The gap between lending and capital.

Every functioning credit market has a shape. At one extreme sit the instruments designed for velocity: credit cards, buy-now-pay-later, payday advances, the thin-margin consumer products whose underlying unit economics depend on processing millions of tickets of a few hundred dollars each. At the other extreme sit the instruments designed for scale: broadly syndicated leveraged loans, collateralised loan obligations, investment-grade revolvers clearing in tranches of a billion dollars or more, trading on desks in London and New York with published indices and a deep secondary bid. Both ends are efficient in their own way. Both ends are, in the language of market microstructure, solved problems.

The question that has troubled credit allocators since at least the early 1990s is what happens in between. Specifically, what happens to the facilities that are too large to fit on a consumer balance sheet and too small to interest a syndicate desk — the $5 million warehouse line to a specialty auto lender in Ohio, the $75 million receivables facility backing a medical-equipment leasing company, the $250 million trade-finance programme for a commodities merchant in Houston. Call it the middle market, though the term has been stretched thin by use. Call it, more precisely, the institutional structured-credit middle: roughly $5 million to $500 million per facility, bespoke in structure, held to maturity, and — until quite recently — almost entirely invisible from the outside.

Two ends that work, and a middle that does not

The two functioning ends of the credit market are functioning for reasons worth naming. At the small end, the underwriting problem has been reduced to a statistical one. A consumer lender does not care who you are; it cares about the distribution of outcomes across a million borrowers who look roughly like you. FICO, transaction histories, merchant categories — the inputs are standardised and machine-readable, and the business is, in the end, a question of loss curves and marketing spend. At the large end, the problem is different but also standardised. A syndicated loan to a $3 billion EBITDA sponsor-backed company has covenants drawn from a small library of market-standard templates; the leads (JPMorgan, Goldman, Jefferies) underwrite, distribute across fifty or a hundred institutional buyers, and the LSTA quotes the paper daily. Risk is priced because risk is visible.

Between these two poles, however, lies a vast territory of credit that looks like neither. Equipment leasing companies financing construction fleets. Specialty finance originators writing small-business loans through merchant channels. Consumer auto lenders serving the sub-prime and near-prime segments. Trade-finance operators moving working capital across borders for mid-sized importers. Warehouse lenders providing the intermediate funding that lets non-bank originators grow before they securitise. These are, collectively, the plumbing of the real economy — the mechanism by which a forklift gets bought, an invoice gets factored, a seasonal inventory build gets funded. By most estimates the stock of US structured-credit assets addressing this middle now sits somewhere north of $1.5 trillion, and the flow is growing faster than the syndicated market it is quietly replacing.

And yet the middle remains structurally underbuilt. Spreads are wider than the risk profile would suggest, not because the credits are bad but because the friction of investing in them is high. Secondary markets are essentially non-existent: a $40 million warehouse facility to a specialty lender in Dallas is held to maturity because there is nowhere to sell it. Reporting is idiosyncratic, delivered in PDFs or Excel workbooks with schemas that vary from originator to originator. Diligence costs, amortised across a facility of that size, can eat fifty to a hundred basis points of annual return. The middle is wide, it is deep, and it is — for most institutional allocators — effectively unreachable at scale.

Why standardisation is not the answer

The obvious response, and the one the market has tried for thirty years, is to make the middle look more like the big end. Rate it. Template it. Pool it into a CLO-shaped vehicle and sell tranches. This works, after a fashion, at the top of the middle — the upper end of direct lending, where $200 million unitranche facilities to sponsor-backed companies now trade with something approaching syndicated-market conventions. But it works by excluding most of what makes the middle interesting. A specialty auto lender's warehouse line is not a unitranche facility. Its collateral is a revolving pool of consumer loans with custom eligibility criteria; its advance rate depends on vintage performance data that no two originators report the same way; its covenants are negotiated facility by facility because the businesses are not fungible. Standardisation, applied here, does not reveal the credit. It erases it.

This is the deeper problem the institutional middle has had with every attempt to industrialise it. Howard Marks has written at length about the tension between efficiency and fidelity in credit markets — the observation that the most efficient markets are often the ones in which the least actual information is being priced, because everything has been normalised to a common grammar. The middle resists that grammar. Its value, to the originators who live in it and the allocators who understand it, is precisely the bespoke-ness: the ability to structure an advance rate around a specific obligor mix, to negotiate a trigger around a specific performance metric, to underwrite a specific business rather than a category. Kill the bespoke-ness and you kill the asset class.

What changes with programmable rails

The premise of River is that the tools now exist to do exactly this. A facility is, after all, a set of rules: eligibility criteria, concentration limits, advance rates, waterfall priorities, reporting cadences, covenant triggers. Historically these rules have lived in legal documents and spreadsheets, executed manually by servicers and agents, reconciled days or weeks after the fact. None of that is inherent to the economics of the facility; it is inherent to the paper-and-email medium in which the facility has been administered. Move the rules onto programmable rails — onchain where that makes sense, in deterministic software where it does not — and several things change at once.

• Reporting becomes real-time and schema-consistent, without the originator surrendering the custom structure of their facility.
• Eligibility checks, advance-rate calculations, and covenant tests run continuously against the underlying collateral, producing a verifiable state that any allocator can inspect.
• Secondary transfers become tractable: a standardised representation of a non-standardised asset is something a buyer can diligence in hours rather than weeks.
• Capital stack composition — senior, mezzanine, equity — can be assembled and re-assembled against the same underlying collateral as investor appetite shifts.
• Diligence cost falls by an order of magnitude, because the facility itself is the source of truth rather than a derivative of it.

None of this requires the underlying credit to be anything other than what it is. The Ohio auto lender still writes the same loans to the same obligors under the same eligibility box; the trade-finance programme still funds the same working-capital cycle. What changes is the substrate on which the facility is administered and the transparency with which it can be observed. A $50 million warehouse line on programmable rails is not a different asset from the same line administered on paper. It is the same asset, legible.

Why the middle matters

It is tempting, in conversations about financial infrastructure, to treat the middle as a technical curiosity — a gap to be closed because gaps in markets are aesthetically unpleasing. That undersells the stakes. The middle is where the real-economy transmission of capital happens: the point at which institutional savings meet the working-capital needs of the businesses that employ most Americans. When the middle is underbuilt, capital sits in Treasuries and the top of the syndicated stack while small and mid-sized originators pay twelve percent for growth capital they ought to access at eight. The spread between those two numbers is a tax on productive economic activity, and it is paid, ultimately, by the firms and workers who cannot reach the capital markets directly.

Closing that gap is not a matter of writing more structured-credit funds, or raising larger direct-lending vehicles, or extending the reach of existing BDCs another hundred billion dollars. Those are all useful, and none of them address the underlying structural problem, which is that the infrastructure for administering and observing middle-market credit has not kept pace with the scale of the asset class. The next decade of institutional structured credit will be defined, we think, by whoever solves that infrastructure problem — by whoever can deliver syndicate-grade transparency to bespoke facilities without flattening them into something they are not. River is built for that thesis. The middle has always been the hard part. It is also, by some distance, the most important part left to build.