Private dollars, public plumbing.
A $280-320bn synthetic dollar now lives inside the US Treasury-bill market — a private analogue of a central bank's reserves, where the issuer accumulates T-bills to back its token the way a sovereign accumulates them as FX reserves. The GENIUS Act mechanically turns its growth into a structural bid for bills; for hundreds of millions in emerging markets it has quietly become the savings account the local currency never was; and the IMF's 2026 Global Financial Stability Report finds it has overtaken unbacked crypto as the primary vehicle for cross-border crypto activity. Read the supply level as a slow structural variable and the weekly mint-and-burn as the fast one.
Dollar-pegged stablecoins have crossed from crypto-trading instrument to durable monetary plumbing. Two demand channels now matter: private dollarization (emerging-market households holding synthetic dollars outside the correspondent-banking system, overwhelmingly USDT on Tron) and a captive, front-end bid for US Treasury bills — because GENIUS-compliant reserves must sit in government paper maturing in 93 days or less. It is still less than three cents of every dollar of bills outstanding; but it is a non-trivial marginal buyer of the new ones — a small ocean next to the money funds, but a large river flowing into T-bills.
Two numbers describe the regime, and they move at different speeds — a slow-compounding stock and a live, week-to-week flow.
Two issuers run the market: Tether's USDT (offshore, El Salvador) and Circle's USDC (onshore, NYSE-listed) together hold above 82% of supply. They diverge most on disclosure — and that gap is the single most important thing to understand about reserve risk.
| Metric | USDT · Tether | USDC · Circle | Note |
|---|---|---|---|
| Supply (mid-Jun 2026) | ~$186.5bn | ~$74.8bn | USDT ~61% share, USDC ~24% (Tether & Circle transparency pages, 11-13 June 2026). A duopoly above 82% of supply; the remaining ~15-17% is a fragmented tail. |
| Issuer / domicile | Tether · El Salvador (offshore) | Circle · US (NYSE: CRCL) | Offshore vs onshore. Circle is MiCA-authorised and on a GENIUS-compliant path; Tether is not a GENIUS permitted issuer and faces the tighter foreign-issuer regime. |
| Reserve disclosure | Quarterly BDO attestation | Monthly Deloitte exam + daily SEC N-MFP | Neither is a full financial-statement audit. Circle names its cash banks and files CUSIP-level holdings; Tether discloses category totals only. |
| Direct T-bills | $117.0bn | $19.1bn | Tether Q1 2026 (BDO, 31 Mar) / Circle Apr 2026 (Deloitte, 30 Apr) reports. |
| Treasury repo | ~$24bn | ~$47bn | Overnight + term, Treasury-collateralised. Adding repo lifts Tether to ~$141bn and Circle to ~$66bn of total Treasury exposure. |
| Dominant chain | Tron (retail / P2P) | Ethereum (DeFi / institutional) | Tron carries the majority of USDT by transaction count; Solana is the fastest-growing rail for USDC. See the chain note below. |
| Use-case centre | EM savings & cross-border P2P | US enterprise, DeFi pricing, regulated B2B | The two franchises barely compete for the same user: offshore retail cash versus onshore institutional rail. |
The chain split is the cleanest proxy for who uses a stablecoin and why. Tron carries the majority of USDT by transaction count — the day-to-day digital cash of the Global South, where low fees matter more than composability. Ethereum hosts roughly $160bn of stablecoin value (DeFiLlama chain table) and is the institutional and DeFi settlement layer. Solana is the fastest-growing rail for USDC, capturing low-cost retail and high-frequency flow. Same dollar token, three different economies.
Regulated US payments token; PayPal distribution is its distinctive moat.
Offshore exchange-liquidity token; concentrated on large centralized-exchange trading pairs.
Consortium model (Paxos-issued) designed to share reserve yield with network partners.
Payments and settlement positioning, leaning on Ripple's institutional corridors.
The largest decentralised, over-collateralised model; USDS is the rebranded successor to DAI.
None individually exceeds a few billion dollars, and together they split the ~15-17% the duopoly leaves. The distinction that matters is not size but lane: the regulated subset (PYUSD via Paxos, USDG, RLUSD) is where institutional Treasury allocators look first for compliant exposure, while the decentralised names (USDS/DAI) answer a different demand. One caution for the T-bill thesis: synthetic dollars such as Ethena's USDe sit outside this list entirely — they are crypto-collateralised, not Treasury-backed, and are not a bill buyer at all.
The attestation reconciles to a clean balance-sheet identity: total assets of $191.768bn against $183.536bn of liabilities, leaving an $8.232bn excess-reserve buffer (Tether Q1 2026 ISAE 3000R assurance, BDO Italia, 31 Mar).
- Gold$19.838bn
- Bitcoin$6.624bn
- Listed equities$3.408bn
- Secured loans$15.830bn
- Other$4.843bn
On the sector's largest issuer, that distinction is the whole story. An attestation, as accounting commentators put it, confirms only that the numbers reported on a given date match the books; an audit asks whether the system producing those numbers is reliable enough to trust over time.
| Reserve leg | Amount | Detail |
|---|---|---|
| Direct US Treasuries | $19.111bn | Held inside the Circle Reserve Fund; disclosed at CUSIP level, filed daily on SEC Form N-MFP. |
| Treasury repo (overnight) | $46.998bn | Treasury-collateralised repo inside the Circle Reserve Fund. |
| Cash in the Reserve Fund | $1.003bn | Cash leg of the SEC Rule 2a-7 government money fund. |
| Cash at regulated banks | $10.567bn | Held at named, regulated institutions including BNY Mellon. |
Total reserves of $77.124bn back $77.048bn of USDC in circulation (Circle April 2026 examination, Deloitte, 30 Apr) — a $0.076bn cushion.
The bulk of the reserve sits in the Circle Reserve Fund — an SEC Rule 2a-7 government money-market fund managed by BlackRock and custodied at BNY Mellon.
- Reserves back tokens at least 1:1, confined to cash, Federal Reserve balances, insured deposits, Treasuries maturing in 93 days or less, overnight Treasury repo and government money funds holding only those — crypto and the stablecoins themselves are excluded.
- Holders may be paid no yield; disclosure is monthly, with a registered-firm examination and CEO/CFO certification.
- Issuers above $50bn add annual audited financial statements; any issuer at or above $25bn holds at least 0.5% of reserves (capped at $500m) in insured deposits.
- A state-qualified issuer up to $10bn may stay on a substantially similar state regime; above $10bn it moves to federal OCC oversight within 360 days or stops issuing.
- Liquidity: at least 10% of reserves liquid daily, 30% weekly; a weighted-average maturity no longer than 20 days — tighter than the 93-day statutory ceiling.
- Concentration: no more than 40% of reserves at any single eligible institution.
- Redemption: T+2 business days, stretching to T+7 if redemptions hit 10% of supply within 24 hours — a circuit-breaker for a run.
- Plus a 12-month operating-expense liquidity backstop, weekly confidential reporting, and monthly public composition.
- Circle's EURC holds above 50% of the euro-stablecoin market.
- MiCA-compliant euro tokens command roughly 91% of euro-stablecoin share.
- Euro stablecoins have grown about 1,200% year on year.
- Article 50 bars interest on e-money tokens, walling EU stablecoins off from yield the same way GENIUS does in the US.
Two of these dates are mechanisms, not just milestones — one converts stablecoin growth into Treasury-bill demand, the other could force a reallocation of existing float.
| Where | Signal | What it means |
|---|---|---|
| Argentina | 71% | of crypto purchases are stablecoins versus a ~40% regional average — a daily hedge against peso depreciation, and the clearest case of the IMF's point that adoption tracks the classic dollarization drivers (FX volatility, inflation, weak institutions). |
| Turkey | 4.3% of GDP | the world's highest stablecoin purchases relative to the size of the economy — lira weakness and capital controls turning a synthetic dollar into a household savings instrument. |
| Sub-Saharan Africa | 43% | of crypto transaction volume is in stablecoins; Nigeria alone runs an estimated ~$22bn a year in stablecoin remittances, the region's low-cost alternative to correspondent-bank transfer. |
| Brazil | ~$82bn | of stablecoin flow routed through a single large exchange (Bitso) — Latin America's busiest regulated corridor, and a marker that the demand is settlement, not speculation. |
| Asia (settlement hub) | ~60% of flow | of the roughly $390bn annual stablecoin payment volume originates in Asia, with Singapore and Hong Kong the dominant settlement hubs — so most of the rail's traffic is not US even though the liability is dollar (Chainalysis-aggregated, Tier-2). |
The IMF links stablecoin adoption to the very drivers of classical dollarization — currency volatility, inflation and weak institutions — and finds the same telling asymmetry that haunts the textbook version: dollarization is sticky. Once households move savings into a harder currency they rarely move back, so demand persists even after local inflation cools — a structural floor under the float, not a cyclical spike.
Payments are the fastest-growing use of stablecoins, and increasingly a business one — evidence the rail is a real flow, not a marketing slogan.
In calm conditions the peg is held not by faith but by arbitrage: a token trading at $0.99 is a free dollar to anyone who can buy it and redeem it with the issuer at par, so authorised redeemers close the gap almost as fast as it opens. March 2023 is the canonical illustration of what happens when that pipe is blocked.
The transmission to the Treasury market is not symmetric, and the asymmetry is where the tail risk lives.
Solvency is about the assets. A run is about the plumbing. The March 2023 episode broke the peg even though the reserves were money-good — four conditions stacked at once.
| Condition | Why it breaks the peg |
|---|---|
| A frozen rail | Par redemption with the issuer is the mechanism that arbitrages the peg back to $1.00. When the banking rail that processes redemptions is shut, that mechanism is offline and only the secondary market sets the price. |
| A banking-hours gap | Stablecoins trade 24/7; the banks behind their reserves do not. A failure that lands on a Friday leaves a weekend with no way to move reserve cash — the gap SVB fell into. |
| Concentration | The $3.3bn stranded at SVB was about 8% of the reserve in a single institution. Diversifying the cash leg is now the first-order defence, and the reason Circle moved its cash into a government money fund. |
| A confidence shock | Once holders cannot see and reach the cash, some will sell a $1.00 token for less than $1.00 rather than wait, even when final recovery is near-certain. The secondary-market discount is the price of the doubt, not of the assets. |
| Stablecoin supply | Bill-equivalent demand | Share of bill stock | Share of new issuance |
|---|---|---|---|
| $230bn | $184bn | 2.7% of the bill stock | baseline (start-2026) |
| $400bn | $320bn | 4.7% of the bill stock | 30.9% of CY26 net bill increase |
| $750bn | $600bn | 8.8% of the bill stock | 94.5% of CY26 net bill increase |
The arithmetic is deliberately transparent: assume full 1:1 reserves and an ~80% allocation to bill-like instruments — direct bills, Treasury-backed repo and government money funds that themselves hold bills — so the bill-equivalent bid is simply supply times 80%, and the incremental bid is (supply minus the $230bn baseline) times 80%.
Duration risk is modest by design.
Hand-off to the Rates desk: read the on-chain front-end bid as one consolidated line, not a stablecoin-only one.
Stablecoin demand is not uniformly rate-sensitive. The cleanest split is roughly two-thirds transactional balances — payment and trading collateral, held for use, with low sensitivity to the level of rates — and roughly one-third idle store-of-value balances, where a non-yielding token's opportunity cost bites. The GENIUS Act bars issuers from paying holders any yield, so when policy rates are high that idle sleeve has somewhere to go: tokenized money funds and Treasury-backed yield tokens. The migration is regime-dependent. On a $400bn market, at a ~5% policy rate about 30% of the idle sleeve migrates; at ~2% only about 10% does. Today's effective fed funds rate of 3.62%, with a 3.75% target ceiling, sits between the two.
| Rate regime | Float split | $400bn market splits into | Issuer income |
|---|---|---|---|
| ~5% policy rate | ~65% transactional / ~35% idle; ~30% of the idle sleeve migrates | ~$358bn payment + ~$42bn yield-token | ~$16bn issuer income (supply x ~80% x rate) |
| ~2% policy rate | ~65% transactional / ~35% idle; only ~10% migrates | ~$386bn payment + ~$14bn yield-token | ~$6.4bn issuer income |
The non-obvious conclusion is the one that matters for the front end: payment-stablecoin float is rate-sensitive, but total on-chain short-Treasury demand is less so, because the substitute the idle balances flee to is itself a bill buyer. High rates shrink the no-yield token and lift issuer carry; low rates do the reverse. Either way the consolidated bill bid largely stays put. These are scenario figures on transparent assumptions, not forecasts — read them as market structure, order of magnitude.
The cleanest demand amplifier, and the one most aligned with the rest of the cross-asset book. A strong-dollar, EM-stress regime pushes the slow structural variable — the supply level — structurally higher.
Genuinely two-sided, which is why the net effect is mixed rather than directional.
A marginal near-term threat, not a structural one. A US retail central-bank digital currency is politically disfavoured, which leaves regulated private stablecoins as the de-facto digital dollar by default rather than by design.
Cuts two ways at once, and the second cut is the non-obvious one.
The second on-chain bid for T-bills, and the regulated answer to the GENIUS no-yield rule. BlackRock's BUIDL (about $2.2-2.5bn) holds 100% cash, Treasury bills and Treasury repo, pays a roughly 3.4% seven-day yield and carries an Aaa-mf rating; Franklin Templeton's BENJI (about $1.98bn at 29 April 2026) is the first US-registered tokenized money-market fund. Unlike a payment stablecoin these are securities that pay their holders — but their reserves buy the same front-end paper. The right way to size the channel is a consolidated line: payment stablecoins plus tokenized money funds plus Treasury-backed yield tokens, roughly $200-230bn of on-chain bill demand today and plausibly $300-400bn by end-2026. That consolidated figure, not the payment float alone, is what the Rates desk should carry.
The most common error in stablecoin commentary is to lump every dollar token into "stablecoin Treasury demand." It does not all belong there, and the distinction is mechanical. Ethena's USDe is a synthetic dollar held to its peg by delta-neutral perpetual-futures basis trades — collateralised by crypto and a hedge, not by Treasuries, so it buys no bills at all (BaFin forced an EU wind-down). Ondo's USDY, by contrast, is a Treasury-backed regulated security yielding about 3.55% on roughly $2.1bn of assets — it is a bill buyer. The working taxonomy: payment stablecoins, tokenized money funds and Treasury-backed yield tokens such as USDY all transmit to the bill market; synthetic yield tokens such as USDe do not. Counting USDe in the aggregate overstates the channel.
Covered in full in the rate-elasticity table above. The one point worth carrying away: because the yield products that idle balances migrate into are themselves bill buyers, the consolidated on-chain Treasury bid is far more rate-resilient than the payment float alone would suggest.
When new stablecoins are minted and their reserves buy bills, is that net-new demand for Treasuries, or money reshuffled out of existing money-fund and bill holdings? The honest answer is both, in proportions no free data set can yet separate. Emerging-market and retail dollarization flows are predominantly net-new — a saver leaving local currency was not previously a bill holder. US institutional flows are partly reallocation — a treasurer rotating from a money fund into a tokenized equivalent adds no new buyer. A clean split needs holder-residency data that has no free public feed, and BNP Paribas and the TBAC have both cautioned the net effect is smaller than the headline. We flag the ambiguity rather than resolve it, because resolving it on the available evidence would be false precision.
The Fed's Senior Financial Officer Survey (September 2025) found roughly half of large banks prioritising tokenized-deposit issuance and about 40% prioritising reserve services for stablecoin issuers — banks are restructuring around stablecoins, not simply losing to them. Where the reserve cash ultimately sits decides how disruptive the shift is:
- Reserves stay as bank deposits — aggregate deposits little changed, only more concentrated.
- Reserves move into bills and repo — pulled out of banks, but recycled to the market through dealers.
- Issuers win remunerated access to central-bank balances — the largest disintermediation risk, draining deposits out of the banking system, and the open question for bank funding models.
The figures that dominate headlines belong in a clearly fenced box, because every one is a projection rather than a current fact. Standard Chartered has put potential new T-bill demand at $0.8-1.0tn by 2028; Citi has modelled roughly $1tn of net-new bill purchases by 2030; and a Bessent/TBAC supply trajectory has been read as pointing toward about $2tn of stablecoins outstanding by 2028. These are scenario estimates, not the live state — today's supply is the $280-320bn band — and they collide with the net-versus-gross question above: BNP Paribas and the TBAC caution the genuinely additive bid is smaller once reallocation is netted out. Useful as an upper-bound sketch of where a structural front-end buyer could go; not a forecast we endorse.
Three near-dated events will move the channel:
- MiCA cliff, 1 July 2026 — post-cliff USDT EU-venue data is the first read on forced reallocation.
- OCC final rule, due by 18 July 2026 — it fixes how tightly compliant reserves are pinned to the front of the curve.
- Tether's Q2 attestation (around July) and the pending KPMG audit — the test of whether the disclosure gap finally closes.
For the high-frequency read, watch the weekly mint-and-burn for the turn and the cross-tracker supply spread for data quality.
- Intra-quarter reserve composition — both issuers disclose only at a reporting date (quarterly for Tether).
- Intraday peg by venue — aggregate prices mask venue-level stress, and the order-book depth that decides a run is not free.
- Tether security-level holdings — category totals only, never the CUSIP detail Circle files.
- Holder residency and end-user attribution — the proprietary chain-analytics that would split net-new EM demand from US reallocation (published classifiers self-report 65-79% accuracy).
- Issuer repo counterparties, intraday redemption queues, exchange-internal netting, and foreign-issuer US-customer segmentation (awaiting Treasury comparability determinations).
HIGH — Anything traceable to a primary, free source: the GENIUS Act text (GovInfo) and the OCC proposed rule (occ.gov); the issuer attestations (Tether/BDO, Circle/Deloitte) and Circle's daily SEC N-MFP filings via EDGAR; Treasury TBAC and TIC Table 5; the BIS transmission work (Working Paper 1270, Bulletin 108); the IMF working paper on stablecoins and cross-border flows; the Fed FEDS Note on the SVB de-peg; and the live macro series — FRED for the policy rate, ICI for money-fund assets.
MODERATE — Figures that are credible but second-hand or single-model.
- B2B payment volume and geography: The business-to-business payment-volume and geography numbers (~$390bn annual, ~60% Asia, ~65% year-on-year growth) are Chainalysis-aggregated and attributed as such, distinct from the IMF's audited $2.019tn of mapped 2024 volume.
- The MiCA EU-liquidity-at-risk figure: The ~$184bn estimate is a single directional model estimate.
- The scenario table's share-of-issuance math: Our arithmetic on hard inputs, not a published series.
Quarantined — Any single-tracker point-in-time supply figure — use the $280-320bn band; the ~$30-40bn cross-tracker spread is a methodology difference (circulating versus total versus in-custody), not an error. We do not publish a single "stablecoins as the Nth-largest Treasury holder" ordinal: the TIC comparison reads only as rivalling the smaller foreign-holder lines, because Tether is a private issuer, its repo is not identical to direct holdings, and TIC country lines cover all maturities, not just bills.
Sources: [1] GENIUS Act (PL 119-27), full text — GovInfo[2] OCC stablecoin proposed rule (March 2026)[3] Tether transparency & Q1 2026 BDO attestation[4] Circle transparency & USDC reserve report (Deloitte)[5] SEC EDGAR — Circle Reserve Fund N-MFP filings[6] BIS Working Paper 1270 (stablecoins & T-bill yields)[7] BIS Bulletin 108 (flow asymmetry)[8] IMF Working Paper 25/141 (stablecoins & cross-border flows)[9] Federal Reserve FEDS Note (Dec 2025) — USDC/SVB de-peg[10] Treasury TBAC quarterly refunding presentations[11] Treasury TIC — foreign holders of US Treasuries[12] ICI money-market fund assets (weekly)[13] DeFiLlama stablecoins tracker & API[14] FRED — effective fed funds (DFF) & target ceiling (DFEDTARU)[15] ESMA — MiCA regulation & registers
Related desks: Rates (the front-end T-bill bid) · Dollar, FX & Gold (private dollarization) · Crypto (plumbing & de-peg mechanics).