The Securities Lending Machine: How a $3 Trillion Shadow Market Runs on Opacity

If you hold equities in a brokerage account — or through a pension fund, a mutual fund, or an ETF — there is a reasonable probability that those securities are being lent out tonight. Not by you. By whoever holds them in custody on your behalf. The borrower is likely a hedge fund executing a short sale. The collateral is sitting in a tri-party account managed by a third institution. Your custodian is collecting a fee for facilitating the transaction. You are receiving a fraction of that fee, if anything at all.

This is securities lending: a market worth roughly $3 trillion in outstanding loans on any given day, generating an estimated $10 billion annually in revenue for the intermediaries who operate it. It functions almost entirely in the dark. The beneficial owners — the pension beneficiaries, the ETF holders, the retail investors — rarely know when their assets are lent, to whom, on what terms, or how the revenue is split. They have delegated custody, and with it, they have delegated the economics.

The opacity is not incidental. It is structural. It is also, from the custodian's perspective, the point.

How the Machine Actually Works

Securities lending exists because short sellers need to borrow shares they don't own before selling them. To borrow a share, a hedge fund approaches a prime broker, who sources the share from a securities lending program. The securities lending program is typically operated by a large custodian — State Street, BNY Mellon, JPMorgan, BlackRock's Aladdin — acting as agent lender on behalf of institutional clients whose assets sit in its custody.

The mechanics are straightforward. The custodian lends the security to the borrower, who provides collateral — usually cash or high-quality government bonds — worth slightly more than the security's market value. The custodian reinvests the cash collateral in short-duration instruments and earns a spread. The borrower pays a fee for the loan. The custodian keeps somewhere between 20% and 50% of the gross revenue and passes the remainder to the beneficial owner. The rates, the split, and the collateral terms are all negotiated bilaterally, outside any public market.

The result is a market where prices are almost entirely opaque. A pension fund has no reliable way to know whether the lending rate its custodian negotiated for a particular security was competitive, because there is no public price feed. There is no exchange. There is no consolidated tape. The custodian knows every transaction that clears through its system, the beneficial owner knows only what appears on a quarterly statement, and the information asymmetry is built into the relationship's terms.

The Revenue Split Problem

The fee split between custodian and beneficial owner varies enormously and is rarely disclosed in a form that allows comparison. A large pension fund with negotiating leverage might retain 75 cents of every dollar generated by its lending program. A smaller institutional client using a pooled program might retain 50 cents. A retail investor whose broker participates in a lending program may receive a single-digit percentage, or nothing.

These splits are not secret, exactly — they appear in custody agreements — but they are negotiated without any reference to a market standard because there is no market standard. The custodian has the data. The custodian runs the program. The custodian sets the terms. The beneficial owner signs the agreement and hopes the revenue split is reasonable, without a reliable method for verifying that it is.

This is the trust problem at the core of securities lending. The arrangement works because beneficial owners have no practical alternative. Moving custody is expensive and disruptive. Running an independent securities lending program requires scale, operational infrastructure, and relationships that most institutional investors don't have. The custodian's position as essential intermediary is self-reinforcing: it controls access to the market, controls the pricing data, and earns a significant fraction of the revenue generated from assets it doesn't own.

What Blockchain Changes

Tokenized securities do not eliminate the need for securities lending. Short sellers will still need to borrow. The economics that make lending attractive to asset owners will still exist. What changes is the information architecture.

If a security exists as a token on a programmable ledger, a lending transaction can be executed directly between owner and borrower through a smart contract — code that automates the collateral transfer, the fee calculation, the revenue split, and the return of the security at loan maturity. Every term of the transaction is visible on-chain. Every fee payment is recorded. The revenue split is encoded into the contract rather than negotiated in private.

More consequentially, a public lending market for tokenized securities would generate real-time price discovery. The fee for borrowing a specific security on a specific day would be observable, comparable, and auditable. The information asymmetry that currently allows custodians to capture a disproportionate share of lending revenue evaporates when the market price is public.

Several projects are working in this direction. Maple Finance and Goldfinch have built on-chain credit markets that demonstrate the basic mechanics. More directly, firms including JPMorgan's Onyx division and Broadridge's distributed ledger platform have run experiments with tokenized repo and securities lending transactions, finding that the settlement friction and manual reconciliation that make the current system expensive can be substantially reduced on shared ledgers.

The Regulatory Wrinkle

Securities lending in its current form operates under a regulatory framework that was built around the intermediated model. In the United States, the Securities and Exchange Commission has historically treated securities lending as a custodial activity rather than a securities transaction, which has kept it outside the disclosure regime that applies to other forms of securities activity.

That is beginning to change. The SEC's Rule 10c-1a, adopted in 2023, requires reporting of securities lending transactions to a registered national securities association, which will make aggregate lending data more available. It is a modest step toward transparency, but it is also an acknowledgment that the current opacity is a problem worth addressing. The rule does not require disclosure of individual fee splits to beneficial owners, nor does it create a consolidated real-time tape of lending rates.

The European Securities and Markets Authority has taken a somewhat more aggressive posture through the Securities Financing Transactions Regulation, which requires transaction reporting and imposes disclosure obligations on fund managers. Even so, the data that flows to individual beneficial owners remains thin.

A tokenized lending market would resolve the disclosure problem structurally rather than regulatorily. When every transaction is on-chain and every fee is encoded in the contract, regulatory reporting becomes a consequence of how the system works rather than a compliance burden layered on top of an opaque private market.

Why the Custodians Are Ambivalent

The large custodians are not ignoring blockchain. BNY Mellon has a digital assets division. State Street has made investments in distributed ledger infrastructure. JPMorgan's Onyx platform has processed trillions in tokenized repo transactions. But there is a material difference between using distributed ledger technology to improve the efficiency of existing securities lending operations and building a transparent, peer-to-peer lending market that eliminates the custodian's informational advantage.

The experiments underway at major custodians are largely aimed at the former. Faster settlement, reduced reconciliation costs, and automated collateral management are real improvements — and they generate cost savings that the custodian can monetize. A genuinely open lending market, where rates are public and smart contracts replace agency relationships, attacks the revenue model directly.

Anyone who has spent time inside institutional asset management understands how this plays out. The technology gets adopted in ways that preserve the intermediary's value-add. The parts of the technology that threaten the revenue model get classified as not yet ready, too risky for institutional use, or lacking regulatory clarity. This is not cynicism — it is a rational institutional response to a genuinely threatening technology.

The Beneficial Owner's Leverage

The dynamic shifts if large beneficial owners — pension funds, sovereign wealth funds, large asset managers — decide that on-chain lending programs serve their interests better than custodian-operated agency programs. They have the negotiating leverage to demand it. They have the scale to make it work. And they have a fiduciary obligation to their own beneficiaries that arguably requires extracting the maximum available return from lending their assets.

The first institutional lending programs built on tokenized rails will likely be private, bilateral arrangements rather than open markets — the same pattern that characterized early electronic trading before exchange competition created genuine price discovery. But the direction is legible. Once a large pension fund runs a tokenized lending program and demonstrates that it can retain a higher fraction of the gross revenue with lower operational risk, the argument for the traditional agency model becomes harder to make.

The securities lending market is worth studying not because it is the largest financial market that blockchain will touch — it isn't — but because it illustrates something precise about how intermediation works. The custodian's revenue doesn't come from taking risk. It comes from controlling information. That is a different kind of moat from the ones that disappear when technology improves operational efficiency. It disappears when the information becomes public. That is exactly what a transparent ledger does.