BNY and State Street indemnify $1.13tn of securities lending off balance sheet

BNY and State Street disclosed a combined $1.133 trillion of securities-lending or securities-financing indemnifications at 30 September 2026. The figure is larger than either bank’s balance-sheet assets in several ordinary business lines, yet it does not represent a trillion-dollar cash obligation or a prediction of loss. It is the maximum contractual amount the banks could be required to cover if borrowing counterparties failed and the collateral available to the lending clients proved insufficient. Against those indemnities, the two banks reported $1.203 trillion of collateral, producing a combined disclosed cushion of about $69.7 billion, or collateral coverage of 106.15%.[1][2][3][4]

The scale is economically important because the indemnities sit outside the intuitive picture of a custody bank merely safeguarding assets. In an agency securities-lending programme, a beneficial owner lends securities through the bank to a borrower. The borrower posts collateral. The bank may indemnify the owner against a shortfall if the borrower defaults and the recovered collateral is not enough to replace the securities. That promise supports the lending market, but it transfers a defined tail risk to the agent. The gross contractual figure therefore measures the protected principal at risk under extreme conditions; collateral, daily revaluation, borrower underwriting and close-out mechanics determine the residual exposure.[1][3]

The two programmes now exceed $1.1 trillion

BNY’s indemnified amount was $706.034 billion, up $39.639 billion, or 5.95%, from $666.395 billion at the end of 2025. State Street’s protected securities-financing amount was $426.989 billion, up $55.021 billion, or 14.79%, from $371.968 billion. Together, the programmes expanded by $94.660 billion, or 9.12%, in nine months. The faster percentage growth at State Street does not by itself establish greater risk appetite: volumes can change with client participation, asset prices, borrower demand, collateral mix and the exact transactions included in each institution’s disclosure.[2][4]

The collateral ratios were strikingly similar—about 106.1% at BNY and 106.3% at State Street—but even that comparison has limits. A coverage ratio is a point-in-time aggregate. It says nothing by itself about the liquidity, volatility, concentration or legal enforceability of individual collateral pools. A programme could appear overcollateralised in aggregate while a particular borrower or security creates a more demanding close-out. Conversely, a large gross indemnification need not imply a probable loss where collateral is high quality, marked frequently and subject to margin calls.[1][3]

Gross contractual exposure is not expected cash loss

Accounting rules require disclosure of the maximum potential amount of future payments for certain guarantees without reducing that amount for collateral or recoveries. That produces an intentionally severe numerator. If every covered borrower defaulted at once and collateral recovered nothing, the gross indemnification figure approximates the protected amount. The scenario is not a forecast. BNY and State Street both describe controls including collateral requirements, daily valuation and credit review. Their disclosed collateral exceeded the protected exposure at the reporting date.[1][3]

The appropriate residual-risk question is therefore not “How can the banks owe $1.1 trillion?” but “What conditions could turn a collateralised promise into a loss?” The chain requires at least a borrower default, a replacement cost for the lent securities, and a collateral shortfall after liquidation or application of recoveries. Market gaps, wrong-way risk, illiquidity, operational delay, legal disputes or correlated defaults could widen that shortfall. The indemnity matters precisely because those events are unlikely but potentially consequential; it should neither be ignored as off-balance-sheet boilerplate nor reported as though the full amount were unsecured debt.[1][3]

Cash collateral creates another chain of exposure

The programmes also connect securities lending to the repurchase market. BNY said its indemnification total included $78 billion of cash collateral invested in indemnified repurchase agreements, up from $65 billion at year-end. State Street disclosed $53.603 billion of securities-lending cash collateral invested in indemnified repos, supported by $57.897 billion of repo collateral. These figures are not additional to every headline amount in the same way: they describe how part of the cash collateral is reinvested and protected within another transaction chain.[1][5]

That reinvestment matters because a cash-collateral programme contains two counterparty relationships. The securities borrower owes the lent securities back and posts cash. The agent or client then invests that cash, perhaps through a repo in which another counterparty supplies securities as collateral. Indemnities can attach to different links. Reporting the lending indemnity, the cash collateral and the repo collateral as three independent exposures would double or triple count the same economic chain. The safer approach is to map which amount protects which obligation.[1][3][5]

BNY’s CIBC Mellon disclosure must be kept separate

BNY also reported a CIBC Mellon securities-lending programme with $72 billion of indemnified exposure, down from $74 billion. BNY and Canadian Imperial Bank of Commerce severally indemnify clients through their joint venture. The amount is excluded from the $706.034 billion BNY programme total and from this article’s $1.133 trillion combined calculation. Adding it without explaining the joint-venture boundary would inflate the comparison and risk attributing the whole programme to BNY alone.[1]

The word “severally” is consequential. It does not mean both parents promise the whole amount jointly in every circumstance. The contractual allocation and recovery rights determine each party’s obligation. The public filing gives the programme scale and explains that it is separately indemnified, but it does not supply a borrower-level allocation or a stress loss for BNY’s share. The correct treatment is therefore disclosure, not consolidation into the bank’s own programme as though the exposures were identical.[1]

What the growth does—and does not—show

The combined nine-month increase of $94.660 billion is larger than the $69.680 billion point-in-time collateral excess, but comparing those two numbers as a coverage warning would mix a flow-like change with a stock cushion. The growth amount measures how much the protected principal increased between dates; the excess collateral is the margin over exposure on one date. A programme can expand while maintaining the same overcollateralisation policy. The relevant tests are whether collateral requirements, eligible assets, borrower quality, concentration and stress assumptions changed with the scale.[2][4]

Neither filing reports a material realised indemnification loss arising from these programmes in the cited disclosures. That absence is relevant evidence against presenting the gross amounts as an imminent liability. It does not prove that future loss is impossible. Securities-lending indemnities are designed for a failure scenario, and the economic value of the service depends partly on the agent’s ability to perform under stress. For investors and lending clients, the useful monitoring variables are the gross protected amount, collateral coverage, collateral composition, borrower concentration, reinvestment exposure and any recognised liability or loss.[1][3]

Questions the public disclosures cannot answer

The filings do not provide a common borrower-level denominator, a distribution of collateral haircuts, the overlap between the two banks’ borrower sets, or a comparable stress loss under one market shock. They also do not allow an outsider to infer that one bank is safer from a 0.17-point difference in aggregate coverage. A stronger cross-bank comparison would require aligned definitions of eligible transactions, netting, collateral valuation, indemnity exclusions and default scenarios.[1][3]

The documentary finding nevertheless stands. Two of the world’s largest custody banks supported more than $1.1 trillion of client securities financing with contractual indemnities at the end of September, while holding reported collateral worth about 106% of that amount. The figure reveals an important mechanism behind agency lending: custody and market infrastructure can create very large contingent promises without placing the full protected amount on the balance sheet as ordinary debt. Understanding the collateral chain is what separates the scale of the service from the size of a plausible loss.[1][2][3][4]

Why a common stress test would be more informative than the headline total

The next analytical step would be to apply one shock to both programmes: simultaneous default of the largest borrowers, a specified decline in collateral values, delayed liquidation and stressed foreign-exchange or settlement conditions. The public aggregates do not permit that calculation. They also do not reveal whether excess collateral is evenly distributed or concentrated in transactions already carrying low market risk. A common stress loss would be a better cross-bank denominator than gross indemnification alone.[1][3]

Client concentration matters on both sides of the arrangement. A custody bank may lend securities for many beneficial owners to a smaller group of major borrowers. Netting, collateral segregation and legal agreements can reduce or redistribute exposure, but the gross tables do not disclose the network. A trillion-dollar combined amount therefore signals infrastructure importance while leaving the actual default topology unresolved.[1][3]

Report an error or submit an update