The Valuation Arbitrage Keiretsu Model - The Diff

The Valuation Arbitrage Keiretsu Model

There were some terms that American businesses borrowed from Japan on the way up in the 1980s, like kaizen. On the way down, there were other borrowings, like zaitech, or financial engineering.1 One term that sits halfway through is keiretsu, a set of companies that do business with one another and have interlocking share ownership. The keiretsu model had several perks:

  1. It meant that companies had an incentive to treat their suppliers and customers well, since they'd capture some of the upside. These relationships can get fairly bare-knuckled when the participants have different levels of access to capital: when US airlines were undercapitalized compared to aircraft manufacturers, the manufacturers knew that they could strike aggressive deals (delay payment for a year in exchange for increasing the price paid per plane, for example).
  2. It reduced the risk that a key supplier would get acquired. The seller of a critical component sometimes has untapped pricing power, and as long as that supplier's customers collectively own most of their stock, such pricing power will remain untapped.
  3. Interlocking ownership makes it hard to value companies. If company A owns part of company B and company B owns part of company A, or companies A through Z have a similar set of overlapping ownership, then in economic terms the number of shares each company has outstanding is overstated. But it's hard to calculate by how much, and there is a difference between a company having 50% fewer shares outstanding and a company that owns shares of other companies that collectively own 50% of it. In the latter case, it's possible that investors won't see that money for a long time, if ever.
  4. If one part of the supply chain gets exciting to investors, the rest of the keiretsu can allocate more profits to it in order to increase the overall entity's value. If investors put a low P/E on banks and a high P/E on manufacturers, then a keiretsu with both banking and manufacturing holdings can maximize its market value by having the banks make more low-interest loans to manufacturers. For example, if banks trade at 10x earnings and manufacturers trade at 25x, then every dollar of profits that can be reported by a manufacturer at the expense of a bank creates an additional $15 in market value.

This model is starting to be recreated in the tech sector. For example:

Part of what makes these deals work is that the entity reporting revenue has some kind of business—cloud computing, enterprise software, ads—that produces repeatable high-margin revenue and has a market size that's hard to calculate. Palantir did $1.4bn in revenue last year, and given the flexible nature of their software it's hard to argue that they couldn't at some point 10x that.

(The big argument for them is over whether they'll keep reporting deeply negative EBITDA margins—interestingly, the company generates positive operating cash flow, because stock-based comp was 60% of revenue in the last year, and spent $207m on investing in other companies. In other words, Palantir is systematically selling its own stock and investing in other companies that increase the value of its stock, and doing so in a way that currently generates net cash.)

The deeper question is whether or not this creates any real value. In one sense, it's easy to argue that this is pure financial engineering, creating an ongoing cycle of manufacturing cash from an expensive stock and then using that cash to buy revenue growth. That could apply to Palantir, but it doesn't work for the other companies mentioned: Alphabet, Oracle, Amazon, and Microsoft are all throwing off gobs of cash.

The "steelman" argument for this as a way to create value goes like this:

  • There are basically two kinds of companies out there: the ones that have to think carefully about capital allocation because cash is scarce and opportunities are abundant, and the ones that have to think carefully because cash is abundant and and good investment opportunities are scarce. Companies in the latter category can make some incredibly stupid decisions (witness cyclicals making acquisitions at the peak of the cycle). But they can also afford to make investments that have a high overall return but a significant upfront cost. This can be because the investment target is something that doesn't generate reliable cash flow, because it's subscale, or because it's a cash-generative business that's highly levered.
  • These deals essentially allow a company to temporarily make its customer less cash-constrained for the purpose of making a single good decision. Selling cloud services has large and hard-to-quantify long-term upside, but so does buying it. Turning scaling into Google's or Oracle's problem instead of Univision's or Cerner's is a good use of resources on both sides of the transaction.
  • This also analogizes to the development cycle of economies. An industrializing country consumes capital, which it can either source the hard way through exports and financial repression, or the relatively easy way through foreign direct investment. Meanwhile, in mature economies the available returns are lower, but there's a lot more capital and a better-developed financial system.

As long as the transactions themselves create value, this seems to be an ideal way to structure them. The companies that need cash don't have to produce as much of it upfront, while the companies with more money than they know what to do with can find a new capital sink to put those funds to work. As it turns out, at least for the moment, one of the cost-effective ways to find new customers for cloud computing or analytics software is to find companies that could use the products and then write them a big enough check that they can afford them.