Depending on whose estimate you prefer, somewhere between $5tn and $27tn sits idle in prefunded nostro and vostro accounts around the world. The larger figure is widely attributed to the Bank for International Settlements and has become the standard reference in recent industry analysis. The industry has spent a remarkable amount of energy arguing about which figure is right. It is the wrong argument. Whatever the true number, it measures the same thing: how much capital the global financial system has to immobilise to compensate for uncertainty it cannot otherwise resolve. The number is the price. The question worth asking is what the system is paying for.
Two narratives currently compete to answer that question, and both are wrong. The first says this is a treasury problem: with better forecasting, better visibility and better optimisation tools, institutions could run leaner. The second, louder narrative says the rails themselves are broken and should be replaced, usually by whoever is selling the replacement. The truth is less convenient for both camps. The system works really, really well. The mechanics of clearing and settlement are not broken. What is wrong is the incentives of the participants doing the clearing, and no dashboard or new rail fixes incentives.
Buffers exist where certainty is missing
Start with why the balances exist. Institutions prefund accounts across currencies, jurisdictions and correspondent relationships, carrying the cost of capital sitting idle in nostro accounts, a bank’s own money held at a foreign bank, and vostro accounts, the reverse arrangement. When settlement timing is uncertain, this is entirely rational behaviour. A treasury team that overfunds is protecting against a failed payout, a delayed settlement, a time-zone mismatch or a missed obligation. The cost of idle capital is real, but the cost of a client-facing disruption is worse. So, the buffers stay.
And the cost compounds. Recent market analysis puts the all-in cost of prefunding, including overhead, at 3% to 5% annually. At a 5% rate, every billion dollars parked in a settlement account represents roughly $50m a year in lost yield or financing. Call it what it is: defensive capital allocation. Institutions are not holding this liquidity because it is strategically useful. They are holding it because settlement is not predictable enough to do anything else. Once you name the behaviour correctly, the limits of the treasury-tools answer become obvious. A forecasting tool can predict funding needs. It cannot remove cut-off windows, take intermediaries out of a settlement route, or change the behaviour of a correspondent bank. You cannot optimise your way out of someone else’s uncertainty.
Speed is not the same as certainty
The replace-the-rails camp makes a different error. For years the conversation has centred on speed: faster rails, faster messaging, faster initiation. The industry has modernised the front end of payments impressively. Interfaces are cleaner, APIs are everywhere, status tracking has improved. None of it has released the trapped capital, because speed and certainty are different properties. An instruction can move in seconds while the underlying funds remain subject to liquidity constraints, intermediary timelines and local market dependencies. A message can arrive before the money is available. The friction lives before and after the message moves: funds have to be in the right place, in the right currency, at the right time, having cleared controls across multiple jurisdictions, in a form that can actually be credited.
This is also why the rail itself is not the answer. Whether a transaction arrives via stablecoin, tokenised deposit or Swift, the requirement is identical: direct access to central bank clearing and accounts that settle with certainty. A new rail that terminates in the same chain of intermediary balance sheets inherits the same uncertainty. The rail is a detail. The settlement guarantee is the point.
The problem is institutional design, not broken infrastructure
So, if the rails work and the tools cannot fix it, where does the uncertainty come from? Institutional design. Traditional correspondent banking sits inside organisations whose economics are shaped by lending, balance-sheet utilisation and net interest margin. Clearing and settlement matter to these institutions, but they are not their primary economic purpose. When clearing sits inside that model, settlement behaviour reflects that model. If an institution earns from deposits, lending and treasury spread, idle balances are not operational residue. They are economically useful. From the bank’s perspective that is rational. From the perspective of a platform, payment institution or corporate treasury trying to move money predictably across markets, it is a structural misalignment. You cannot optimise for yield and velocity at the same time.
The structural data make this sharper, not softer. Data published by the Committee on Payments and Market Infrastructures at the BIS show active correspondent banking relationships contracted by roughly a quarter between 2011 and 2020, even as message volumes rose. The network is concentrating: more of the world’s settlement flow is running across fewer balance sheets, and those balance sheets belong to institutions whose incentives were never designed around moving client funds. This is not a criticism of banks. Banks play a necessary role in global finance, and most of the controls in the system exist for good reason. But concentration without realignment means the uncertainty premium grows. That is what the trapped trillions are pricing.
What infrastructure designed for certainty looks like
Reducing trapped liquidity therefore requires participants designed for the job, not better workarounds. That means fewer intermediary balance sheets in the settlement path and more direct access to local clearing schemes. It means cash management built around client funds rather than around the economics of a lending book. And it means custody structures that make ownership unambiguous.
This last point is the part the existing model cannot easily replicate. A Named Account Custody approach gives each account holder a legal and operational relationship to the custody framework itself, rather than a position buried inside a nested chain. That structure reduces the opacity and chain risk that force institutions to overfund. When ownership is clear and the institution holding the funds has no competing claim on them, the defensive instinct to hold capital in more places than necessary starts to fall away.
The certainty benchmark
Here is a prediction. The next phase of correspondent banking will not be measured in settlement speed, because speed is largely solved at the messaging layer. It will be measured in certainty: whether an institution can know when money will arrive, where it is held, how it can be deployed and how much liquidity it genuinely needs to meet its obligations. Providers will be benchmarked on the size of the buffer their clients no longer need to hold. That is a number every treasurer can calculate for themselves, and it is a far more honest benchmark than transaction speed.
Whether the trapped figure is $5tn or $27tn, it is the market signalling the same thing: the system still leans on buffers because the participants at the centre of it were never built to provide certainty at the point of movement. Treasury teams have become very good at managing that problem. The opportunity now is to stop asking them to.
George Davis, CEO and co-founder of Lorum
