For thirty-six hours in March 2023, the second-largest dollar stablecoin traded at eighty-seven cents. The reflex is to file the episode under "depeg risk" and move on. That framing is too generous to the asset and too lazy about the mechanism. What the market actually witnessed was a reserve-portfolio mark-to-market event leaking into a payment rail: a bank failure inside the reserve stack, a redemption queue, and a token price that briefly stopped being a price and became a probability. The lesson was never that stablecoins can break. The lesson was that a stablecoin is not a rail at all. It is a claim โ a short-duration Treasury position wrapped in a token, carrying every ounce of duration and custody risk that implies.
Begin with the macro record, because that is where the signal actually lives. A dollar stablecoin is, mechanically, a levered bet on the front end of the yield curve. The issuer holds reserves โ Treasury bills, overnight repo, and a shrinking slice of bank deposits โ and issues a token that pays nothing. The spread between what the reserve earns and what the token pays is the entire business model. At a five percent policy rate, a hundred-billion-dollar float throws off roughly five billion in annual gross reserve income with almost no marginal cost. At zero percent, the same float generates nothing, and the issuer must monetize transaction fees or rehypothecate reserves โ the latter being a regulatory red line that effectively kills the model. This is the hidden transmission channel that most crypto commentary still misses: stablecoin issuance is not a function of crypto sentiment; it is a function of the front end of the curve. When the Fed expands its balance sheet and pins the short rate low, the float expands because the opportunity cost of holding a zero-yield token collapses. When policy tightens and the curve inverts, the float does not necessarily shrink, but its composition and its profitability both get re-priced.
Context matters here, because the architecture is deceptively simple and the simplicity is precisely what hides the risk. The reserve stack has three layers. The first is direct bill holdings, which carry clean duration but require a custodian. The second is repo and money-market funds, which add counterparty exposure in exchange for liquidity. The third is commercial bank deposits, which look like cash on a dashboard and behave like credit risk in a run. The income statement only sees the blended yield. The balance sheet, however, is a stack of concentrated custody relationships, and the March 2023 depeg was fundamentally a custody-and-credit event disguised as a peg event. I have spent enough time inside CBDC and reserve-management working groups to know that the number nobody models is not the peg; it is the correlation between reserve custodians. Diversification on a term sheet is not diversification in a stress scenario.
Now to the core technical question, which is the one the market keeps avoiding: can a payment token pay yield? The answer is a structural no, and the reason is legal rather than technological. A token that pays a variable return on a pooled reserve portfolio is, functionally, a money-market fund. The moment it pays yield, it triggers securities law, disclosure obligations, and a fiduciary wrapper that no issuer wants. So the design is forced: yield stays at zero, and all reserve income accrues to the issuer. That single constraint explains almost everything about the sector. The business is a spread business, not a network business. And it is why "stablecoin as payment" and "stablecoin as savings" are diverging into two different products that regulation will eventually separate by statute. The savings version is already being absorbed into tokenized Treasury products; the payment version is being pushed toward a licensed, low-margin settlement layer.
Here is the observation that reframes the whole debate. On-chain velocity of stablecoins is low. The overwhelming majority of supply sits idle or is used as collateral, not as payment. If you measure the float against actual settlement volume, the "payment revolution" narrative is a minority use case wearing a majority costume. The float is collateral first. And collateral demand is a derivative of leverage demand, which is a derivative of policy. This is the liquidity tether in its purest form: the stablecoin float is a shadow monetary aggregate, and it moves with central-bank posture, not with the narrative cycle. When the float grows, on-chain leverage grows mechanically; when it contracts, deleveraging is not a choice but an accounting identity.
Every serious protocol critique needs a stress test, so let us run one honestly. Assume a rapid easing cycle: two hundred basis points of cuts inside twelve months. Reserve income on a hundred-billion float falls by roughly two billion annually. The issuer has three responses. First, cut operating costs โ finite. Second, monetize transaction fees โ viable only if the payment use case is real, which we have just shown is a minority of volume. Third, extend reserve duration to chase yield โ which reintroduces interest-rate risk into an instrument whose entire promise is price stability. The breakeven is brutal: below a threshold rate, the spread model inverts and the issuer becomes dependent on float growth to survive. Based on my audit work on reserve-backed instruments, this is the failure mode that never appears in a whitepaper but always appears in a rate-cut cycle. Yields dissolve; infrastructure remains โ and what remains is a settlement layer that must justify itself on throughput, not on spread.
There is a deeper trilemma here that the industry refuses to name. Price stability, yield, and unconditional redeemability cannot all be maximized at once. Maximize yield and you import duration risk. Maximize redeemability and you must hold liquid, low-yielding reserves. Maximize stability and you accept that the token is a utility, not an investment. Code enforces what contracts cannot: the peg is a rule, but the reserve composition is a judgment, and judgment is exactly where the fragility hides.
This is where the contrarian angle bites. The prevailing narrative says crypto is decoupling from traditional finance โ that on-chain markets are becoming an autonomous system. The evidence says the opposite. Stablecoins are not the decoupler; they are the coupler. The more crypto-native a market becomes, the more it depends on dollar policy, because every unit of on-chain liquidity is ultimately a claim on a dollar reserve managed by a handful of custodians. The sovereignty story is backwards. The correct read is that crypto has built the most efficient transmission belt for US monetary policy ever devised โ a private, programmable extension of the dollar system that responds to the front end of the curve faster than any bank deposit.
And that leads to the regulatory endpoint that the maximalists keep misreading. The state does not compete; it absorbs. CBDCs will not kill stablecoins. They will regulate them into a licensed dollar-extension layer, exactly as Europe's framework is already doing. The absorption is not hostile; it is thermodynamic. A payment token that settles in dollars and holds Treasury reserves is already inside the monetary perimeter. Formal regulation merely writes down what was always true. From speculative frenzy to institutional ledger, the arc is not a betrayal of crypto โ it is the maturation of a claim into an instrument.
The unglamorous risk that deserves more attention than the peg is reserve concentration. The peg is a price; concentration is a structure. A float backed by a handful of custodians, each holding bills through a small number of clearing relationships, is not diversified no matter how many line items appear on the disclosure. In a genuine run, the binding constraint is not the token contract; it is the settlement speed of the underlying reserve. This is the gap between a narrow bank and a real bank: stablecoins replicate the liability side of banking without the deposit insurance or the lender of last resort that make the liability side survivable. The architecture is honest about what it holds and silent about who it depends on.
So where does this leave positioning? The next cycle will not be decided by whether the float reaches some round number. It will be decided by whether the reserve stack survives a genuine rate shock with its duration intact. Watch two variables that almost nobody quotes on a screen: the float-to-M2 ratio, which tells you how much shadow money is circulating outside the banking system, and the average duration of the T-bill ladder, which tells you how much rate risk the peg is quietly carrying. When the float expands and duration stays short, the system is healthy. When the float expands and duration stretches to defend the spread, the peg is borrowing against its own stability.
The question worth sitting with is not whether stablecoins become systemically important. They already are. The question is whether the market can distinguish a payment rail from a reserve portfolio before the next thirty-six hours force the distinction for it.