Key takeaways
- Tether's BDO attestation at 30 June 2026 shows $187.75 billion of assets against $183.64 billion of liabilities, a buffer of $4.11 billion or 2.24% of what it owes.
- Of Tether's assets, 61.2% are US Treasury bills and 74.9% are cash equivalents. The other 25.1% is gold, bitcoin, secured loans, equities and other investments.
- Circle's 10-Q shows $73.16 billion segregated for USDC holders against $72.93 billion owed, a 0.32% cushion, with 84.6% of it in one government money market fund.
- USDC traded at 86 cents in March 2023 when 8% of its reserves were stuck at Silicon Valley Bank. TerraUSD fell from $0.9964 to $0.7934 in a day in May 2022 and its market cap was down $18.47 billion by month end.
- A 5% stablecoin sleeve marked at USDC's 2023 low costs a portfolio 0.70%. A sleeve in a coin that goes to zero, as the New York Fed says Terra collapsed, costs its full weight: 5%.
What backs a stablecoin, and has the peg ever broken?
You've watched a stablecoin sit at $1.00 for months and want to know what's behind that number. The short answer is that it depends on which coin, and the 2 largest are backed in very different ways. It also depends on what "backed" means when part of the reserve is in a bank that failed on Friday.
The peg has broken, more than once. The 2 episodes that matter are TerraUSD in May 2022, which suffered what the New York Fed calls a collapse, and USDC in March 2023, which fell to 86 cents and recovered once US authorities stepped in. Both are worth reading closely. The stablecoin reserves question is really a question about the worst day, not the average one.
This is an educational piece, and it has to be. Cryptoassets have been restricted mass-market investments in the UK since the FCA's financial promotion regime came into force on 8 October 2023, with a 24-hour cooling-off period among its required frictions. The FCA's own wording is that consumers "should not expect to be compensated under the Financial Services Compensation Scheme if things go wrong." Nothing below is a recommendation to hold one.
Tether holds $187.75 billion, and a quarter of it isn't cash-like
Tether publishes a quarterly attestation prepared by BDO. The one for 30 June 2026 lists total assets of $187,751,426,411 against total liabilities of $183,641,897,215. The difference, $4,109,529,196, is what Tether calls its excess reserves. Expressed against liabilities, that's a buffer of 2.24%.
The composition matters more than the total, and the chart above plots it. US Treasury bills, with a weighted average maturity under 90 days, come to $114.96 billion, or 61.2% of assets. Overnight reverse repurchase agreements add $18.63 billion and term reverse repos another $6.99 billion. Together with a rounding error of non-US bills, that cash-equivalent subtotal is $140.64 billion, or 74.9%.
The remaining 25.1%, $47.11 billion, is where the reserve question lives. Physical gold, held at fair value, is $18.84 billion, and Tether's July 2026 release puts the pile at more than 146 tons. Bitcoin, marked at a Bloomberg close of $58,642.15, is $5.80 billion. Secured loans, described as over-collateralised and subject to margin call, are $13.45 billion. Public equities are $3.76 billion and "other investments" $5.24 billion. Cash actually sitting in bank deposits is $40.3 million, which is 0.02% of the pool.
Put the buffer next to the volatile assets and the arithmetic gets uncomfortable. Gold and bitcoin together are $24.64 billion. A 16.7% fall in their combined value would use up the entire $4.11 billion excess. Bitcoin alone would need to fall 70.8% to do it, and gold and bitcoin falling together is the case the buffer is really sized against. Tether's own release notes it cut secured lending by $2.38 billion, or 15%, in the quarter. It also reports a net operating profit of about $1.5 billion, which is the mechanism by which the buffer gets rebuilt when it isn't paid out.
Circle holds almost nothing but Treasuries, and almost no buffer
Circle, the issuer of USDC, files with the SEC. Its 10-Q for the quarter to 30 June 2026 shows $73,161,172 thousand of cash and cash equivalents segregated for the benefit of stablecoin holders, against $72,927,544 thousand of deposits from those holders. The difference is $233.6 million, a cushion of 0.32%. USDC in circulation at quarter end was $73.3 billion.
Where Tether spreads the pool across 7 categories, Circle concentrates it. The same filing discloses $61.9 billion held in the Circle Reserve Fund, a BlackRock government money market fund whose holdings are subject to Rule 2a-7, down from $66.3 billion at 31 December 2025. That's 84.6% of the segregated pool, with the other 15.4% outside the fund. BlackRock's own page for the fund, ticker USDXX, showed net assets of $62,317.4 million on 3 September 2026, a weighted average maturity of 9 days and weekly liquid assets of 100%.
So the 2 largest dollar stablecoins answer the reserve question in opposite ways. Tether runs a 2.24% buffer over a pool that's 25% gold, bitcoin, loans and equities. Circle runs a 0.32% buffer over a pool that's almost entirely 9-day Treasury paper. One is a bank-like balance sheet with a capital cushion. The other is closer to a money market fund with a thin wrapper. Neither is a bank deposit, and neither carries deposit insurance.
The economics explain the design. Circle earned $668 million of reserve income in the quarter at a reserve return rate of 3.5%, and that income was 95% of its $701 million total revenue. Tether's $1.5 billion quarterly profit is, in its own words, "led by U.S. Treasury and repo." The issuer keeps the interest. The GENIUS Act, signed on 18 July 2025 as Public Law 119-27, makes that explicit: no permitted issuer may pay a holder "any form of interest or yield" solely for holding the coin. You're lending the issuer dollars at 0%. The reserve is what you get instead of a rate.
USDC went to 86 cents because 8% of its cash was in one bank
The March 2023 episode is the best-documented run on a fully reserved stablecoin, and the Federal Reserve published a detailed account of it in December 2025. When Silicon Valley Bank failed, Circle announced it couldn't access $3.3 billion of USDC reserves held there, around 8% of a total that was then roughly $40 billion. The other 92% was fine. It didn't matter. With primary redemptions paused over the weekend, USDC traded on secondary markets at 86 cents to the dollar at its trough.
The run then moved through the plumbing. The Fed's authors note that around 1 billion USDC was deposited into the DAI stablecoin's peg stability module on each of 2 days, and that over 400 million USDP, more than half its supply, was pulled out of the same module. By 15 March Circle had redeemed $3.8 billion and minted $0.8 billion. Its cash at US financial institutions fell from $11.5 billion on 6 March to $3.7 billion on 31 March, even though the price "recovered sharply after the backstop announcement on Sunday."
The lesson the Fed draws is the one that matters for the reserve question. The SVB episode "demonstrates the potential for at least a proportion of such assets to become inaccessible in a stress scenario." A reserve is only as good as the holder's ability to get at it on the day it's needed. That's why the regulatory fights below are almost entirely about where the cash sits, not just what it's invested in.
A separate New York Fed staff report, by many of the same authors, finds a discrete "break-the-buck" threshold of $1 below which redemptions accelerate, the same dynamic the authors document for money market funds. A coin at 99.9 cents is a pricing quirk. A coin at 99 cents is a run in progress.
TerraUSD fell 20% in a day because there was no reserve at all
The counter-case to "reserves are the answer" is that the worst stablecoin failure had none to argue about. TerraUSD was an algorithmic coin, held at $1 by a mechanism that minted its sister token Luna to absorb redemptions. The New York Fed's July 2023 account records the price dropping from $0.9964 to $0.7934 between 7 and 8 May 2022 as that mechanism broke. Each dollar redeemed increased Luna's supply, which pushed Luna's price down, which needed more Luna for the next dollar. Luna's supply went from 365 million units on 9 May to more than 6 trillion by 13 May.
The crash wiped out $17.17 billion of Terra's market value and $20.77 billion of Luna's, and Terra's market capitalisation was down $18.47 billion by the end of the month. Between 1 and 16 May 2022, stablecoin circulation as a whole fell by $15.58 billion. What's instructive is where the money went. The New York Fed notes that USDC and BUSD, "backed by traditional safe financial assets," took significant inflows in the same fortnight. The run was a flight from unreserved coins to reserved ones. That's the strongest evidence in the record that reserve composition is what holders are actually pricing.
The GENIUS Act now requires reserves "on an at least 1 to 1 basis" for any US permitted issuer, from a list that runs to cash, insured deposits, Treasury bills of 93 days or less, overnight repo and reverse repo backed by those bills, and government money market funds. A Terra-style design isn't eligible. That's the regulatory response to May 2022, arriving 3 years later.
UK rules on stablecoin reserves: 5% in bank deposits, up to 30% at the Bank
Notice that the GENIUS list is about what the reserve is invested in. The UK's 2 regulators went further and specified where it has to sit, which is the SVB lesson written into rules.
The FCA's final rules for qualifying stablecoins, PS26/10 of 30 June 2026, set an on-demand deposit requirement of 5% of the backing pool that has to be held in bank deposits available on demand. On top of that, a core backing asset requirement is the higher of 5% of the pool or the highest daily redemption percentage in the past 180 redemption days. The pool sits in a statutory trust for holders, an issuer may keep an excess of at most 5%, and a payment order for a valid redemption has to go out by the end of the next business day. Those rules come into force on 25 October 2027. The earlier consultation, CP25/14 of May 2025, had already said the FCA would not prescribe specific asset compositions beyond that floor, and that it intended to stop issuers passing interest on the pool to holders.
The Bank of England's regime for systemic sterling stablecoins is stricter, because it's designed for coins big enough to matter to the payment system. Its November 2025 consultation proposed 40% of backing assets in unremunerated deposits at the Bank and 60% in short-term gilts, with holding limits of £20,000 per individual and £10 million per business. The June 2026 policy statement softened both. The central bank deposit floor fell to 30%, the interest-bearing ceiling rose to 70%, and the per-person limit was replaced by a temporary issuance guardrail of £40 billion per coin, with launch expected in 2027.
Set Tether's 30 June 2026 balance sheet against either rule and it fails on the deposit line, not the Treasury line. Bank deposits of 0.02% aren't 5%, let alone 30%. Set Circle against the FCA's rule and the question is what's in the 15.4% outside the reserve fund, which the 10-Q doesn't itemise. Neither issuer is applying for a UK licence with its current pool. The point is that regulators who've studied the run record have converged on cash location as the binding constraint on stablecoin reserves.
What a depeg costs a portfolio depends only on sleeve size
A stablecoin sleeve in a portfolio is usually there for 1 of 2 reasons. Either it's parked between crypto positions, or it's earning a yield on a lending platform. In both cases the holder is treating it as cash. The table converts the 2 historical depegs into portfolio outcomes, assuming the whole sleeve is marked at the low and nothing is sold. These are arithmetic on the fetched prices, not forecasts.
| Stablecoin sleeve | USDC at 86 cents (March 2023), 14% haircut | TerraUSD at $0.7934 (7 to 8 May 2022), 20.7% haircut | Total loss |
|---|---|---|---|
| 2% of portfolio | 0.28% | 0.41% | 2% |
| 5% of portfolio | 0.70% | 1.03% | 5% |
| 10% of portfolio | 1.40% | 2.07% | 10% |
| 20% of portfolio | 2.80% | 4.13% | 20% |
The 86-cent USDC print was a secondary-market low that recovered after the US backstop announcement. A holder who redeemed at par through Circle lost nothing. A holder who sold on an exchange at the low locked in the 14%. The Terra column is different in kind. The mechanism never recovered and the coin's market capitalisation fell $18.47 billion by the end of May, so the 20.7% print on 8 May understates the eventual loss. That's why the table carries a total-loss column.
Two things follow. First, a stablecoin sleeve sized like an emergency fund, say 10% or 20% of a portfolio, carries a tail loss that no bank balance carries. It's a claim on an issuer, not a deposit, and the FSCS protection limits that apply to a bank balance don't apply to it. Second, the sleeve's risk is nothing like the bitcoin allocation it usually sits next to. A stablecoin's worst day is far smaller than the ones in the crypto drawdown record, but it arrives without warning and, for an unreserved coin, without recovery.
The case that reserves are the wrong thing to worry about
The Bank for International Settlements makes the strongest argument against the framing of this piece. Its June 2025 annual report chapter on the future monetary system argues that stablecoins fail as money not because reserves are thin but because they "often trade at varying exchange rates." That breaks the principle that a dollar is a dollar regardless of who issued it. On that view the 86-cent print is a design flaw of bearer tokens, not a reserve problem, and no reserve rule fixes it.
The BIS also flags the macro side. It estimates that a $3.5 billion increase in stablecoin market capitalisation can already depress Treasury yields by around 2.5 to 5 basis points, with effects up to 3 times larger during redemption episodes. Tether's $114.96 billion of bills makes it a large, price-insensitive holder that has to liquidate during a run. The BIS frames that as a financial stability issue. For a holder it cuts the other way: the reserve most likely to be sold in a stress is the one everyone else is also selling.
The strongest pro-stablecoin reply is the March 2023 record itself. A coin with 8% of its cash trapped in a failed bank still redeemed $3.8 billion at par within the week. That happened with no deposit insurance and no central bank facility of its own. Reserves worked. What failed was the secondary market, which priced a bank failure at 14 cents on the dollar when the exposure was 8 cents at most.
What the attestations can't tell you
An attestation is not an audit of the issuer. BDO's opinion covers the reserves report and the total assets and liabilities at a single time and date, 30 June 2026, and states that activity before and after was not considered. Tether's secured loans and other investments, $18.70 billion together, are described in footnotes rather than itemised. The liabilities figure also uses a contractual redemption value that takes account of Tether's 10 basis point redemption fee, so a holder redeeming directly gets 99.9 cents on the dollar.
Circle's numbers sit inside an SEC filing, but the 10-Q doesn't break out the 15.4% held outside the reserve fund, and it's more than 2 months old by the time this is published. Both issuers report in US dollars, so a UK holder has currency risk on top of reserve risk. And 2 depegs is a sample of 2. Neither happened during a broad market crisis, which is the scenario the BIS is worried about and the one no reserve has been tested against.
What would change the conclusion
If Tether's buffer stopped covering its volatile assets. At 30 June 2026 the $4.11 billion excess absorbs a 16.7% combined fall in gold and bitcoin. If the next attestation shows the buffer below that line, the coin is exposed to the varying exchange rates the BIS describes on the way down, which is exactly what the buffer exists to prevent.
If an issuer moved 5% or more into on-demand deposits. That's the FCA's floor and it directly addresses the SVB failure mode. Tether's 0.02% is the figure to watch. A move toward 5% would change the reading of its balance sheet more than any change in its Treasury holdings.
If a reserved coin depegged during a wider crisis. March 2023 was a single bank failure that US authorities backstopped within days. A run coinciding with a Treasury market dislocation, where the reserve itself is hard to sell, is the case the record doesn't contain. If that happens and a coin holds par, the stablecoin reserves question is answered. If it doesn't, the sleeve-size table above is the only arithmetic that matters.
The figure to check each quarter is the one that moves: excess reserves over liabilities, and what sits in the deposit line. LedgerTouch shows a stablecoin sleeve at its market price rather than at $1, which is the difference between a 5% weight and whatever it's worth on the day.