Here's the number that should stop you cold: a bank's Liquidity Coverage Ratio can slide from 120% to 109% while its total deposits stay exactly the same. No wire. No withdrawal. No headline. The dollars never moved. The owner did. And that single reclassification is the whole story.
I've spent the last two weeks pulling apart the BIS, Fed, and Basel research on stablecoin reserves, cross-checking reserve attestations against the accounting they imply, and I keep landing on the same uncomfortable conclusion β the entire public debate is arguing about the wrong variable. Everyone is asking how much money is leaving the banks. The real risk lives one layer beneath that question, in the quality of what stays behind.
Let me show you the plumbing, because the plumbing is where this gets decided.
Stablecoins now hold hundreds of billions of dollars in reserve assets β mostly short-dated US Treasuries, plus a slice of bank deposits. That scale makes them a genuine participant in the money market, not a novelty. The Bank for International Settlements, the Federal Reserve, and the Basel Committee have all published research on how this growth interacts with bank funding. Three of the most conservative institutions on earth, studying the same question, roughly simultaneously. That is not academic curiosity. Institutions of this caliber don't spend research budget on hypotheticals. They do it because a rule is coming.
The framework that actually matters is the money hierarchy. Not all dollars are equal. A retail deposit held across a thousand households is sticky β it sits through volatility, it doesn't flinch at a headline, it forgets it exists. A wholesale deposit held by a single institution is flighty β it moves on a rumor, it rebalances on a basis point. Same currency. Radically different behavior under stress.
Stablecoins sit in the middle. They are programmable, they settle instantly, they cross borders without friction, and they carry no deposit insurance. They depend entirely on an issuer's credit and custody. So when a household moves $100 into a stablecoin, the bank does not lose $100. It loses a retail depositor and gains an institutional one. The ledger is unchanged. The stability profile is not.
The stakes aren't theoretical. If you hold a floating-rate loan, a business credit line, or a mortgage that reprices, you are already exposed to a variable you've probably never heard named.
I first internalized this distinction during the Terra collapse in 2022. I was exposed through a DeFi aggregator, and I spent 72 hours tracking oracle price feeds on-chain, watching the peg break in slow motion. What I learned there wasn't about algorithms. It was about who holds the money and how fast they can move it. The UST holders who moved first ate a haircut. The ones who moved last ate everything. Ownership and velocity decided the outcome, not the total value on the books.
Walk the actual money flow, because the accounting is where this argument is won or lost. There are three paths, and they are not equivalent.
Scenario A: the issuer parks the money in its own bank account. The household's $100 becomes the issuer's $100. Total deposits in the banking system: unchanged. Owner of record: changed from dispersed retail to concentrated institutional.
Scenario B: the issuer buys Treasuries on the secondary market. Its bank balance drops by $100; the seller's bank balance rises by $100. Aggregate deposits: still unchanged. Ownership: shifted again, this time toward whoever sold.
Scenario C: the issuer buys Treasuries directly off a bank's own book, or buys new issuance. This is where the model breaks from intuition. When an issuer buys a bank's own Treasury holdings, the bank's balance sheet contracts on both sides β the asset leaves, and the deposit liability is extinguished with it. When the issuer buys new issuance, the cash flows into the Treasury General Account, and only returns to the system later through government spending.
Scenario C is the contractionary path, and it is the one almost nobody models. If stablecoin growth routes mainly through this channel, the tightening effect on the banking system is materially larger than the headline suggests.
Now the regulatory mechanism that converts all of this into real, cash cost. Under Basel III, a bank must hold enough high-quality liquid assets to survive a 30-day stress scenario. The ratio is HQLA divided by projected net outflows over those 30 days. The critical input is the run-off assumption β the regulator's estimate of how much of each deposit type walks out under pressure.
Retail deposits carry a low run-off assumption. They are assumed to stay. Institutional and wholesale deposits carry a high one. They are assumed to leave.
So when a stablecoin issuer absorbs retail balances and redeposits them as institutional money, the denominator of the LCR calculation deteriorates. The bank hasn't lost a dollar, but its projected outflow just went up. To repair the ratio, it must either hold more liquid assets β which earn less β or switch to longer-term, more expensive funding. That is the mechanism behind the 120% to 109% move. No theft. No run. Just a quiet reclassification that forces the bank to pay more for the same balance sheet.
And here is the part that should genuinely anger people. That cost does not vanish. It travels. Higher funding costs push up loan pricing. The borrower who has never touched a stablecoin, who couldn't describe one, ends up paying for a deposit migration happening somewhere else entirely on the ledger. The cost is socialized onto people who never opted in.
Then there's the seigniorage question, and I've watched enough reserve attestations to know this is where the real money sits. The issuer holds interest-bearing Treasuries against the tokens it mints. That yield β the coupon on the bills, the interest on reserves β accrues to the issuer's shareholders. The token holder typically receives zero. So a structure that markets itself as a neutral payments rail is, economically, a private claim on quasi-sovereign money issuance. The float is the product. The transaction fee is a rounding error.
Two more data points worth pinning to the wall. The Fed's own work documents the same pattern β distributed household balances converting into concentrated institutional accounts, net total unchanged, quality degraded. And the banks are not sitting still. Several are piloting tokenized deposit products, effectively a stablecoin that never leaves the bank's own balance sheet. That tells you the institutions closest to the plumbing already see the funding-quality erosion coming.
I don't buy the framing that this is a technical flaw in stablecoin design. The reserve architecture works. It settles. It reconciles. What's under-modeled is the systemic spillover β the way a well-run product can still degrade the stability of the host system it plugs into. I learned that lesson the hard way during the 2020 DeFi freeze, when I rushed into Yearn vaults without reading the whitepaper, then watched withdrawals lock up in a gas war. Speed without a stability model is fatal. The same logic applies at the banking layer.
Here's where I part ways with almost everyone covering this.
The dominant question β "are stablecoins draining bank deposits?" β is the wrong question, and answering it "no" produces false comfort. Of course the aggregate is roughly stable in normal times. Money doesn't evaporate. But aggregate stability is precisely the camouflage. It hides a structural downgrade that only surfaces in stress.
Think about what happens in a genuine liquidity event. A retail base is slow. It takes days, sometimes weeks, of fear before households move. An institutional base holding stablecoin-linked balances is fast. It moves in hours, sometimes minutes, because the rail is programmable and the exit is one signature away. Stablecoins don't just change where deposits sit. They change how quickly deposits can run. That's bank run 2.0 β same fire, dramatically faster ignition.
The honest reading: stablecoins may reduce the probability of a slow drain while increasing the velocity of a fast one. That trade is invisible in calm markets and decisive in panicked ones.
The second blind spot is the counter-move nobody is pricing. Banks are already building tokenized deposits β programmable money that keeps the balance on the bank's own books. If that route wins, it doesn't just compete with stablecoins; it reshapes the entire competitive map, and the moat today's issuers enjoy β regulatory clarity plus reserve scale β gets thinner with every quarter.
And one calibration note I won't skip. The research I'm reading carries a timestamp that doesn't reconcile with the present β a date stamped into the future, plus a key quote from the BIS general manager that arrives truncated. I flag both deliberately. I've been burned by a mistranscribed date before, and I don't make the same mistake twice. If you're building a position on this narrative, pull the primary documents yourself.
Watch the run-off assumption, not the headline flow. If Basel formally widens the outflow haircut on stablecoin-linked deposits, banks will reprice custody overnight, and the cheap reserve model underpinning today's stablecoin economics starts to buckle.
The next twelve months won't be decided by how many billions leave the banks. They'll be decided by a spreadsheet cell most people will never see β the one that decides how much of a deposit is assumed to stay.
Which side of that cell are you actually holding?
