Who Owns The Liquidity Layer?

Tokenization is solving issuance and wallets. And recreating, one network at a time, the trapped liquidity problem that the Fed master account and a century of netting were built to escape.

Jun 14, 2026

JPMorgan’s Kinexys platform has moved more than $3 trillion since inception and now settles over $5 billion on an average day, with a stated target of doubling that to $10 billion. Citi runs Token Services for institutional clients. HSBC has extended tokenized deposits from Hong Kong and Singapore into the United States. Five US regional banks have stood up the Cari Network to mint deposit tokens on a shared ledger. The stablecoin market sits above $320 billion, with Tether and Circle holding more than nine tenths of it between them. Every one of these is a production system, not a pilot, and every one of them is solving the same half of the problem.

They are all building issuance. The token, the wallet, the mint, the redemption back to a dollar. That half is crowded now, well capitalized, and moving fast.

The other half is the part that makes a tokenized dollar usable once it exists. A token moves inside its own network in seconds. Getting value from one network to another, from a regional bank’s deposit token to a money center bank’s, from a stablecoin to a tokenized deposit, requires liquidity to be sitting inside each network at the moment the payment fires. That liquidity layer, the settlement plumbing between the rails, is the part almost nobody is building. And it is the part that decides whether tokenized money stays one fungible thing, or fractures into dozens of separate balances that cannot reach each other.

What a century of plumbing was built to avoid

Step back from the tokens and look at what the existing system actually does, because the existing system solved this problem a long time ago, and solved it well.

A US bank holds one balance at the Federal Reserve. That single master account balance works across every system the bank touches, so think about wires through Fedwire, batch payments through ACH, and instant payments through FedNow. The bank does not pre-position a separate pot of money inside each network. It funds one balance, and that balance is fungible across all of them. That is the entire point of a master account, and it is the thing that keeps a dollar a dollar no matter which rail it travels on.

On top of that, the system economizes. ACH settles on a deferred net basis, which collapses thousands of gross obligations into a single end of day figure and cuts how much money actually has to move. CHIPS, the large value network that clears most of the country’s dollar wire traffic, runs a liquidity algorithm that in 2024 settled $29 of value for every $1 of intraday funding, up from $26 the year before. By another measure it settles close to $62 for every dollar of excess funding in the system. The Clearing House put the annual savings from that netting at more than $5 billion. Instant payments pull the same trick a different way. RTP runs off a joint prefunded account at the Federal Reserve Bank of New York, owned collectively by its participants, with each bank’s running position backed dollar for dollar out of the shared pool.

Notice the pattern. Every one of those mechanisms exists to do one of two things, keep liquidity fungible across many systems, or economize the total amount of liquidity the system has to hold. One master account instead of many. Netting instead of gross. A shared pool instead of a funded balance in every silo.

We also know exactly what it looks like when that infrastructure is missing, because we live with one version of it every day. It is called correspondent banking. To make a cross border payment, a bank pre-funds a nostro account in the destination currency, sitting idle at a foreign correspondent, earning little, available for nothing else. The BIS estimates more than $27 trillion sits trapped in these accounts globally. A single large global bank can have $10 billion to $25 billion tied up across its correspondent relationships. That is the cost of a system with no shared settlement asset, where to transact inside a network you must first hold value inside it. Every dollar parked in a nostro is a dollar that cannot be lent or deployed. Liquidity that cannot move is not really liquidity at all.

Atomic and gross is the opposite of netted

So this is the bit that I think the issuance story skips. Tokenized settlement does not merely fail to net. It almost makes netting impossible by design.

Tokenized settlement is atomic and gross. Atomic means all or nothing, both legs of a transaction happen together or neither does. Gross means per transaction, settled in full, with no batching and no net position at the end of the day. Both properties are genuinely valuable. Atomic delivery versus payment removes the risk that you pay and the other side fails to deliver, which is real, and it is most of why institutions want this at all.

But run it at scale and the liquidity arithmetic inverts. A netted system lets a bank send $100, receive $95, and move only $5. An atomic gross system makes it move the full $100, at the instant the payment fires, or the payment fails. There is no settlement window. There is no intraday overdraft, no daylight credit, no soft landing while an outbound payment waits for an inbound one to fund it. If the money is not there at the moment of execution, the transaction does not settle late. It does not settle at all.

The official sector has been clear eyed about this even while the industry has stayed quiet. The Financial Stability Board, the IMF, and the BIS have all made the same point, that tokenized mechanisms settling on a gross basis require far higher prefunding than the netted systems they replace, and at scale that raises liquidity demand rather than lowering it. The IMF’s 2026 work on tokenized finance said it plainly, instant atomic settlement means transactions can only occur if cash and assets are immediately available, which eliminates the possibility of netting obligations. And where liquidity pools and settlement assets differ across platforms, fragmentation impairs par convertibility and degrades netting efficiency across the whole system.

So the structural picture is this. Liquidity demand goes up, precisely at the moment liquidity becomes most fragmented. You need more of it, in more places, with less ability to net it down. That is the exact inverse of what a master account and a netting engine were built to deliver, and it is being engineered back into the system one tokenized network at a time, mostly by people looking only at the issuance side of the ledger.

Who ends up being the clearinghouse

When there is no shared facility, each institution solves the liquidity problem alone, and the system as a whole solves it badly.

A bank that wants to transact across five networks pre-funds a balance in all five. That is five pots of idle liquidity where there used to be one master account, the nostro problem brought home and multiplied. For the largest banks, this is an irritation rather than a constraint. They have the balance sheet to hold inventory everywhere, and increasingly they do. JPMorgan’s deposit token is already live on Base, deployed on the Canton Network, and settling cross border tokenized Treasury redemptions on the XRP Ledger. A handful of institutions can afford to be everywhere at once.

For everyone else, the math is brutal, and I think it concentrates. Cross network settlement gravitates toward whoever already holds inventory in every network, which means the few largest institutions quietly become the de facto clearinghouses for everyone else, on terms they set themselves. A smaller bank either pays one of those institutions for access, or falls back on slow off chain correspondent workarounds, which reintroduces the precise friction tokenization was supposed to remove. The community banks that chose FedNow over the bank owned instant rails already understand how this story ends. They have lived the question of who owns the pipe and what they get charged to use it.

This is the barbell effect, expressed in liquidity rather than assets. The institutions that can fund everywhere consolidate the clearing function. The ones that cannot are left renting access or having to route around it. The middle hollows out, not because anyone designed it that way, but because that is where the absence of shared infrastructure naturally pushes the system.

The attempts, and what they leave unsolved

I’m clearly not the only one thinking about this, but most of what is being built around this problem feels like its solving something adjacent to it.

The most serious US attempt was the Regulated Settlement Network proof of concept, which in late 2024 brought together Citi, JPMorgan, Mastercard, Swift, TD Bank, US Bank, Wells Fargo, Visa, Zions, and others to test round the clock multi asset settlement of tokenized central bank money, commercial bank money, and Treasuries on a single shared ledger, with each institution running its own partition. The findings were encouraging and the legal workstream found no fundamental blockers. It was also a proof of concept, and it has not yet become production infrastructure.

Partior, the JPMorgan, DBS, and Temasek venture now backed by Standard Chartered and Deutsche Bank, runs live atomic cross border settlement today. But it is a closed, member owned network, which is to say it is one more rail that needs its own liquidity positioned inside it, not a layer that makes liquidity fungible across rails. Fnality, which grew out of the old Utility Settlement Coin idea, raised $136 million last year from Bank of America, Citi, and others to push into dollars and euros, and it took roughly a decade to get from concept to live payments, gated the entire way on Bank of England settlement finality. Ubyx, founded by Tony McLaughlin, the former Citi payments architect who designed the Regulated Liability Network, raised $10 million from Galaxy, Coinbase, and Paxos, took a first stablecoin investment from Barclays, and is building a real clearing system. But Ubyx clears redemption at par across issuers. It solves acceptance, getting a stablecoin back to a dollar in a bank account, which is necessary and valuable, and it is a different problem from funding the intraday gap when an outflow fires before an inflow lands.

Add them up and you get a great deal of motion around the settlement question and almost nothing aimed directly at the liquidity layer itself, a shared, jointly funded pool that any participant can draw on to cover a cross network shortfall, governed by a common rulebook for who can draw, against what collateral, and how a loss gets shared. That facility is exactly what CHIPS and the RTP joint account already are for the legacy rails. For tokenized networks, it does not yet exist.

Who provides it, and on whose terms

A shared facility is not free of risk, and pretending otherwise is how you build the next crisis. Mutualizing liquidity also mutualizes exposure. Pool reserves without a rulebook and you have built a channel for moral hazard and a single point of failure in the same structure. The hard questions are all governance questions, who gets to be a member, how votes are weighted, whether a community bank’s voice survives in a structure the largest members fund, and who actually operates the thing versus who merely convenes it and writes the standards.

We have run this experiment before, and the result is instructive. RTP and Zelle were built and owned by the large banks. Community banks largely declined to join, and waited instead for the Fed to build the neutral alternative, which became FedNow. When the institutions that need shared infrastructure do not trust the institutions that own it, they do not adopt it. They route around it, or they wait for a neutral party, and the fragmentation persists in the meantime. Any tokenized liquidity facility that gets the ownership question wrong will run into the same wall.

There is one development that could make much of this net down on its own, tokenized central bank reserves. If the top tier of the monetary system arrives on programmable rails with a round the clock intraday credit equivalent, a large part of the cross network funding need collapses back into the master account it came from, and a private pool matters far less. I think that is probably years away, and I would not build a strategy around it landing soon. Until it does, the function does not disappear simply because nobody has built the facility. It gets performed informally, expensively, by whoever already holds the inventory.

That is the real state of play. The issuance layer is the part everyone can see, so it is the part everyone is building, and it is turning into a land grab. The liquidity layer is the part that decides who clears whom, and right now it is empty. Empty infrastructure does not stay empty. It gets filled by whoever was already standing in the room, on whatever terms they care to set. The wallets are the announcement. The liquidity between the networks is the business.

Share Atomic Settlement

References

Tokenized issuance and platform volumes

Legacy settlement liquidity and netting

Trapped liquidity and correspondent banking

Atomic gross settlement and liquidity demand

Settlement-layer attempts

Atomic Settlement

Discussion about this post

Ready for more?