The financial services industry is built to intermediate. It matches people who need capital with the ones who have it and are looking for a return, or, in the case of zero-sum derivatives, it matches people who want to hedge a risk with people who either want to take that risk or have a corresponding risk to hedge. This isn't a problem in some cases, because each side's demand is pretty obvious: it's very easy to see why a neocloud might borrow money to buy GPUs, and also easy to see that the lender would appreciate the interest they earn. But in other cases, the risk is more one-sided.

Consider oil futures and mortgages, two products that are a disproportionate share of their broader asset class (oil makes up ~25% of commodities futures and mortgages ~20% of the bond market). The global market for oil continuously clears, in the sense that global storage capacity is equal to about a month and a half of production. So, in one sense, there's just as much of an aggregate incentive to hedge against oil price declines as there is to hedge against increases. You could theoretically buy oil futures to hedge against the cost of your commute, for example, and your indirect counterparty might be an oil company that's implicitly hedging against the risk that you'll use less. But if you try to do this practically, you run into problems: trading oil futures means hedging in thousand-barrel increments, which is a lot more oil than you personally consume. (At ~20 gallons of gasoline per barrel, someone driving a car that gets 27 miles per gallon is hedging a little over half a million miles of driving. Also, they're long diesel and jet fuel.) And even if you could buy a miniature contract that could hedge your economic risk, you now have a mark-to-market mismatch: if oil prices plunge, you'll get a margin call and need to put up more cash, even though your net worth hasn't changed because your hedging losses are made up for by your future savings. To keep the hedge, you'd essentially need to borrow against the future increase in disposable income you'd get from hedging.

30-year fixed mortgages with an option to refinance have a similar problem. As a borrower, you know exactly what your future payments will be, and you also know that if they change, it will be because you refinanced and they went down. But a lender is in a more annoying situation: they're buying a fixed-income asset, but unlike the typical case where a 30-year bond is worth more as rates decline, this one has a habit of getting repurchased at 100 cents on the dollar any time it's worth more. Whereas if you take a big immediate loss—you lent a homeowner money in early 2022, when rates were low, and now they're up and your loan is trading at 70 cents on the dollar—it's staying on your balance sheet.

Given that American homeowners retain access to mortgages, the system clearly works, but it raises the question of who wants to take that risk. And the answer is, generally, someone who is running some kind of fixed-income strategy that mostly targets duration risk, and is willing to continuously hedge, rearrange their investments, etc. in response to changes in the interest rate environment. A homeowner could decide that the prepayment option is an interest rates derivative that they, as a regular person, have no business speculating in, and they, too, could hedge it. But they have the same problem as the hypothetical oil trader, just over a longer timeframe. If you hedge the value of the prepayment option, and rates rise instead, your locked-in mortgage with predictable payments over the next thirty years suddenly requires an unpredictable amount of cash right now. And brokers are not especially excited about letting you cross-margin your mortgage liability and your treasury futures position, even if that is theoretically elegant.

There are other kinds of transactions that end up with some sort of leftover risk that nobody particularly wants. In fact, a lot of the industry can be modeled this way:

• Venture debt means lending to some of the most unpredictable companies in the world, and lenders don't like unpredictability much. Depending on the business, they can either cordon off the parts that are good collateral (loans secured by their biggest long-term contracts, for example) or just pair those loans with warrants that pay them for some of the volatility.

• In the CDO boom, there were many buyers who wanted a lift in rates in exchange for a little bit more credit risk, but these structures require an equity tranche that carries the most risk, pays the highest returns, and has the first loss. Sometimes that tranche ended up being owned by hedge funds (especially if those hedge funds used the returns from the equity piece to fund bets against the highest-rated tranches). There was an investor base for credit that yielded tens to hundreds of basis points more than treasuries, but there wasn't as developed a market for something that was halfway between debt and equity, didn't have an active market, and was also hard to analyze.

• Sometimes the residual risk is timing. Even if a market ultimately clears—over time almost every barrel of oil has a buyer and a seller, as does every unit of currency, share of stock etc.—it doesn't always clear instantaneously, and middlemen can earn a spread from trading in exchange for warehousing some of the risk they take.

• The venture studio model is, arguably, a way to warehouse extremely-early-company risk. VCs are reluctant to back a founder who knows they need a cofounder but hasn't found one yet, but an incubator can make that part of their value-add and part of the reason they deserve their equity.

• Professional services companies are at least partly in the business of warehousing talent—Bain, BCG, and McKinsey can all be confident that over the next year, corporate America will encounter many problems that temporarily need the help of 23-year-olds who are wizzes at PowerPoint and Excel (and, increasingly, Claude Code), but they don't want to source that talent on a just-in-time basis. So they buy big batches of talent upfront, and then provide talent liquidity to their customers. But they also run the risk of the talent sitting “on the beach” (underutilized) when things slow down, not dissimilar to a market maker warehousing stock they can’t immediately offload.

• Merger arbitrage, convertible bond arbitrage, and index rebalancing are all strategies that swoop in when a given asset is suddenly subject to a new set of forces. A few months ago, Warner Brothers Discovery was a bet on IP, a successful transition to streaming, and big box office successes. Now it's a bet on the balance sheets of Netflix and Larry Ellison and perhaps the whims of Donald Trump and US antitrust regulators. When a company gets added to or booted from an index, everyone who holds it has the choice of either a) getting up to speed on the latest advances in modeling the impact of these changes on financial flows (and also the latest advances in modeling how other people will model this) or b) they can just assume that the newly-prevailing price is as good an estimate of fair value as they're likely to get.¹

The places where intermediaries warehouse risk illustrate why this can be such a lucrative line of work. They have a classic problem: if they passively take on risk, they end up with peak inventory whenever the selling really gets going, and might have promised more than they can deliver when prices rise. This is yet another lens for understanding why prices are so volatile when markets drop fast: an oil trader's job is to find a buyer who is willing to accept a slightly worse price in exchange for certainty, and a seller who wants the same, and to take the risk that the buyer and seller don't show up at the same time. There's probably a price the seller could offer that would make it worth the buyer's time to go direct, and vice-versa. But that price is pretty far outside the narrow bid/ask spread established by the middlemen. If some of those middlemen have big losing positions they're in a hurry to exit, that warps the market even more. (Of course, not all of them do, and market-makers' profits tend to peak in the aggregate at extremes. But depending on the market, some of them will turn out to have executed a pivot to directional trading that lasts for the remainder of their career. Which is typically measured in weeks to hours.)

Some asset classes have natural holders. Pensions and endowments with predictable financing needs should be taking on the liquidity risk of PE and VC, at least as long as those asset classes can deliver good returns. Households saving for retirement should be cautious about opaque, high-fee vehicles, but will be basically unaffected if their 401(k) temporarily drops 40% when they're still two decades from retirement, especially since that's the cost of stocks delivering higher unlevered returns than other asset classes. For some weird assets, like weather derivatives, the natural holder is just someone looking for an uncorrelated return, so they have the chicken-and-egg problem that they need to be big and liquid enough to fit into an institutional portfolio, but won't get there unless they have counterparties. Whenever there's a new asset class, or a new kind of derivative like a prediction market, it's worth asking: who is the natural counterparty to the natural bettor. And, if that counterparty is just someone who wants to risk capital in order to get a return, where does their return come from and is the other side of the trade willing to pay them indefinitely?

Many pieces in The Diff are implicitly about this intermediation role, and where it goes wrong.

• We've looked in more depth at the question of how individual consumers should hedge ($), including the idea of renting as a natural hedge if you work in your town's biggest industry.

• How central banks monetize land ($).

• We've examined the CLO asset class, and how it does and doesn't pattern-match previous structured products.

• And we've looked at how neoclouds fit into AI financing.

• We've also considered how Affirm creates asset-backed securities that match investor preference by pairing very different kinds of loans ($).

• We’ve asked what happens when asset classes get democratized.

• And don’t miss where prediction markets fit into portfolios.

Join the discussion!

1  Interestingly enough, in merger arbitrage one of the things that sets the best performers apart is that they think like non-merger arbs, too. If you're betting on the completion of some media deal, or a biotech acquisition, or a mining merger, it helps to have enough real-world grounding to know when there's a chance of a bidding war.