Executive Summary

Four years after TerraUSD erased roughly €16 billion in a week and pushed the ECB from analysis to alarm^[1], stablecoin reserves are regulated in every major financial center: MiCA in force and enforced in the EU^[2][3], the GENIUS Act signed as US law with implementing rules in flight^[4][5], and licensing regimes live in Singapore, Hong Kong, and the UAE^[6][7][8], with the UK’s rules finalized for an October 2027 start^[9]. The convergence headline hides the finding that matters: the regimes agree on principles and disagree on every mechanic your treasury and diligence teams actually touch. Three findings drive this assessment.

For a board, the whole report compresses to one diligence question per stablecoin on the approved list: whose paper do we actually hold — issued by which entity, backed by what, redeemable how fast, and treated how under Basel’s Group 1b tests^[11]? An issuer that answers with a consolidated global number has answered a different question.

Regulatory Background

Every reserve rule now on the books traces to one week in May 2022. TerraUSD lost its peg on 9 May and traded below $0.10 within days; roughly €18 billion of market value compressed to under €2 billion, and the ECB moved from analysis to demand: MiCA, it wrote that July, “needs to be implemented urgently”^[1]. Terra was an algorithmic design with no reserve to speak of, but the legislative reaction did not distinguish: what emerged everywhere is a fiat-backed model with hard reserve mechanics, and the algorithmic category was regulated out of the institutional market.

The EU moved first — and enforced first

MiCA’s stablecoin titles (III for asset-referenced tokens, IV for e-money tokens) became applicable on 30 June 2024, a full year before any comparable regime^[2]. The enforcement followed quickly: ESMA’s January 2025 statement required crypto-asset service providers to restrict non-compliant ARTs and EMTs by the end of that month, with a sell-only wind-down through Q1 2025^[3]. That statement, more than the regulation itself, is what moved the market — the largest offshore stablecoin lost its EU venue listings not because a court ruled, but because every exchange’s MiCA licence depended on delisting it. Issuers who dismissed MiCA as a paper regime in 2024 spent 2025 restructuring.

The US arrived by statute, not agency action

After years of proposals that died in committee, the GENIUS Act was signed on 18 July 2025 as Public Law 119-27^[4]. It is narrower than MiCA — payment stablecoins only — and deliberately federalist: issuers at or below $10 billion may remain under “substantially similar” state regimes, while the OCC takes federal nonbank issuers. The statute’s requirements bite on the earlier of 18 January 2027 or 120 days after final implementing rules; as of mid-2026 the OCC’s proposed 12 CFR 15 and the Treasury, FDIC, and NCUA proposals are all out for comment, and none is final^[5][12]. US issuers are therefore in the awkward interval where the law is settled but the examination manual is not — a familiar position for anyone who lived through the early Basel implementations, and the period in which supervisory expectations are actually formed^[13].

Asia and the Gulf built licensing gates

Singapore finalized its single-currency stablecoin framework in August 2023, earliest of the majors, trading a narrow scope (SGD and G10 pegs, issued in Singapore) for depth of prescription^[6]. Hong Kong’s Stablecoins Ordinance took effect 1 August 2025 with HKMA licensing^[7] — and the first licensing round is the story: roughly three dozen applicants produced exactly two licences in April 2026, for Anchorpoint (the Standard Chartered HK/HKT/Animoca venture) and HSBC^[14]. That ratio is a policy statement. The UAE central bank’s Payment Token Services Regulation (effective 31 August 2024) is the strictest on reserve composition anywhere — dirham-token reserves sit as cash, in same-currency escrow, at a non-group UAE bank or the central bank itself^[8]. Japan runs stablecoins through its Payment Services Act as electronic payment instruments, restricting issuance to banks, funds-transfer providers, and trust banks^[15]. The UK closed the arc in June 2026: FCA policy statement PS26/10 finalized backing-asset rules ahead of an authorisation gateway opening September 2026 and a regime in force from October 2027^[9].

The prudential overlay nobody opted into

Above all of this sits the Basel Committee’s cryptoasset standard. The July 2024 amendments (d579) tightened the conditions for a stablecoin to reach Group 1b — the classification that spares bank holders punitive capital — including bankruptcy-remote reserve custody and a defined tolerance for short-term sovereign-collateralised reverse repos, with implementation from 1 January 2026^[11]. An issuer can be fully licensed locally and still fail Group 1b; for any bank-facing stablecoin, Basel is the binding constraint that no local licence waives.

The pattern, having mapped these regimes clause by clause: they converged on the two principles Terra made non-negotiable — full backing and redemption at par — and diverged on nearly every mechanic that determines what a reserve portfolio actually looks like. The next section takes the mechanics apart.

Reserve Requirements Comparison

Every regime says “fully backed, redeemable at par.” Your treasury desk cannot invest against a principle; it invests against an eligible-asset list, a deposit floor, and a redemption clock. Those three mechanics — not the shared slogan — decide the carry, the credit exposure, and the run profile of a stablecoin reserve, and they differ enough across the five live regimes that the same coin cannot hold the same portfolio in two of them.

1. Eligible assets: the deposit-floor divide

The deepest split is what the reserve may hold. MiCA anchors reserves in the banking system: an EMT issuer must place at least 30% of the funds received in separate accounts at credit institutions, with the remainder in secure, low-risk, highly liquid instruments under Article 38(1), currency-matched to the peg (Art 54); for significant EMTs the deposit floor rises to no less than 60% per official currency (Art 45(7)(b))^[2]. The EBA’s technical standards add maturity ladders (instruments mapped to one-to-five-working-day liquidity buckets) and bank concentration limits^[16].

The GENIUS Act points the other way — at the sovereign, not the banks. Section 4(a)(1) permits currency and Federal Reserve balances, insured demand deposits, Treasury bills of 93 days or less, overnight repos and reverse repos collateralised by such bills, government money market funds invested solely in the same, and tokenized versions of each^[4]. There is no deposit floor; a US payment stablecoin can lawfully run a reserve that is almost entirely short Treasuries and repo.

This is not a technical nuance — it is opposite risk philosophy. MiCA treats bank deposits as the safe core and drafted its floors as run-liquidity insurance; GENIUS treats the T-bill curve as the safe core and lets deposits be incidental. March 2023 already adjudicated the trade once: the largest reserve-related depeg to date came not from a bad asset but from reserve deposits trapped in Silicon Valley Bank over a weekend, resolved only when Treasury, the Fed, and the FDIC jointly guaranteed all depositors on the Sunday night^[17]. A 60% deposit floor writes that exposure into law. We think the US drafting has the better of this argument, and the EU knows it — the EBA’s concentration limits exist precisely because the deposit floor concentrates bank credit risk.

The outer positions belong to the Gulf and London. The CBUAE requires dirham-token reserves to sit as cash in same-currency escrow at a non-group UAE-licensed bank or at the central bank itself — no instruments at all^[8]. The FCA’s final rules put the backing pool on statutory trust with at least 5% in on-demand deposits, the lightest floor of any regime^[9]. Singapore allows cash, cash equivalents, and government or central-bank debt of the peg currency with residual maturity of three months or less — a 93-day-style list with an explicit currency match^[6]. Hong Kong requires the reserve pool’s market value to at least equal the par value of outstanding tokens, held on trust and segregated^[10].

2. Redemption windows: the run clock

A five-day window and an at-any-time obligation are different products under stress. The window is the time the issuer has to liquidate reserves in an orderly market; the obligation is the speed at which a run can drain them. MiCA’s fee-free, any-time redemption is the most holder-protective and the most run-exposed — it removes every friction that slows a panic, which is exactly why MiCA pairs it with the deposit floor. Regimes are internally coherent packages; the danger is comparing one element in isolation.

3. Attestation: from reporting to criminal liability

MiCA runs on supervisory reporting and audit within the EU framework; MAS requires monthly independent attestation and an annual audit^[18]. The GENIUS Act escalated the model: monthly public disclosure of reserve composition on the issuer’s website, examination by a registered public accounting firm, and a CEO/CFO certification carrying criminal exposure under 18 U.S.C. §1350 — the Sarbanes-Oxley enforcement mechanism pointed at reserve reports^[4]. Having spent years reading issuer “attestations” that were carefully-worded agreed-upon-procedures letters, we regard the certification requirement as the single most consequential clause in the Act: it converts reserve disclosure from a marketing exercise into a personal-liability event, and it will become the de facto global disclosure standard because no institutional counterparty will accept less from a competing issuer.

4. Issuer capital: small numbers, different shapes

Own-funds requirements are second-order but reveal each regulator’s model of failure. MiCA: 2% of the average reserve, 3% for significant EMTs (Arts 35(1)(b), 45(5))^[2]. MAS: the higher of S$1M or 50% of annual operating expenses^[6]. Hong Kong: HK$25M paid-up capital^[10]. CBUAE: AED 15M plus 2% of outstanding face value as an alternative^[8]. GENIUS leaves capital and liquidity minimums to the regulators, and the OCC’s proposed 12 CFR 15 covers exactly that ground^[5]. None of these numbers recapitalises a failed issuer; they are wind-down funding, and the percentage-of-reserve designs (EU, UAE) scale with the risk while the flat minimums (HK) merely gate entry.

The comparison, in one table

The operational consequence: build the reserve book per jurisdiction, not per coin. A portfolio that satisfies the strictest composition rule (UAE cash-escrow) fails no one on quality but bleeds carry everywhere else; a GENIUS-optimal book fails MiCA’s deposit floor outright. Multi-jurisdiction issuers are already running separately incorporated issuers with separately managed books — the subject of the next section, because separate books are exactly where the arbitrage lives.

Cross-Border Regulatory Arbitrage

Divergent rules for the same instrument create a market in jurisdictions. Some of that market is legitimate structuring; some of it is risk quietly relocating to wherever the rulebook is thinnest. Four vectors matter, and they are not equally dangerous.

1. The multi-issuer structure is now the default — and it fragments the claim

No single reserve portfolio satisfies MiCA’s 60% significant-EMT deposit floor^[2], GENIUS’s Treasury-and-repo list^[4], and the CBUAE’s cash-in-escrow rule^[8] at once. The industry answer is separate issuing entities per jurisdiction, each holding a locally compliant book under one brand. It works — and it quietly changes what a holder owns. The “same” token is a claim on a different legal entity with a different reserve, depending on where and from whom it was issued. In a failure, those books do not backstop each other; the EU holder’s claim on the EU reserve is unaffected by the health of the US book, which is the design working as intended — until holders discover mid-run that fungibility in the market does not mean fungibility in the claim. Diligence teams onboarding a stablecoin should demand the entity-level reserve attestation for the entity their tokens were issued by, not the consolidated marketing number.

2. Redemption-window asymmetry concentrates runs on the fastest door

The same brand redeems at any time in the EU^[2], within one business day in Hong Kong and the UAE^[10][8], and within five in Singapore^[6]. Under stress, sophisticated holders route redemptions to the fastest venue — the run happens where the window is shortest, against the local book, regardless of where the loss of confidence originated. The mitigation is unglamorous: per-entity redemption liquidity sized to that entity’s window, plus contractual limits on cross-entity token migration during stress. We have not yet seen a multi-jurisdiction stablecoin run; when it comes, the window map above is the playbook for predicting where it lands.

3. Regime-shopping gateways: state regimes and comparability

The GENIUS Act contains two deliberate openings. Issuers at or below $10 billion may remain under a state regime Treasury certifies as “substantially similar” (§4(c)(1)); and §18 lets foreign issuers serve the US market under Treasury comparability determinations^[4]. Both are sensible federalism and both are arbitrage surfaces: fifty state regimes reviewed against a standard that is itself still a proposed rule (Treasury’s state-regime proposal is April 2026, unadopted^[12]) is an invitation to charter where review is lightest. The gray zone is genuine — as of mid-2026 no comparability determination has been published, so nobody knows how strict the gate will be. Hong Kong ran the opposite experiment: roughly thirty-six applicants, two licences^[14]. A regulator that treats the licence as scarce forecloses the shopping vector at the cost of market breadth; we expect the HKMA’s ratio, not the state-charter route, to be what other supervisors copy after the first state-chartered failure.

4. The Basel overlay caps the arbitrage — for bank money

Whatever the local licence says, a bank holding or dealing a stablecoin answers to the Basel cryptoasset standard. The July 2024 amendments tightened Group 1b conditions — bankruptcy-remote reserve custody, defined reverse-repo tolerance — with implementation from January 2026^[11]. A stablecoin engineered to the thinnest local regime may be legal to issue and ruinous for a bank to touch: fail Group 1b and the exposure prices like an unbacked cryptoasset. This is the quiet convergence force. Institutional distribution runs through banks, banks run on capital, and capital runs on Basel — so the strictest-common-denominator book wins commercially even where the local rulebook permits less. ESMA’s 2025 delistings showed the same mechanism at venue level: distribution infrastructure, not the issuer’s own regulator, enforced the standard^[3].

Where we land

Rank the vectors: entity fragmentation (1) is a disclosure problem today and a loss-allocation problem in the first multi-entity failure; window asymmetry (2) is the acute stress risk; regime-shopping (3) is real but self-limiting wherever Basel-regulated distribution is the goal (4). For a compliance team, the practical test for any stablecoin on your approved list is a single question with four answers: which entity issued our tokens, what does its book hold, how fast must it redeem, and does it clear Group 1b? If the issuer’s answer starts with the consolidated global number, the diligence has not begun.