Ten days of maturity, and a pile of gold

This desk has written about stablecoin settlement repeatedly this month — a card book settled in an issuer's own coin, a bank consortium moving tokenised deposits, a clearing house adopting the rails. Each is an argument about payments. None opens the balance sheet. The sector's float is now roughly $300bn, of which USDT is about $183bn and USDC about $74bn — some 88 per cent between two issuers — and the interesting question is no longer whether the money is there. It is what it is sitting in, and for how long.

Circle is the easier of the two to inspect, because it files. Its Form 10-Q for the quarter ended 30 June reports $74.05bn of segregated reserve assets, of which $73.16bn backs stablecoins held by customers. Of that, $61.9bn sat in a single vehicle: the Circle Reserve Fund, a registered government money market fund managed by BlackRock. The remainder — roughly $12bn — is cash at banks. Reserve income for the quarter was $667.7m, against $410.4m of distribution and transaction costs, which is a reminder that the economics of a stablecoin are the economics of a deposit franchise whose distribution is rented.

The fund publishes a factsheet, the most informative document in the sector. As of 31 August the Circle Reserve Fund held $62,317m. Its weighted average maturity was ten days. The portfolio was 52.0 per cent direct US Treasury debt and 48.0 per cent overnight Treasury repurchase agreements, and 83.3 per cent of it matured within one to seven days. The seven-day SEC yield was 3.62 per cent.

Ten days is an extraordinary number. A government money market fund is permitted 60 days of weighted average maturity; a conventional bill ladder would run 30 to 45. Circle is running at a sixth of the regulatory allowance and giving up yield to do it. That is not conservatism for its own sake but a recognition that the liability is redeemable on demand, around the clock, and that the only way to guarantee dollar-for-dollar redemption during a fast outflow is to hold assets that mature into cash before the outflow finishes. Nearly half the fund is overnight paper. The reserve is not, in any meaningful duration sense, invested in Treasuries. It is parked in them.

Tether is a different business wearing the same label. Its attestation at 30 June, prepared by BDO, reports total assets of $187.75bn against liabilities of $183.64bn, of which $183.62bn is USDT in circulation. Excess reserves — the equity cushion — were $4.11bn, and the company's own release confirms gold holdings above 146 tonnes after adding 14 tonnes in the quarter, a $2.38bn or 15 per cent reduction in secured lending, and net operating profit of $1.50bn for three months' work.

The line-item composition is not in that release; Tether publishes totals and leaves the breakdown to the reserves report itself. As relayed by accounts of that report, Treasury bills account for about $114.9bn or 61.2 per cent of assets, with direct and indirect Treasury exposure including repo reaching roughly $140.6bn; precious metals stand at about $18.8bn or 10 per cent, bitcoin at about $5.8bn, and secured loans at about $13.45bn. This desk could not open the reserves report and flags those figures accordingly; the totals above are the company's own.

Put the two structures side by side and the difference is not a matter of degree. Circle holds overnight government paper and bank deposits. Tether holds a large government portfolio plus roughly $38bn of gold, bitcoin and secured loans against a $4.11bn equity buffer. That buffer is 2.2 per cent of liabilities. It is a fraction of the assets it is meant to absorb losses on: an 11 per cent drawdown in the non-government book would exhaust it. Tether's excess reserves were reported at $8.23bn a quarter earlier and halved as gold and bitcoin were marked down — which is the mechanism working exactly as described, in a quarter that nobody would call stressed.

American law has already reached a conclusion. The GENIUS Act's Section 4 lists what a permitted payment stablecoin issuer may hold: currency, insured bank deposits, Treasury securities with 93 days or less remaining, repurchase agreements of seven days or less, overnight-collateralised reverse repo, and government money market funds restricted to the same. Reserve assets may not be pledged, rehypothecated or reused, directly or through a custodian. Gold is not on the list. Bitcoin is not on the list. Secured loans are not on the list.

That is why Tether launched a second token. USA₮, issued in January through Anchorage Digital Bank with Cantor Fitzgerald as reserve custodian, is a separate instrument designed to sit inside the federal perimeter. Global USDT remained offshore and unchanged. The largest dollar stablecoin in the world is not a GENIUS Act stablecoin, and its issuer has said so by building another one rather than reshaping the first. The rulemaking on what an examination must establish matters a great deal more once you notice that the statute's reserve list and the market leader's reserve sheet are different documents.

Now scale it. Treasury's own Monthly Statement of the Public Debt shows $7,248bn of marketable bills outstanding at 31 August against $40.2tn of total public debt. Tether's roughly $140.6bn of direct and indirect exposure plus the Circle Reserve Fund's $62.3bn come to about 2.8 per cent of the bill market: enough to matter at the margin, not enough to set the price.

The best measurement of that margin is the Bank for International Settlements'. Working paper 1270, by Rashad Ahmed and Iñaki Aldasoro, uses daily data from January 2021 to March 2026 and a granular instrument built from idiosyncratic variation in token-level growth. It finds that "a $3.5 billion (2-standard deviation) inflow lowers 3-month Treasury bill yields by 0.71 basis points on impact, and up to 4 basis points within 10 days, with limited spillovers to longer tenors". The effects, the authors add, "are state-dependent, increasing under Treasury market stress and with the scale of the stablecoin sector".

Both halves of that sentence deserve attention. The first says stablecoins are a front-end phenomenon: they buy bills, they do not buy bonds, and anyone arguing that stablecoin growth will fund the long end is arguing against the evidence. The second says the coefficient is not stable. It grows with the sector and it grows in stress — which is precisely when a reserve manager is a forced seller rather than a buyer, and precisely when the bid for bills is thinnest.

Projections of where the sector goes should be read as projections. Standard Chartered's widely quoted estimate is $2tn of stablecoins by the end of 2028, implying up to $1tn of incremental bill demand. That is a house view, not a fact, and it assumes a growth rate the sector has not sustained this year. But even a fraction of it changes the Treasury's issuance calculus, because a structural buyer of bills that never extends duration is an argument for issuing more bills — for shortening the average maturity of the national debt in exchange for a cheaper front end. Regulators have begun saying out loud that this plumbing is being rebuilt.

The thing to watch is not the peg. It is the roll. Circle's ten-day weighted average maturity means the entire reserve turns over roughly three times a month, which in a $62bn fund is a standing order in the bill and repo markets of a size that shows up. That is a design choice made to survive a redemption wave, and it is the right one. It also means the sector's contribution to the front end is not a stock of holdings quietly sitting still. It is a flow that has to be re-established every ten days, by a buyer whose own funding can leave overnight.