Wholesale settlement scales by getting out of bespoke, name-by-name credit and leaning on standardized, secured, or central-bank-backed money. Judged by that standard, a single-issuer stablecoin is a step backward: a private workaround that leads at the edge only because the more neutral public settlement option is locked by policy. The question was never deposits versus stablecoins. It is who gets to stand on the floor underneath both.
Most of the stablecoin argument happens at the wrong layer. People debate whether stablecoins beat cards or eat into bank deposits. The question that actually lasts is quieter: what asset does the network settle on. Ask it that way and the popular answer starts to look backward. The instrument everyone treats as the future is, by the logic of how settlement has always scaled, a step in the wrong direction. It leads the market anyway, and why it leads is the interesting part.
Networks scale away from bespoke credit
There is a pattern that runs through the whole history of market plumbing. The systems that scale stop relying on credit underwritten name by name and move toward exposure that is secured, standardized, mutualized, or settled in central bank money.
This is not because credit cannot move money. It obviously can, and most of the plumbing is still credit underneath. The narrower point is that bilateral credit, priced one relationship at a time, does not scale across a network. Each link is its own piece of work: a view on the counterparty, the legal terms, the liquidity, the capital you hold against it. Add one participant and you add that work in both directions with everyone already there, so the relationships grow with the square of the players.
The systems that beat this did not abolish credit. They made it less bespoke. Repo turned unsecured bilateral lending into secured, standardized exposure. Central counterparties took credit that used to sit between two firms and mutualized it behind a shared rulebook. CLS cut principal settlement risk by tying payment in one currency to payment in the other. Even card networks, which get misread here all the time, did not mutualize credit. They standardized the rulebook and the interbank settlement and left cardholder credit with the issuers who actually carry it.
The systems that scale don’t remove credit. They make it secured, standardized, mutualized, or settled in central bank money, instead of bespoke and bilateral.
So where does a stablecoin sit
Hold that pattern up against the instrument everyone is excited about, and the answer is awkward.
A single-issuer stablecoin is the opposite of mutualized. It is concentrated counterparty risk in one private company: its reserves, its banking relationships, its operations, its freeze switch. That puts it among the least mutualized things anyone seriously proposes as settlement money. By the same logic that explains why CCPs, repo, and CLS scaled, a single-issuer token is not a step forward. It is a regression.
Not a fraud, and not useless. A regression specifically against the way wholesale settlement has always scaled, which is exactly why a stablecoin is genuinely useful at the edge and structurally imperfect at the core. The cleaner wholesale endpoint is not any single issuer’s token. It is neutral settlement moving at message speed: tokenized central bank money, or a mutualized tokenized clearing layer that stands in the middle the way a CCP already does. A stablecoin is a privately built stand-in for that.
A fair reader will ask the obvious follow-up: if the stablecoin fails that test, does the tokenized deposit pass it? Not really, and not for a different reason. A tokenized deposit is faster bank credit, a claim on one institution, so it carries the same single-balance-sheet limit. Its edge is elsewhere. It keeps the dollar inside the regulated perimeter, which is precisely why most of traditional finance prefers it and why bank-led settlement is moving first there. So the deposit token wins inside the banking system; it is not the neutral layer underneath it. Two private claims, two different jobs, neither one the floor.
So here is the claim the rest of this piece defends. The stablecoin is a transitional workaround, not the destination. It leads not because it won the design argument but because the more neutral public option is shut. The honest test is whether that survives the hardest facts, and there are four of them.
The two instruments, read through the pattern
A tokenized deposit is bank-credit money in a faster wrapper. Its value still rides one bank’s balance sheet, with that bank’s credit, legal, and interoperability questions attached, so it is not mutualized either. It is a claim on one institution. That is the real gate on bank-to-bank interoperability, which still looks hard to scale, and the block is not mainly technology, it is credit. For two banks to settle in each other’s deposit tokens, each has to get comfortable with the other’s credit, legal, liquidity, and capital treatment, and every bank you add multiplies those bilateral relationships. It is the same scaling problem from before, now sitting inside the regulated perimeter. So the deposit token moves fastest where it never has to cross that gate: in-house and bank-led settlement, where keeping the dollar inside one regulated perimeter fits the existing capital and AML treatment. JPMorgan’s Kinexys is the live proof of that. A stablecoin sidesteps the gate by handing the market one more neutral asset to settle on, which is why it keeps landing as the bridge layer, even though it brings its own issuer, reserve, and redemption risks. Slower as neutral cross-network money, faster as in-house bank settlement.
A stablecoin is a reserve-backed issuer liability built to move like a payment. In most structures the holder does not hold the reserves directly. They hold a claim on the issuer, and the issuer holds the backing. What makes a well-built one feel like neutral settlement is not a security interest in the holder’s hands. It is that the backing is meant to be standardized, high quality, and ring-fenced. But that only goes so far. Better reserves make the single issuer safer; they do not make the exposure mutual. You are still standing on one private balance sheet.
Why the weaker design still leads: the door is locked
If the single-issuer token cuts against the pattern, why does it dominate? Because the better option is shut, by choice, not by physics.
Non-banks reach for a private token because they cannot reach central bank money directly. The floor exists; most of the economy just isn’t allowed to stand on it. And there are really two doors here, both closed on purpose. The wholesale door is master-account access to central bank settlement, which stays policy-gated, as Custodia’s denial and the long fight over TNB show. The retail door is a public digital dollar, which is blocked by executive action and a separate anti-CBDC bill, not by the stablecoin statute itself. Neither is an engineering delay. Both are decisions.
With the more neutral public option closed by decision, a concentrated private substitute wins by default. That tells you something about the door, not about the merits of the token.
The model and the market are not the same thing
The clean, ring-fenced model I just described is where regulation is pushing the category. It is not what leads the market today.
The biggest token by float is Tether, around three-fifths of the market on its own, with Tether and USDC together making up the large majority of stablecoin value. Tether’s published attestations show the bulk in Treasuries alongside allocations to secured loans, gold, and bitcoin. It operates offshore and is not authorized under the EU’s MiCA regime. The instrument carrying most of the volume, in other words, is not the standardized, fully ring-fenced asset the clean story assumes.
That does not weaken the argument. It sharpens it. If the leading token wins while holding gold and bitcoin against demand liabilities, and while direct redemption with the issuer is subject to verification and minimum thresholds that put it out of reach for most retail holders, who reach the peg through exchanges and on-chain liquidity instead, it is not winning on reserve quality or neutrality. It is winning on distribution, and on the fact that nobody has shipped a better-built public option. The market leader is the workaround in its least clean form, and it still leads. That is the locked door at work, not the design.
The law is coming, not here, and it reads like a patch
In the US, as of 2025, there is finally a legal standard aimed at the reserve model, and the tense matters, because a careful reader will catch it. The GENIUS Act was enacted in July 2025. It is not yet in force. It phases in on the earlier of January 2027 or 120 days after the final rules land. The agencies are still writing those rules through 2026, and non-compliant tokens stay sellable through a multi-year transition after that.
Once live, the Act sets a one-to-one reserve model in permitted assets such as cash, insured deposits, short Treasuries, certain repo, and qualifying money-market-fund shares. It requires segregation, limits rehypothecation, bans issuer-paid yield to holders, and carves reserve assets out of the issuer’s bankruptcy estate for the benefit of holders. The exact ranking on a shortfall is still being argued, so the honest line is to claim the carve-out, not perfection.
What is striking is what the carve-out is doing. It does the work a security interest would, without the holder ever touching the collateral. The law is making an unsecured claim on one private issuer behave more like a ring-fenced, secured-style payment claim. What it does not do is make the exposure mutual. Every holder is still standing on one issuer; the statute just makes that issuer safer to stand on. You only write a segregation-and-priority regime for an instrument whose core weakness is that holders are unsecured creditors of a single company. The law is the clearest admission that the workaround needs reinforcing.
Funding transformation, plus some real flight
A stablecoin can look like money walking out of the banking system. Part of that read is too clean, and part of it is true.
The reserves do not vanish. They reshape, landing across short Treasuries, repo, money market funds, and bank deposits. To the extent they recycle back as deposits at the issuer’s reserve banks, this is funding transformation, not exit. A deposit from a household or a company is granular and cheap to fund with. Pool a lot of them into one issuer and the balance that comes back to the reserve banks is larger, more concentrated, more rate-sensitive, and quicker to move. The liquidity rules treat it accordingly: a stablecoin issuer’s balance generally lands in the least favorable bucket, non-operational wholesale funding from a financial entity, with assumed outflows near the top of the scale, against single-digit assumptions for stable retail deposits. Same dollars, worse funding.
Some of this is funding transformation. Some of it is real deposit flight. Being honest about which is which is the whole point.
And the part the comforting version skips: where reserves sit in Treasuries instead of cycling back as deposits, it is not transformation at all. It is genuine funding loss for banks, concentrated on the smaller ones that leaned on those operating balances, plus a new structural bid for T-bills. One more concession a treasurer will force, so I will make it up front: the sticky-retail-deposit comfort is thinner than it sounds for the balances in scope here. Insurance caps at $250k, the large operating balances we are talking about are mostly uninsured, and uninsured balances are exactly what ran from SVB in March 2023. The runnable comparison is closer than the clean story admits.
Convertibility is the real bottleneck, and it is single by design
Among bank deposits, par is propped up by a public bundle you never see: central bank settlement, deposit insurance, supervision, a lender of last resort. That backstop is shared and public, mutual by design in a way one issuer’s reserves are not. It is not unconditional, either. Insurance is capped, support is discretionary, and SVB showed uninsured par can break without intervention. But the contrast holds: a deposit’s par leans on a mutual public net, a stablecoin’s par leans on one private issuer.
And that par is not symmetric. Institutions with direct issuer access can redeem at par. Most everyone else exits through secondary markets, where par depends on depth, spread, and confidence. Underneath sits an arbitrage that somebody has to fund on their own book, capacity that is cheapest when nobody needs it and thinnest when everybody wants out at once.
The bottleneck is not moving the token. It is convertibility. Can holders get back to bank money at par under stress, and through how many doors?
The SVB weekend in March 2023 was the proof, and it proved the thesis as much as the risk. USDC did not wobble because tokens stopped moving. It wobbled because part of the reserve was trapped in a failed bank and primary redemption was choked over a weekend, so the secondary market repriced convertibility. The token-transfer layer worked the entire time. The boundary back to bank money was where the single issuer’s concentration showed up, running through a handful of banking access points. A more mutualized settlement layer would not have the same single-issuer seam. The fragility was not a stablecoin accident. It was the predictable cost of standing on one private balance sheet instead of a shared one.
One tension in my own view
Stablecoins are most useful out at the edges of the system: offshore, in weak-currency economies, and for players who can’t get a bank account. Those edges happen to be dominated by the token with the least clean reserves and the hardest redemption access for ordinary holders, where direct redemption with the issuer is subject to verification and minimum thresholds that put it out of reach for most retail holders, who reach the peg through exchanges and on-chain liquidity instead. So the weak spot is widest exactly where the asset is most useful. That is a real problem, not a footnote.
Programmability cuts the same way. The admin controls that make 24/7 settlement possible are the same ones that allow freeze, blacklist, and seizure. That is a feature to a compliance team and a censorable settlement asset to a treasurer. It is a property of single-issuer control, not of tokenization as such.
Where it goes
The contest was never deposits versus stablecoins, and never which token has the slickest features. It is two old questions asked again with new plumbing. What do you settle on, bespoke credit or standardized, neutral backing? And who is allowed to touch the most neutral settlement asset of all?
The non-bank edge already leans toward standardized, reserve-backed money, and it built a private path to the floor because the public one is shut. Banks and regulators pull the other way, toward the tokenized deposit, because it modernizes settlement without leaving the regulated perimeter. If the public door opens, through tiered settlement access, a synthetic arrangement, or eventually tokenized central bank money, the private substitute loses its structural reason to exist at the wholesale core. Distribution, programmability, and offshore reach would still keep it alive at the edge. If the door stays shut, the workaround quietly hardens into the architecture. Which of these becomes the base layer is not settled, and pretending otherwise would be a forecast dressed as analysis.
Infrastructure has a way of becoming permanent long before anyone decides it should be.
The views here are my own and do not represent my employer or any client. This is general analysis, not investment, legal, tax, or accounting advice, and not a recommendation to buy or sell any asset.