USDT — Risk Report

Moderate risk · 6.0/10

The most-used dollar in crypto, held back by the least transparent issuer. USDT is the deepest, most liquid stablecoin on Earth — you can move billions in and out at peg on any chain, on any exchange — and it has held its dollar through every stress event of the last decade, including the TerraUST collapse. The catch is the issuer. Tether has never completed a full audit (a first one, by KPMG, is underway in 2026), publishes point-in-time attestations rather than continuous verification, and carries a regulatory-settlement history. Holding USDT means trusting unmatched liquidity and a ten-year track record over institutional-grade transparency.

Yield	Exit method	Primary redemption	Age	Chains
None	Sell on any CEX/DEX at peg; direct redeem is qualified-customer only ($100k+ mins)	Mint/redeem 1:1 with Tether for verified customers; retail exits via secondary market	Since 2014 — the longest-running major stablecoin	Ethereum, Tron, BSC, Solana, Arbitrum, Avalanche, Optimism, Polygon + more

Summary

USDT is the largest stablecoin in existence — roughly $183.5 billion in circulation — issued by Tether Limited, a BVI-domiciled company affiliated with iFinex, the parent of the Bitfinex exchange. It is backed primarily by short-term US Treasuries (around 77% of reserves), alongside cash equivalents, Bitcoin and gold, and it is verified by quarterly reserve attestations from BDO Italia. A first full audit — engaged with KPMG in early 2026 — is in progress but not yet delivered.

Two things define USDT’s risk. First, it has the deepest liquidity of any crypto asset, on every major venue and virtually every chain, and it has held its peg through a decade of FUD cycles, regulatory actions and market crashes. Second, its issuer is the least transparent among the blue-chip stablecoins: attestations rather than a completed audit, no on-chain proof of reserves, a complex corporate structure, and two resolved regulatory settlements in the record.

The 6.0/10 reflects that balance — unmatched liquidity and a proven peg against issuer transparency as the dominant residual risk. It lands one notch below fully-audited, US-regulated USDC, and the gap is driven by issuer and backing transparency, not by any weakness in the peg itself.

Tether’s Q2 2026 attestation, published 2026-07-31, moved one part of that picture. USDT is still over-collateralized and the peg did not budge, but the cushion of reserves above what Tether owes holders halved in a single quarter, from a record $8.23 billion to $4.11 billion. Backing was cut from 6.5 to 6.0 on that, and the overall held at 7.0 at the time — liquidity and peg resilience were untouched, and one quarter of evidence on one axis does not re-price a decade-proven asset. ⚠️ Overall has since moved to 6.0 for an unrelated reason — two axes that had never been measured were measured — see the Overall row for the current figure. The detail is in “What backs it” below.

1 · Stability — 7.5

The reassuring counterweight to all of the above is the track record. Despite a decade of FUD cycles, regulatory actions and one brief dip to about $0.9975 during the May 2022 TerraUST collapse, USDT has always returned to peg. Massive secondary liquidity plus arbitrage redemption is a robust peg defense, and it has never suffered a sustained loss of peg. When the rest of the market was on fire, USDT held.

2 · Backing — 6.0

Start with what did not happen. USDT is over-collateralized. Tether holds more assets than it owes to holders, the Q2 2026 attestation confirms it, and the peg never moved — not by a basis point that mattered. Nothing below is a shortfall, and none of it is a reason to expect one tomorrow.

What changed is the size of the margin, and how fast it changed.

Reserves are majority short-term US Treasuries — around 77% of what Tether owes, counting T-bills and reverse repo together. That is a major improvement over the 2021 commercial-paper era and it is the load-bearing part of the book. Around the Treasury core sit cash equivalents, secured loans, physical gold and bitcoin — and that outer ring is where the Q2 story is.

The cushion halved in one quarter

The excess reserve is the buffer: total assets minus everything owed to USDT holders. It is the only thing standing between a loss on Tether’s investments and the dollar you are holding. Here is what one quarter did to it:

	Q1 2026 (Mar 31)	Q2 2026 (Jun 30)
Excess reserves	$8.23B — a record	$4.11B

✅ Update 2026-08-25: the cushion has rebuilt, and this is a checked-and-held result rather than a re-rate. Re-read on 2026-08-16, the excess reserve is up about 53% off the Q2 low — a genuine recovery on the axis this report cut in August, and the direction the Backing move was watching for. Every score holds, because one quarter’s rebuild does not undo the structural point the cut was about: the buffer remains levered to gold and bitcoin, and the composition question is unchanged.

⚠️ Tether has stated a restore trigger, and it is worth having as a number rather than a sentiment: $8B. That is roughly the Q1 record, and it is the level at which the cushion returns to where it stood before the halving. A reader now has a specific figure to check rather than a direction to feel — and it is the cleanest thing on this page to watch, because it is stated by the issuer and observable in the next attestation. | Collateralization | about 104.5% | about 102.2% | | Physical gold | $19.84B / 132.2 tonnes | $18.84B / 146.2 tonnes | | Bitcoin | $6.62B / 97,137 BTC | $5.80B / 98,933 BTC | | Secured loans | about $15.9B | about $13.5B (down 15%) | | Owed to holders | about $183.5B | $183.64B (flat) |

Liabilities were flat. The buffer fell by half anyway, because the assets fell — mostly gold and bitcoin, which had a bad quarter. Gold dropped about 15%; bitcoin fell from $68,200 to $58,600.

That is the structural point worth taking away: the cushion is roughly 6× levered to two volatile assets. Gold and bitcoin together are worth about $24.6 billion, sitting on top of a $4.11 billion buffer — so a combined drawdown of a bit under a fifth in those two positions would, on paper, consume the entire surplus. Q2 was that arithmetic running in slow motion. Not a crisis; a demonstration.

Two things make it worth a score change rather than a shrug:

Tether grew the volatile book in the same quarter the cushion halved. It bought 14 tonnes of gold and 1,796 BTC — $1.91 billion of buying — while the buffer was falling. Prices went down; the position sizes went up. That is a deliberate choice about risk posture, not something the market did to them. It may well be a good long-term bet. It is still the opposite of what a conservative reserve manager does when the margin is thinning.
About $1.9 billion left the reserve and nobody has said where it went. Run the arithmetic on prices alone: Tether earned $1.5 billion in operating profit, and gold and bitcoin were marked down $3.74 billion, so the buffer should have landed near $5.99 billion. It came in at $4.11 billion. The gap is around $1.9 billion. The likely explanation is mundane — distributions to shareholders, or transfers into Tether’s separate investment arm — but “likely” is doing real work in that sentence, because no one has confirmed it.

Why that gap is the best argument in this report

Look at the two numbers moving in opposite directions. Gold and bitcoin were marked down $3.74 billion on price, and Tether bought $1.91 billion more of them. That purchase is the same money as the “growing the volatile book” point above, seen from the other side of the ledger — and it is exactly what makes the leftover impossible to pin down.

The size of the unexplained gap depends on where the cash for those purchases came from, and an attestation cannot tell you:

Funded from inside the reserve — Tether shifted existing reserve assets into gold and bitcoin — then the unexplained outflow is about $1.9 billion.
Funded by new capital put in from outside, the purchases are not an outflow at all, and the unexplained amount is about $3.8 billion.

The evidence favours the first, which is why $1.9 billion is our central estimate: the $1.80 billion of gold buying alone exceeds the entire quarter’s $1.5 billion of operating profit, secured loans fell $2.38 billion and freed up cash, and the Treasury book barely moved (down $0.42 billion).

But “favours” is the whole problem. Nothing in the attestation lets you distinguish “Tether moved $1.9 billion of reserve into gold” from “$1.9 billion left the reserve” — and that is a sharper illustration of what “attestation, not audit” costs you than any bare missing sum. One of those is a risk-posture decision you might disagree with; the other is money out the door. A full audit would have to say which. So the two durable caveats below are not abstract:

Attestations, not a full audit. Tether publishes quarterly reserve attestations through BDO Italia — point-in-time snapshots confirming that assets were at least equal to liabilities on a given date. That is not a continuous Big-Four audit of reserve quality, encumbrance, or intra-quarter availability — and, crucially, it carries no obligation to explain movements. An attestation confirms the numbers added up on June 30th. It does not have to trace how they got there, so it does not have to say whether that $1.9 billion was reallocated or withdrawn — and it didn’t. A full audit would have to. A first one (KPMG) is underway in 2026 but has not been delivered.
No on-chain proof of reserves. Unlike newer stablecoins that publish a Chainlink proof-of-reserves feed, USDT’s reserves are verified only by the periodic report. You are trusting the attestor, not the chain.

The other side of the ledger

Three genuine positives in the same report, and they are why this is a 0.5 move on one axis rather than something louder:

Treasury exposure held flat at about $140.6 billion. The safe core of the reserve did not shrink; the volatile ring around it did.
Secured lending was cut 15%, from about $15.9 billion to about $13.5 billion. That is real de-risking of the least transparent asset class in the book.
USDT kept growing — issuance was up $446 million into a quarter when the stablecoin market overall contracted. Demand for the product is not the problem here.

Since June 30th both gold and bitcoin have partly recovered, which on a mark-to-market basis puts the cushion back near $4.8 billion. That recovery is a reason not to over-read a single quarterly snapshot — and also a reminder of the underlying point, since a buffer that moves that much on price alone is a buffer that depends on prices.

3 · Liquidity & Exit — 9.5

For retail, exit is the secondary market — and it is the best exit in all of crypto. USDT trades at near-zero slippage at any realistic size, on every major CEX and DEX, on every chain. Direct redemption with Tether does exist, but it is gated to verified, qualified customers, typically with $100k+ minimums; ordinary holders never touch it directly.

What keeps the peg tight is that those qualified redeemers do arbitrage. Any discount to $1 gets bought up and redeemed at par, which has reliably pulled USDT back to peg through every historical wobble. You don’t need access to the redemption window yourself — you benefit from the people who have it.

4 · Dependencies — 4.0

✅ The majority of the book is the most substitutable asset there is. Roughly 77% is US Treasuries — $114.96B of T-bills plus $25.62B of reverse repo. On composition alone this is a strong reserve, and that belongs first.

⚠️⚠️ What holds the axis to 4.0 is that the counterparty set has never been enumerated at all. No full audit has ever been completed. What exists is quarterly attestation by BDO Italia — an accounting firm, not a Big Four auditor — and an attestation reports point-in-time composition, not counterparty identity. So for a $183.64B liability book, not one custodian or banking counterparty can be named.

⚠️ That is unenumerability rather than concentration, and it is worse in a specific way: it cannot be assessed in either direction. A concentrated set that is named can be judged — a reader can look at the name and form a view. An unnamed set offers nothing to judge. The absence is the finding.

⚠️ Two further pieces of the set are unnamed by construction. About $13.5B of secured loans is a counterparty book with no disclosed borrowers. And the iFinex / Bitfinex affiliation means some portion of the arrangement may not be third-party at all.

Why this sits below frxUSD’s 4.5, which has ~90% in a single custodian: Frax’s custodians are named. A named concentration is more assessable than an unnamed one, even when the named concentration is severe.

✅ The counter-argument, because overstating risk is not the safe direction either. 77% Treasuries is a great deal of good collateral, and it is the majority of the book. If the loan book and the custodian set were disclosed and clean, this axis would plausibly be a 6.0. What is missing is disclosure, not demonstrated quality.

⚠️ Deliberately not priced here: the gold ($18.84B), the bitcoin ($5.80B), or the 6.0× volatile-asset-to-surplus ratio. Those are market risk and are already carried by Backing above. Importing them would price one fact on two axes.

5 · Contract & Admin — 3.5

⚠️ On Ethereum, USDT’s owner is a 3-of-6 multisig with no delay of any kind. Measured 2026-09-11 from the token outward:

Token 0xdAC17F958D2ee523a2206206994597C13D831ec7 — symbol() USDT, totalSupply() 88,306,028,997.33, deprecated() false, upgradedAddress() zero
owner() → 0xC6CDE7C39eB2f0F0095F41570af89eFC2C1Ea828, a contract
That contract returns required() = 3 with six owners, all externally-owned accounts with live non-zero nonces
getThreshold() fails ABI decoding — the positive discriminator for the older MultiSigWallet surface. It is not a Gnosis Safe, and is not classified as one by address shape.

✅ The permission was measured, not inferred. eth_call of deprecate() reverts from 0x…dEaD and succeeds from the owner, with a fabricated selector absent from the bytecode as a control. A selector present in bytecode proves a function exists; it does not prove who may run it. This establishes the caller gate.

⚠️ deprecate(address) is the power that carries the score. It redirects the token’s behaviour to an address the owner selects — upgrade-equivalent, on a contract that has no proxy. It is dormant, not absent: deprecated() reads false today, and three signatures can change that with no on-chain delay. The same three also reach issue, redeem, addBlackList, destroyBlackFunds, pause, unpause, setParams and transferOwnership.

✅ 3-of-6 is a real quorum, which places this above the any-of-1 assets — USDC at 3.0 and PYUSD at 2.5, where a single signature suffices. ⚠️ And below RLUSD at 4.0, because RLUSD needs seven signatures to change code while USDT’s three do everything, deprecate included.

⚠️ Tether is the only issuer in this cohort that discloses these powers — in its legal terms rather than its technical documentation. “freezing or confiscation of any … Tether Tokens”, “blacklisting any Digital Tokens Address”, and being “under no obligation to … maintain support for any particular blockchain” map one-to-one onto destroyBlackFunds, addBlackList and deprecate. A contract read and a legal document agreeing is real corroboration, because they are unrelated sources rather than two methods sharing one operand.

⚠️ Disclosure does not soften this score. This axis prices the reaction window, and “in its sole discretion” with no stated threshold or process is the explicit absence of one. Who signs remains undisclosed, exactly as with the other four assets in this cohort.

Coverage: Ethereum, Solana and Avalanche measured. Avalanche’s token owner and proxy admin converge on the same 3-of-6 signer set as Ethereum — one signer set across two rails, not two independent ones. Solana’s mint and freeze authority both resolve to a 2-of-3 SPL Token multisig (numRequiredSigners 2 of numValidSigners 3 — read m, not n). ⚠️ Tron is Tether-declared and unmeasured. BNB, Arbitrum, Optimism and Polygon appear on this report’s chain list but are not Tether-declared as native USD₮ — they are present-but-not-issuer-declared and unread on this axis, which is not the same as clean.

6 · Issuer — 5.0

Tether Limited is BVI-domiciled, sits inside a complex iFinex/Bitfinex corporate structure, and holds no US banking license. Two resolved regulatory matters are in the record, and they are worth stating cleanly:

The 2019 NYAG finding — commingled funds and an undisclosed credit facility to Bitfinex — settled for $18.5 million.
The 2021 CFTC fine — $41 million, for reserve misrepresentation over the 2016–2019 period.

Both are settled, and both reflect a historical pattern of opacity that improved transparency has since narrowed but not erased.

USDT also sits outside the US GENIUS Act framework, and that timetable is now firm rather than speculative:

Implementing rules were due 2026-07-18.
The broad provisions take effect on the earlier of 2027-01-18 or 120 days after final rules.
From 2028-07-18, US exchanges and custodians may generally offer only permitted stablecoins. For USDT to still be on a US venue that day, Tether would need to qualify through the foreign-issuer comparability pathway.

For a US retail holder this is a two-year horizon, not a today problem. Nothing about your access changes this year, and the 2028 date is the one to diarize. But it is worth understanding Tether’s actual strategy, because it is not “get USDT compliant”. Tether’s US-regulated product is a separate token: USA₮/USAT, launched 2026-01-28, issued by Anchorage Digital Bank (OCC-chartered) with Cantor Fitzgerald as reserve custodian, and expanded to Celo on 2026-07-29. The read is that USAT exists so that USDT does not have to comply — USDT is being positioned for offshore and emerging-market demand, where most of its use already is. That is a coherent business plan. It also means USDT itself is unlikely to ever pick up a US regulatory wrapper.

What you actually earn

Nothing — plain USDT pays no yield. It is a settlement and trading dollar, not an income product. If you want yield on a Tether-adjacent dollar, that lives in separate staked or lending products, not in USDT itself.

Audits & security

No full financial audit has ever been completed. The KPMG engagement (2026, with PwC assisting on internal-systems readiness) would be the first, and it is the single biggest forward catalyst for this score. Today the verification layer is BDO Italia’s quarterly attestations — useful, but not the same thing.

The token contracts themselves are simple, long-lived and heavily battle-tested. A common framing is that the residual risk here is “reserve and issuer transparency, not contract bugs” — true as far as it goes, and misleading, because issuer control over the contract is a third thing that is neither a bug nor a disclosure question. It was measured for the first time on 2026-08-23.

On Ethereum, USDT’s owner is a 3-of-6 multisig with no delay of any kind. The owner address 0xC6CDE7C3…Ea828 is a contract, and its selector set identifies it as the 2016-era Gnosis MultiSigWallet rather than a modern Safe (required(), getOwners(), confirmations() and MAX_OWNER_COUNT() present; getThreshold(), VERSION() and masterCopy() absent). It returns required() = 3 against six owners, and all six are plain externally-owned accounts. There is no delay surface anywhere in its bytecode — no executionDelay, delay, MINIMUM_DELAY, GRACE_PERIOD or timelock. It submits, confirms, executes and revokes, and nothing else, so execution fires inside the transaction that lands the third confirmation. This is not a short delay; it is no delay, by construction.

What those three signatures can do — all owner-gated, all present in the token’s own bytecode, against an $88.3B Ethereum leg: pause(), issue(), redeem(), setParams(), addBlackList(), destroyBlackFunds(address), which permanently destroys a blacklisted balance rather than merely freezing it, and deprecate(address), which sets a deprecation flag and forwards every ERC-20 entry point to an arbitrary new contract.

deprecate() deserves singling out, because it is stronger than a proxy upgrade in one specific respect. A proxy at least separates the admin from the implementation, and those can be different parties under different constraints. Here there is no separation to exploit and no admin role that could be held by someone else — three signatures replace the behaviour of the token itself, in one call, immediately.

Keep this in proportion, in both directions. A 3-of-6 of independent keys is a materially harder set to compromise than a single key, these powers have existed since 2017, and their use has been narrow and overwhelmingly law-enforcement blacklisting. Nothing here is a bug and nothing suggests misuse. But a reader told that contract risk reduces to “not bugs, and otherwise a transparency question” would not know any of it exists, and on an $88.3B position that is the wrong impression to leave.

Two comparisons place it fairly, and they cut opposite ways. In Tether’s favour: this multisig is better evidenced than most. All six owners are known addresses, individually verified as independent externally-owned accounts — where thBILL, which carries a 3-of-5 and is docked for the same missing delay, does not disclose its signers at all. A quorum you can enumerate is worth more than one you are asked to take on trust. Against Tether: the powers are broader than USDC’s, over a larger book. USDC’s freeze immobilises a balance; destroyBlackFunds permanently destroys it, and deprecate replaces the token’s behaviour outright.

⚠️ Scope this carefully: the above is Ethereum only. The earlier claim that the contracts are battle-tested “across every chain USDT lives on” was never evidenced, and this pass does not evidence it either. Tron carries the larger USDT float and runs a different contract that this coverage has not read. Treat the admin posture on Tron and other non-EVM deployments as unmeasured rather than as either better or worse than Ethereum’s.

Score breakdown

Dimension	Score	Notes
Peg Mechanism	7.5	Standard fiat mint/redeem, proven across a decade including TerraUST contagion. High redemption minimums are offset by the deepest arbitrage network in crypto.
Backing	6.0	Down from 6.5 (2026-08-03). Reserves remain majority US Treasuries (about 77%), but the Q2 2026 attestation halved the surplus from $8.23B to $4.11B (104.5% to 102.2% collateralization) while the issuer bought a further $1.91B of gold and bitcoin into a book now roughly 6× the cushion, and about $1.9B left the reserve unexplained. Still attestations only (BDO Italia, not Big Four), no on-chain proof of reserves, and a history of misrepresentation settlements.
Liquidity	9.5	Best-in-class — deepest stablecoin liquidity on every major CEX and DEX, the highest trading volume of any crypto asset, on virtually every chain. Near-zero slippage at any realistic size.
Issuer	5.0	Two resolved regulatory settlements (NYAG + CFTC), BVI domicile with a complex iFinex/Bitfinex structure, no completed full audit, and USDT sits outside the US GENIUS Act framework. Offset by 10+ years operating, improved transparency, and no sustained depeg. These are entity facts and they carry the 5.0 on their own. ⚠️ The on-chain control surface is not priced here — the undelayed 3-of-6 holding `deprecate`, `destroyBlackFunds` and blacklist authority is argued under 5 · Contract & Admin, and counting it on both axes would price one fact twice.
Overall	6.0	Held at 7.0 (2026-08-03). Unmatched liquidity and a decade of peg resilience carry real weight, and neither was touched by the Q2 print — peg and liquidity scores are unchanged, and the drawdown that halved the surplus was substantially a gold and bitcoin move that has since partly reversed. Backing comes down one notch on a single quarter of evidence, which is documented here rather than compounded into the headline score. Issuer transparency and regulatory history remain the dominant residual risks. A second consecutive quarter of surplus erosion in Q3 would take this to 6.5. ⚠️ Held at 7.0 on 2026-08-23 despite the Issuer cut, and the reason should not be left to inference. This composite is liquidity-weighted rather than floor-tracking — it already sits above Backing 6.0 and well above Issuer, because Liquidity 9.5 and Peg 7.5 carry it, and neither was touched by the admin finding. The Issuer move is also a recalibration on a drag that axis never counted, not a deterioration in Tether’s conduct. And the one trigger already allocated to this number is a backing trigger; collapsing an unrelated issuer half-notch into it would blur the signal a reader is being told to watch. (Contrast AUSD, whose Overall tracks its weakest axis and therefore did follow its Issuer cut.) ⚠️ Cut 7.0 → 6.0 on 2026-09-11, and nothing about USDT got worse. Dependencies 4.0 and Contract & Admin 3.5 were measured afterwards, pulling the axis mean to 5.92. ⚠️ 6.0 sits +0.08 above that mean and is stated rather than hidden — 5.5 would have surrendered severity to rounding rather than to evidence. ✅ What holds it at the top of its range is liquidity 9.5 and a peg demonstrated through TerraUSD, FTX and the SVB window. ⚠️ What stops it going higher is that two axes are unenumerable rather than merely low — no independent audit of the counterparty set, no disclosure of who signs the 3-of-6 — so their downside is unobserved, not bounded.

How the Issuer cut was decided, and why it is not about quorum size. When this finding was first published on 2026-08-23 it was documented but left unpriced, pending a pass across every report in this coverage with a measured upgrade path. That pass has run, and the discriminating variable is not the signing threshold — it is whether a holder gets a reaction window. USDS holds the same upgrade capability and is not docked, because its path carries a verified 48-hour delay — ⚠️ a window that is narrower than this comparison once implied: USDS’s own walk finds an undelayed UsdsJoin route and no independent canceller, so its 48 hours is notice rather than veto (see USDS axis 5). ✅ The distinction still holds, and it holds on the thing the dock actually prices: two days of public warning against none at all. ⚠️ Qualified 2026-08-25: that path’s cancel power is not independent — schedule and cancel answer to the same authority — so the 48 hours is notice a holder can act on rather than an interruption a third party can apply. The cohort distinction survives the correction, because the discriminating variable was always whether a holder gets a reaction window, and USDS still gives one where the docked cases give none. thBILL has a 3-of-5 quorum and is docked, on an axis rationale that names “no timelock on admin actions” outright. So a quorum does not offset the absence of a delay; USDS is the control that shows what does. USDT has no delay at all, which places it on the same side as USDC and thBILL despite a harder signer set. The quorum credit is real and is paid in the prose above rather than in the number — three independent, individually verified keys is a better posture than one key or than five undisclosed ones, but it does not hand a holder time to act.

Who it’s for

Anyone who needs maximum liquidity and cross-chain reach, values a decade-proven peg, and accepts issuer-transparency risk in exchange. USDT is the default settlement dollar of crypto for a reason — if your priority is moving size at par, anywhere, this is the deepest market there is.

Who should avoid

Anyone who requires audited, on-chain-verifiable reserves or a regulated US issuer — prefer USDC.
Anyone who needs an EU-MiCA-compliant stablecoin. Tether has declined to seek MiCA authorization, and EU venues are acting on it: Revolut announced a USDT delisting for EU users dated 2026-08-31, with a staged wind-down converting residual balances to fiat. ⚠️ That date has passed and the outcome is not verified here. This page does not claim the delisting completed, and does not claim it was abandoned — the announcement is the established fact; the execution is unchecked. If you hold USDT on an EU platform, check your venue’s timeline now rather than later.
Anyone uncomfortable holding the single most systemically concentrated asset in crypto.

What to watch

The Q3 2026 attestation, due around the end of October. This is now the decisive datapoint. A second consecutive decline in the surplus — or a surplus under about $3 billion, or collateralization under about 101.5% — would argue for a further cut to the overall. ⚠️ That trigger was written against an overall of 7.0 and has not been re-derived since the number moved to 6.0, so treat the direction as live and the arithmetic as pending. A rebuild toward $8 billion would argue Q2 was a drawdown artifact and nothing more.
Whether the missing $1.9 billion repeats, and whether Tether ever publishes a distribution policy. A one-off transfer is unremarkable. A recurring, undisclosed drain would be an issuer problem rather than just a backing one, and would hit a second axis. It would also swamp the post-quarter price recovery described above.
Gold and bitcoin prices. Unusually for a stablecoin, these are worth watching directly: the two positions are roughly 6× the cushion, so a large drawdown in either is the actual mechanism by which USDT’s buffer erodes. As of 2026-08-03 both had partly recovered (bitcoin around $63,000, gold around $4,063/oz), putting the marked cushion near $4.8 billion.
The KPMG full audit. Still the biggest forward-positive. It was engaged in March 2026, has no delivery date, and Tether describes it as a multi-year initiative — so do not expect it soon. A clean first audit would plausibly lift Backing to 7.0+ and Issuer to 6.5; the scores are held until it lands.
Any change to the Ethereum owner multisig, or a delay appearing on it. Today it is an undelayed 3-of-6 of plain keys with total authority over the token’s behaviour. Owner changes are visible on-chain, and a timelock or a larger/harder signer set appearing would be the single change that most improves the contract-authority picture. A deprecate() call would be the most consequential single transaction in USDT’s history and would be immediately visible.
⚠️ Whether the Revolut wind-down executed as announced — its stated date of 2026-08-31 has passed unverified, and it is the first thing to check before this page quotes it again. Then further EU delistings, and any sign Tether pursues the GENIUS foreign-issuer pathway for USDT itself rather than ceding the US market to USAT.

This report is based on Tether’s public attestations — most recently the Q2 2026 Consolidated Reserves Report (BDO Italia, published 2026-07-31, as of 2026-06-30) — plus reserve reports and on-chain data through 2026-08-03. USDT’s reserves are attested, not fully audited (a KPMG audit is in progress), and the reserve composition shifts over time. Corrections or attestation links welcome at info@tidresearch.com.

Revision history

2026-08-25 — admin comparison recorded; no score change. Set against USDS’s 48-hour GSM delay, USDT’s upgrade path carries no equivalent notice window.
2026-08-23 — reserve composition restated. Short-term US Treasuries are around 77% of liabilities counting T-bills and reverse repo together.
2026-08-03 — Backing 6.5 → 6.0 following Tether’s Q2 2026 attestation; Overall held at 7.0.