The real cost of pre-funding

Pre-funding is the cost a payment pays before it exists. To settle across the correspondent network, an institution keeps money parked in nostro accounts at partner banks in every corridor and currency it serves, sized not for the payments it expects on an average day but for the peak it might face on the worst one. That capital is the quietest line in the economics of cross-border payments: it appears on no invoice, gets compared in no pricing negotiation, and costs money every hour of every day.

Why the system demands it

Correspondent settlement does not move money across borders so much as adjust balances between banks that already hold money with each other. When a payment arrives for a beneficiary in another country, it is paid out of an account that was funded in advance. Three properties of the system decide how large that advance funding has to be.

First, settlement is slow and runs on banking hours. An account that can only be topped up during a business day, subject to cut-off times, must carry enough buffer to survive nights, weekends, and the days a replenishment payment itself spends in transit.

Second, flows are uncertain. The account must be sized for peak outflows, because a nostro that runs dry means failed payments and a damaged relationship with the correspondent. The difference between peak sizing and average usage is pure idle capital.

Third, every corridor multiplies the problem. Each currency, each correspondent, each market is its own pool, and the pools cannot help each other. A surplus in the Singapore dollar account cannot cover a shortfall in the Mexican peso account over a weekend. This is liquidity fragmentation in its most literal form: the same institution’s capital, divided into puddles that must each be deep enough alone.

What the idle capital costs

The cost has three layers, and only the first is ever estimated.

The direct layer is opportunity cost: capital sitting in low- or non-yielding accounts instead of being deployed. Whatever an institution’s marginal return on capital is, the nostro float earns almost none of it, and the difference is a running loss that scales with every corridor added.

The second layer is funding cost. Buffers must be financed. A bank funds its nostro positions on its balance sheet; a corporate funds pre-positioned balances out of working capital or credit lines. Either way, someone pays a spread to keep money idle, which is the least favorable trade in finance.

The third layer is the one Oliver Wyman and J.P. Morgan pointed at without measuring: their estimate that global corporates incur more than $120 billion a year in cross-border transaction costs explicitly excludes trapped liquidity and delayed settlement. The honest reading is that the industry’s best-known cost figure is a floor, with the pre-funding burden sitting somewhere above it, unmeasured because it is distributed across thousands of balance sheets as capital that could have been something else.

The BIS has meanwhile documented the system shrinking around its remaining participants: active correspondent relationships fell about 22% between 2011 and 2019 even as volumes grew. Fewer correspondents means longer chains and more concentrated nostro relationships, which pushes buffer requirements up, not down, for those still in the network.

What actually reduces it

Pre-funding is rational given slow, uncertain, hours-bound settlement. It shrinks only when those properties change.

Netting attacks the volume that needs to move at all. Where flows run in both directions, multilateral netting collapses gross obligations into small net positions, and the buffers required scale with the net, not the gross. This is the oldest liquidity technology in finance, and it is covered in the multilateral netting guide.

Faster settlement attacks the buffer window. An account that can be replenished in minutes, at any hour, no longer needs to carry a weekend’s worth of peak flow. This is the logic of just-in-time liquidity: funding positioned when a payment is known, rather than warehoused against payments that might come.

Atomic settlement attacks the failure window. Where the two legs of an exchange complete together or not at all, participants stop holding capital against the possibility of paying and not being paid, and settlement finality in seconds makes incoming funds usable immediately rather than provisionally.

None of these requires abandoning correspondent relationships wholesale. But all three are properties of the settlement layer, which is why the pre-funding cost cannot be negotiated away with any provider on the existing architecture. It is priced by how settlement works, and it falls only when settlement works differently, a comparison mapped across the alternatives to correspondent banking.

Where Frame fits

Frame is the settlement layer for global finance: one integration to orchestrate payments at scale across fiat rails, stablecoins, and tokenized deposits, with compliance enforced on every transaction and settlement in seconds instead of days.

Pre-funding is one of the costs that design exists to attack. Settlement in seconds, around the clock, means liquidity can be positioned when a payment is known instead of warehoused against a forecast, and a payment routed per transaction across fiat rails, stablecoins, and tokenized deposits settles over whichever rail lets capital arrive rather than wait. For enterprise treasuries, Frame works as an overlay on existing bank rails, where netting and pre-rail settlement reduce how much money has to move before any rail is touched at all, and no stablecoin is required. The buffers exist because settlement made them necessary. Change the settlement, and the capital comes back.