The New York Fed Just Published 43 Pages on Stablecoins. Here's what else needs to be said.
The Federal Reserve Bank of New York released Staff Report 1179 https://www.newyorkfed.org/research/staff_reports/sr1179 this month — "Stablecoins vs. Tokenized Deposits: The Narrow Banking Debate Revisited" by Huang and Keister.
It's a serious piece of general equilibrium modelling, and it frames a question that matters: Should blockchain-native money be issued as stablecoins backed by safe assets, or as tokenized deposits that fund bank lending?
Their headline finding is a neutral result.
Strip away regulation and moral hazard questions, and the composition of tokenized money doesn't matter.
Reintroduce those frictions, and the answer depends on the size of the regulatory burden on banks relative to their risk-shifting incentive. High regulatory costs favour stablecoins. Low costs favour tokenized deposits. In between, let them compete.
If that sounds like it should have been obvious before you built the model, that's because it is.
The paper essentially confirms that if you control for legal treatment, stablecoins and deposits are interchangeable as settlement media.
The interesting question from my perspective was always why you can't control for legal treatment, and on that question the paper is almost entirely silent.
As someone who spends time thinking about the minutiae of market infrastructure and talking to those who've actually run settlement infrastructure, treasury desks, and clearing houses see that this model misses.
"Cash-like" is not cash
The entire paper's welfare analysis revolves around interest rate differentials between stablecoins and deposits. But the binding constraint on institutional adoption has never been yield. It's balance sheet treatment.
When a bank treasurer holds central bank reserves, the prudential treatment is unambiguous: Level 1 HQLA with zero haircut, 0% required stable funding under NSFR, 0% risk weight, potential leverage ratio exemption.
When that same treasurer holds a GENIUS Act–compliant stablecoin, even one perfectly backed by short-dated Treasuries, a cascade of consequences follows. Under the Basel crypto-asset standard, qualifying stablecoins cannot be classified as HQLA at any level. (Currently we hope). That's a categorical exclusion, not a haircut adjustment. Every dollar moved from reserves to stablecoins reduces the HQLA stock with zero offset.
NSFR assigns 50–100% required stable funding factors to claims on private issuers. The leverage ratio includes the full exposure. Large exposure limits cap concentration to a single stablecoin issuer at 25% of Tier 1 capital (and the EU's CRR3 assigns a 250% risk weight to MiCA-compliant stablecoins).
None of this changes because the issuer holds Treasuries. The bank is holding a claim on a private entity, and the prudential framework treats it accordingly.
The paper models this entire dimension as a single parameter θ, a tax on bank size. That's a long way from the operational reality where the gap between "cash" and "cash-like" is measured across six simultaneous regulatory metrics.
Then the big wrong assumption when modelling stablecoin bank adoption, Nostro balances aren't idle: they're a revenue engine
There's a persistent assumption in the tokenisation literature that banks hold large correspondent balances out of inertia, and that tokenised instruments can unlock trapped liquidity. The fed paper inherits this framing implicitly through its treatment of safe asset intermediation as socially costly but otherwise neutral. A poor assumption.
In practice, cross-border payments represent roughly 20% of global transaction volumes but generate approximately 50% of transaction-related revenues. JPMorgan's Payments division alone produced nearly $19 billion annualised in Q4 2024. Nostro balances support FX spread capture, intraday liquidity provision, overnight sweep economics, and the cross-selling relationships that underpin an entire client franchise. When correspondent banking relationships are severed, corporate clients experience measurable declines in exports, revenues, and employment, effects that persist for years because rebuilding international trade links is slow.
Banks hold these balances because each dollar of Nostro generates material returns. A widely cited industry estimate suggests removing $300 million from a G-SIB Nostro costs approximately $50 million in annual revenue, a figure that becomes plausible when you decompose it across direct yield, FX spread revenue on supported flows, fee income, and attributed relationship value. The idea that blockchain eliminates "inefficient" Nostro holdings fundamentally misunderstands the incentive structure.
Central bank money is irreducible for final settlement
The paper treats settlement as implicit, buyers and sellers meet, exchange money for goods, and move on. There's no concept of finality, no distinction between intraday and end-of-day positions, no recognition that the identity of the settlement medium matters independently of what backs it.
Banks insist on central bank money for final settlement because it eliminates the need to underwrite the settlement asset itself. Central bank reserves carry zero credit risk, zero liquidity risk, cannot be rehypothecated, and provide legal finality that discharges obligations outright such as in the US under UCC Article 4A. A stablecoin transfer delivers a claim on a private issuer with redemption mechanics, operational dependencies, and a legal structure that behaves differently in insolvency. CLS settles approximately $6.5 trillion daily in FX using central bank money exclusively. That design choice is not aesthetic. It eliminates principal risk by construction.
The paper's implicit assumption that stablecoins can substitute for central bank money in institutional settlement is also increasingly empirically stale on the timing argument. FedNow already operates 24/7/365. TIPS provides continuous instant settlement in the eurozone. Fedwire is expanding toward 22 hours per day. The Bank of England targets near-24/7 CHAPS settlement by decade's end. The windows where stablecoins could claim a timing advantage are closing on published timelines. There is a question as to whether adoption was fast enough to make a difference, it think there is still time, but the gap is smaller than most stablecoin issuers would suggest.
The political economy the model assumes away
The paper's stablecoin issuers operate in a frictionless environment they intermediate safe assets into money at zero cost, entering and exiting freely. In reality, the political economy of central bank account access is the binding constraint, and it's adversarial.
TNB USA — a narrow bank founded by a 28-year Fed veteran and former EVP at the New York Fed — applied for a master account to hold only central bank reserves. The safest possible portfolio. The application that normally takes days dragged on for over six years before formal denial. If the Fed denied access to a narrow bank holding only reserves, the institutional logic clearly extends to stablecoin issuers or any other narrow banking. And if you are actually reading this not parsing with AI, I meant those emphasis hyphens. :) I did actually write this not generate it, just grammar improvement.
The GENIUS Act deliberately preserves this ambiguity: it neither grants nor bars Fed account access for stablecoin issuers, leaving the question entirely to the Fed's discretion. The ECB has explicitly prohibited safeguarding accounts for payment service providers. Norway's full-reserve bank surrendered its licence after the central bank withdrew access. Banking trade groups have warned Congress that stablecoins could drain up to $6.6 trillion in bank deposits and lobbied aggressively against any form of central bank access for non-bank issuers.
The paper's "efficient intermediation" channel may not clear in practice because the access it assumes is politically foreclosed. Modelling this as a parameter choice understates the structural nature of the barrier.
The market won't look like the model predicts
The paper assumes competitive stablecoin issuance with free entry and perfectly elastic supply at the safe asset return. Post-GENIUS Act, the market is consolidating into a regulated oligopoly. Compliance costs: BSA treatment, reserve governance, attestations, audits, operational resilience, all create substantial fixed cost barriers. USDT and USDC already command 82% of the $314 billion market. Tether's launch of USAT through Anchorage Digital Bank, with Cantor Fitzgerald as reserve custodian, targets the regulated US market directly. NBER research reveals that USDT has only six arbitrageurs in an average month, with the largest accounting for 66% of redemptions. The competitive dynamics the paper models won't materialise as described. Or even close.
What the paper reveals without intending to
The most useful thing about SR1179 is what it exposes by omission. The entire academic and policy establishment is focused on the settlement instrument question, who should issue tokenised money and what should back it. That's a legitimate question, but it's the wrong first question for institutional adoption.
The right first question is: what happens between the point of trade and the point of final settlement? Who intermediates credit risk? How does default management work when markets run continuously? What infrastructure exists for portfolio-level risk calculation across counterparties posting tokenised collateral? The FCM infrastructure that underpins traditional cleared derivatives cannot support tokenised collateral — legacy systems average 16 years old, no major back-office vendor has production-ready support, and no major CCP accepts tokenised collateral in production. You can only move collateral at the speed of the slowest participant in the chain. Ok, some of those really are the issues that I worry about related to stablecoins rather than everyone.
The Fed has written an elegant paper about money creation. The infrastructure gap that actually determines whether institutions can participate in tokenised markets remains unaddressed. Those are different problems, and conflating them leads to policy conclusions that fail on contact with the balance sheet.
Overall the issues here are that the payments businesses based on stablecoins are becoming very similar to the payments business that are traditional. It is a legal and regulatory difference primarily. I fully understand the value in these mechanics of zero knowledge proof, but its not adding a change to the returns or implementation complexity. Convergence will happen further here. Major differentiation based on legal difference will disappear.
The interesting place looking forward is the really close boundary, when a stablecoin looks more like a safekeeping receipt, without unnecessary minting for payments business cases. Where the reserve can be held in an approved entity rather than a narrow bank. One really interesting aspect is that the CFTCs Dec 15 publications explicitly gives equivalence to digital infrastructure. The implication is that a limited use Stablecoin that is issued in a DCO purely as an SKR against HQLA deposits, tokenised representations or not, should be treated on the balance sheet with zero haircut. It not the core value of a DCO, but it is super interesting for the scaling and development of tokenised asset capable DCO.
Thanks for reading if you got this far.
James Davies is Co-CEO of DCN. He has 25+ years in derivatives infrastructure.
This first appeared on LinkedIn on February 10, 2026. If you want to comment or discuss, that's the place.