should either grow over time, or at least not diminish. You don’t want to invest in an asset whose cash flows only diminish over time.

How do such cash flows shrink? The answer is the central problem of business. The name that I like to use for this is ‘competitive arbitrage’: you discover a lucrative opportunity, make a lot of money, and then everyone sees that you’re making a lot of money and copies you but with cheaper prices. This repeats as other people copy them, until everyone’s profits are destroyed and returns in the entire sector are ‘competed down to the opportunity cost of capital’ — that is, until no investor is willing to put money into such businesses.

The tricky thing about competitive arbitrage is that it can take a long time to play out. If you opportunistically chase business opportunities only according to how lucrative they are, not how protected they are, you will find yourself subject to such competitive forces, perhaps over the course of decades. This means you can waste a significant portion of your life. It is therefore important to study instances where the process is accelerated, in order to avoid it. To cite one recent example, I found it interesting to study the phenomenon where direct-to-consumer e-commerce rollups have failed. The majority of D2C brands have no moat, and therefore no defence against competitive arbitrage. Attempts to collect such businesses on leverage, so shiny for a brief moment, have resulted in terrible outcomes for the investors involved.

Good businesses are businesses that find some way to resist competitive arbtirage. When they do so, they are said to have a ‘moat’. It’s clear that having a moat at the individual business level is important — as a conglomerator, you don’t want to buy assets that become less valuable over time.

Moats at the Conglomerate Level Matter

Moats also matter at the conglomerate level. A conglomerate competes in a market of acquirers for the businesses it wants to buy (or build). For a conglomerate to become large, it, too, must have an advantage against its competitors in the same way that its individual businesses must be able to defend against their competitors.

It is instructive to study what the great conglomerates have as their moats. In some cases, such as with Samsung and with many other Asian conglomerates, the competitive advantage is the controlling family’s close ties with a country’s political power structure.

But let’s take Berkshire Hathaway as an example. The core advantage that Berkshire had in its intermediate years was access to insurance float. In other words, Buffett had access to capital at a cost far lower than any other acquirer on the market; the ‘interest rate’ of internal float was lower than the cost of borrowing capital from external sources.

If this insight is well known — which it is — why has the advantage persisted? The answer is simple: it is extremely difficult to run an insurance company in the way that Buffett has done; most insurance companies are not run for the benefit of float. In some ways, the history of Berkshire Hathaway may be read as the discovery of this insight (Buffett, in the words of Charlie Munger to his friend Rick Guerin in 1963, “knew more about float than anyone” years before he even bought into Berkshire), followed by a few years of experimenting with float in banks vs in insurance companies, followed by a decade of missteps in insurance whilst getting the strategy to work, followed by the discovery of Ajit Jain through an executive search firm in 1986, followed by two beautiful decades of having gotten the strategy to work and reaping its benefits.

There are not many Ajit Jains in the world. Nor are there many insurance companies that are run in the way that Berkshire runs theirs. For instance, Buffett has pointed out that there are periods in super-cat insurance where underwriting new policies is simply unprofitable, and therefore the solution is to have everyone ‘play golf for a few years’. This is only possible if employees have no fear of losing their jobs during this period. This is a structural advantage. No other insurance company can do as the Berkshire insurance companies do, because they are not owned by Berkshire.

In the terminology of 7 Powers, Ajit Jain is a Cornered Resource, and the structural advantage is a form of Counter-Positioning. Together, these advantages — in addition to a handful of others I have elided for the sake of brevity — have protected Berkshire’s growth and made the conglomerate more successful than its creators have thought possible.

Contrast this story with another successful conglomerate, Constellation Software. Constellation’s key insight, starting in 1995, was that businesses that provided mission-critical software to specific industries were excellent businesses to buy. Sure, many of these businesses could not grow as much, since they had maxed out in whatever small markets they were in. But the capital demands of the software business are negligible; their markets are small enough that no large competitor would think to enter; many of these businesses completely dominate their niches, and — most importantly — these companies made software that could never, ever, be ripped out of their customers. (We are talking about the software that runs nuclear power plants, or metro systems, or dental offices.) You could say that the cash flows of these special types of software businesses were better and stronger and more guaranteed than first lien debt.

Constellation took this insight and built a large conglomerate around it. They bought these high quality businesses, harvested their cash flows, and then used that cash flow to buy more such businesses, and on and on in an ever compounding loop. The conglomerate went public in 2006, and then proceeded to compound at an average rate of 34% a year for the next 17 years.

But unlike Berkshire, Constellation had no moat at the conglomerate level. And so Mark Leonard, its founder, eventually stopped writing annual letters for fear of copycats, and began to say in their annual shareholder meetings: “the only thing you need to get started in this (software conglomerate) business is a phone and a cheque book.” CSI has been struggling to keep up its rate of growth in recent years. There is now a name for the software businesses that they buy: “vertical market software” — which is a sure sign that Leonard’s original insight is now fully legible; private equity has become a huge player for such companies. Needless to say, there is a lot of competition in the space.

But: for a glorious period, Constellation had the field almost entirely to themselves. They grew large on that wonderful run.

The Game

What is the shape of the game of conglomeration? The game goes something like this:

1. You find, raise, or build an initial pot of capital.
2. You use it to acquire businesses that are protected, with moats.
3. You harvest cash from the businesses that you own to purchase more companies.
4. Rinse and repeat.

There are variations on this form but this is the basic shape. You will find many people who will tell you that they want to build a ‘holdco’, and many of them will cite this basic formula.

But actually this is a lie. If this were the basic shape, everyone would be able to play this game and build large companies. This is clearly not the case; your average holdco of small companies do not ever become very large. No, the real game is this:

1. You find, raise, or build an initial pot of capital.
2. You acquire long duration, cash generative assets that are overlooked. It isn’t enough to acquire businesses — anyone can do that. What you need to do is to find a secret — something that throws off cash over a suitably long time horizon that is illegible (or mispriced) by other acquirers. It is no accident that Buffett was obsessed with float through the 60s, and then took a full decade to figure out how to use it; Constellation wouldn’t nearly be as successful if Leonard did not have his observation about vertical market software companies in the 90s (when most investors did not understand the dynamics of the software business).
3. You exploit that initial period of misunderstanding to grow. And then hopefully you build a moat at the conglomerate level.
4. You end up with a big business.

A logical conclusion here is that it is not intelligent to talk about the secret. If you’ve found one, you should move quietly as you grow, whilst looking for ways to build a conglomerate-level moat. Such a moat is not always possible.

But perhaps that is not necessary: by the time others have found out, you’d have built a large business already. The lack of a conglomerate-level moat is only a problem for your ability to acquire new businesses. The quality of each individual business in your portfolio is what determines if your conglomerate can die.

And this is why the conglomerate model is so powerful. It is very hard to kill a business that is constructed of many different long-duration, moat-protected assets. Conglomerates tend to last for a long time because they are diversified; they tend to produce generational wealth for the folks involved.

Which is why the game is worth playing in the first place.

Cedric Chin
26 January 2025