Start with the constraint that governs everything else.
A payment stablecoin promises redemption at par, on demand, with no notice period. That promise is what makes it money rather than a fund share, and it is also precisely what makes it runnable. Any reform that genuinely removes the run risk removes the par-on-demand promise, and a token that cannot be redeemed at par on demand is not a digital dollar.
It is a short-term bond with a marketing problem.
So the honest framing of the policy question is not how to eliminate the run. It is where to put the loss-absorbing capacity that a run will eventually require, and who pays for holding it there.
The GENIUS Act answered a narrower question.
It made the assets safe: one-to-one backing, qualifying reserve assets, public redemption procedures, disclosure, reporting. That was the right first move, and it worked. What it did not do — what no reserve rule can do — is make the liability behave.
Federal Reserve Bank of New York researchers have already documented the behavior empirically, finding that stablecoin holders exhibit the same flight-to-safety dynamics as money-market fund investors and estimating a discrete break-the-buck threshold near $0.99 below which redemptions accelerate.1 Reserve quality does not appear in that mechanism. Neither does it appear in the Federal Reserve working paper published in June 2026 showing that runs can occur even when digital money is backed by perfectly safe reserves, driven instead by the interaction of network effects and rising transaction fees.23
Note
There are three places to intervene, and they are not equally attractive.
Move one: take the plumbing out of the loop
The most direct fix is to change what issuers hold. If reserves sit as balances at the Federal Reserve rather than in bills and repo, redemption stops being a market event. An issuer meeting redemptions would shrink a Federal Reserve liability instead of selling collateral into a stressed market, and the transmission chain running from redemption through repo withdrawal to margin calls on leveraged Treasury positions would simply not exist. No forced selling, no funding-spread widening, no feedback into the next day's redemptions.
The question of whether stablecoin issuers should have access to central bank facilities is already live. Researchers at MIT's Digital Currency Initiative have framed it directly: regulators will have to decide whether issuers get discount window access, and if not, what structural safeguards protect the Treasury market instead.4
Two objections carry real weight.
-
The first is competitive. A private company with a Federal Reserve master account, earning interest on reserves while paying nothing to token holders — the GENIUS Act prohibits issuers from paying interest directly — collects a spread that no bank subject to capital requirements and supervision can match. The float income is the entire business model, and moving reserves to the Fed converts that business model into a public subsidy unless the interest is clawed back or passed through.
-
The second is that this move guts the Treasury-demand argument that motivated much of the enthusiasm in the first place. If reserves sit at the Fed, stablecoin growth stops generating bill demand. The advisory committee projection of issuer Treasury holdings approaching $1 trillion by 20285 is a projection about where reserves are allowed to sit. Change the rule and the number goes to zero.
Important
The policy cannot simultaneously use stablecoins as a Treasury distribution channel and remove them from the Treasury market for safety. That trade-off is the whole design problem, stated plainly.
Move two: slow the run without freezing it
If reserves stay in market assets, the next lever is redemption mechanics. Money-market funds have already run this experiment, and the results are unambiguous about which tools work.
The 2014 reforms gave funds the ability to impose redemption gates tied to weekly liquid asset thresholds. In March 2020, that design produced exactly the wrong behavior:
- As funds approached the threshold
- Investors redeemed preemptively to get out before a gate could fall
- The possibility of being locked in accelerated the run it was meant to stop.
The 2023 amendments removed gates entirely and replaced them with mandatory liquidity fees for institutional prime funds, on the reasoning that a price on redemption during stress makes the redeeming investor bear the liquidation cost rather than pushing it onto whoever stays.10
Gates cause runs; fees priced to the cost of liquidation do not, because they change the payoff of running rather than the possibility of it.
A velocity-triggered liquidity fee — accruing to remaining holders, disclosed in advance, activating only above a redemption-rate threshold — is the closest thing to a run brake that preserves par-on-demand in normal conditions.
There is a second mechanic specific to tokenized money, and it is the gap the Federal Reserve paper identifies. The Act regulates issuers and reserves. It sets no capacity or price standard for the public blockchains that carry the tokens, and Treasury's proposed rulemaking in August 2026 kept that focus on issuers rather than network fees.3
When exit runs through a congested chain, the cost of leaving spikes exactly when the most people want to leave.
Tip
The remedy is unglamorous: a mandatory issuer-direct redemption channel that does not depend on chain conditions, with a committed settlement window, so that a fee spike on Ethereum cannot by itself convert a liquidity preference into a depeg.
Move three: reinforce the market the reserves sit in
The third intervention leaves stablecoins alone and hardens what they would be selling into.
A stablecoin redemption wave is dangerous less for its size than for its timing, because it withdraws short-term cash from a Treasury market financed overnight. Hedge fund long Treasury positions reached roughly $2.4 trillion by late 2025, financed by about $1.8 trillion of net repo borrowing, with nearly half of those positions in leveraged basis and swap-spread arbitrage.67
Any large, correlated withdrawal of repo cash lands on that structure. The tools here are already identified and partly built:
- Central clearing for Treasury cash and repo transactions
- Minimum haircuts on non-centrally-cleared repo
- Margin requirements calibrated so that a funding shock does not translate one-for-one into forced sales
- A standing facility that lends against Treasury collateral at a known rate when private repo widens
Important
The Standing Repo Facility exists for exactly this purpose. Extending eligibility to regulated stablecoin issuers against their own bill collateral would let an issuer raise cash without hitting the market — which is move one wearing a smaller hat, and worth noting as such.
Reserve composition drifts, and rules decide which way
One more design detail deserves attention, because it changes over time without anyone voting on it.
Reserve composition responds to incentives.
After Silicon Valley Bank's failure, one of the largest issuers cut the average maturity of its reserves and surged into repo concentrated at FICC, while shifting bank deposits toward the largest institutions — trading interest-rate risk for counterparty risk.8 That was a rational response to a specific shock, and it moved the systemic exposure from one channel to another rather than reducing it.
The Bank for International Settlements has mapped where each choice lands.
- If issuers hold bank deposits, retail funding is replaced by concentrated wholesale money.
- If they hold bills, banks lose high-quality liquid assets.
- If they could hold central bank reserves, conversion would strip banks of both deposits and reserves.9
There is no neutral reserve composition. There is only a choice about which part of the system absorbs the stress, made either deliberately by rule or accidentally by whichever asset happens to be yielding most this quarter.
The pattern nobody wants to say out loud
Line up the three moves and a pattern emerges that has repeated for a century and a half.
Private banknotes were runnable, so the National Banking Acts required them to be backed by government bonds. That was not enough, so 1913 created a lender of last resort. That was not enough, so 1933 created deposit insurance. Money-market funds broke the buck in 2008, and the Treasury guaranteed them within days. They seized again in 2020, and the Federal Reserve stood up a liquidity facility for them.
Important
Every one of those episodes ended the same way: a private liability that the public had come to treat as money acquired a public balance sheet behind it, generally in a weekend, generally under conditions that made the terms worse than they would have been if written in advance.
Each of the three moves above is a version of that same arrival. Reserves at the Fed put the central bank behind the token directly. Liquidity fees and a committed redemption channel are a private substitute for a backstop, which works until the day the fee is not enough. Standing repo access for issuers is a backstop with a haircut attached. The design space contains no option in which a multi-trillion-dollar par-on-demand instrument stands entirely on its own during a genuine panic, because no such instrument ever has.
Which leads to the conclusion that is politically awkward from every direction.
The stablecoin framework was built in part by people who explicitly opposed a central bank digital currency.
Follow the run-prevention logic to its end and the safest version of a regulated payment stablecoin is a Federal Reserve liability distributed through a private interface — public money, private wrapper, competitive front end. That is not what anyone legislated. It is what the mechanics select for.
The choice available now is not whether the public balance sheet ends up behind these instruments. It is whether that happens by design, with the terms written in daylight and the subsidy priced, or in a weekend, with a facility announced on a Sunday night because $1.8 trillion of repo has to roll on Monday morning.