yzUSD — Risk Report

Significant risk · 4.0/10 (3.0/10 on Monad)

The 4.0 sits above its own six-axis mean — 4.0 against 3.92 — by a rounding margin; the axes are current and the headline is the judgement call.

⚠️ What is current and what is not. The on-chain figures below — supply, the vault relationship, the authority split and the reserve address — were measured 2026-08-29. The rest of the body, including the strategy description and the redemption mechanics, dates from 2026-06-08 and has not been re-read. The card shows both dates.

Yield	Exit method	Primary redemption	Age	Chains
None on yzUSD itself — yield accrues only inside the syzUSD wrapper	Secondary DEX only for retail	KYC-gated, best-effort	~12 months	Plasma (canonical), Monad, Ethereum

What this actually is

yzUSD is the base unit of Yuzu Money, a yield-bearing stablecoin protocol. yzUSD does not itself pay yield — it is the unit you hold, and yield accrues only when it sits inside the syzUSD wrapper. The peg is held by overcollateralization against an actively managed basket of curated DeFi yield strategies, with a junior tranche restricted to accredited investors and a Reserve Fund behind it.

⚠️ The single most important thing measured on 2026-08-29 is that yzUSD is barely a circulating stablecoin at all. Of 58,497,074 tokens on Plasma, 57,990,877 — 99.1% — sit inside one syzUSD vault. Float outside the wrapper is roughly 506,000 tokens, under one percent. That changes what “supply” means for every ratio computed against it: yzUSD is best read as a wrapper input, and any depth or coverage figure quoted against total supply describes a quantity almost entirely locked in a single contract.

⚠️ It is also a Plasma asset, not a multi-chain one. Yuzu’s own key-addresses documentation records “Not deployed” on every chain but Plasma:

chain	supply	since 2026-06-08
Plasma (canonical)	58,497,074.10	+38.9% (was ~42.1M)
Monad	23.77	unchanged — a stub
Ethereum	0.10	unchanged — a placeholder

Two of the three deployments hold a rounding error between them. What travels is the syzUSD wrapper, not this token.

1 · Stability

Reference: a $1 par target. yzUSD is not yield-bearing at this layer, so unlike its wrapper it has a peg to hold rather than a NAV to accrue.

It is holding it. Market 0.9999 against par, a −0.009% deviation.

⚠️ But the mark is thin and it is old. 24h volume is $24.74, and the price carries an 11.5-hour age against a wrapper priced off a live pool. A deviation struck on that little turnover is weak evidence about fair value and strong evidence about exit cost. The peg is not the interesting axis on this asset; axes 2 and 4 are.

The token barely moves — 2 Transfer events in 80,000 blocks — which is coherent for an asset where 99% of supply is locked inside a wrapper.

2 · Backing

The backing is published, and it reconciles. Accountable’s feed at yuzu.accountable.capital carries a full look-through on a 15-minute verification interval. totalSupply reconciles exactly against chain, and the satellite conservation is wei-exact.

backing    $63,747,863.77
supply     $58,497,074.10
CR            108.98%          <- read 2026-08-30 01:51 UTC
surplus     $5,250,789.67

So the asset is overcollateralized on its own published basis. ⚠️ The defensible criticism is not that the backing is missing — it is what the backing consists of, and that belongs to axis 4.

⚠️ A 108.98% CR computed over a book that is roughly 70% levered is not an 8.98% cushion in spot terms. The ratio is real; what it is a ratio of is the thing to read.

The Reserve Fund is thin, not empty: $503,428.89, or 0.86% of supply. Almost all of it is 502,675.95 aMonUSDT0 on Monad. ⚠️ A check for yzUSD on Plasma or ETH on Ethereum returns zero and looks like an empty reserve — the balance sits on a different chain in a different token, and enumerating the address’s transfer history finds it where guessing token contracts does not. The supportable criticism is that it has not grown with issuance: supply added roughly $16.4M over the same window. ⚠️ A reserve that does not scale with what it protects is not functioning as one, which is why audited_reserves: false remains correct.

⚠️ Expect the live dashboard to show a different CR from the one above. The reserve moves in steps rather than drifts — five discrete steps in the last thirty days — so a page figure and a tile figure taken hours apart can differ by around two percentage points against a 30-day range of under four. Neither is stale; they are separate moments in a series that jumps.

One measurement caution this report follows: the composition table carries its own timestamp, roughly 26 hours staler than the headline, and sums to about $63.61M against the $62.55M headline. ⚠️ So shares are quoted as shares and no CR is re-derived from them — mixing the two bases would produce a number neither source published.

3 · Liquidity & Exit

Both exit paths, and the axis takes the worse one.

Primary redemption — the binding leg. KYC-gated and best-effort. ⚠️ Reachability has never been probed, and unmeasured is not open. This is why axis 3 sits at 3.0 rather than at the 4.0 the venue leg alone would support.

Secondary — the better leg, and healthy. ⚠️ Do not read low turnover as thin depth. The Curve yzUSD/USDT0 pool on Plasma clears $250,000 inside 13 basis points:

$1k     −1.33 bps
$10k    −1.56 bps
$250k  −12.13 bps

Turnover and depth are different quantities. Low volume means nobody is trading it, not that you cannot get out — and on this leg you can, at retail size and well beyond it.

⚠️ The constraint sits one layer up, on the wrapper — and further out than the chain sizes suggest. See syzUSD: $11.0M of issuer vault stands on Monad against an enumerated $483K of swap TVL, but a routed $100k exit there still clears inside 10 bps, and the enumeration misses venues the router actually uses. The gap is real and it is a large-holder problem past $100k, not a retail one.

4 · Dependencies

⚠️ This is the axis that carries this asset’s risk, and the old rubric had nowhere to put it. The CR is fine; what stands behind it is the finding.

Exposure	Share of backing	Why it is a dependency and not just an asset
Levered loop positions	~70% — about $44.5M across 12 `_Loop` strategies	A ratio measured on a levered book is not that ratio on spot
Ethena (USDe + sUSDe)	34.6% — $22,046,996	⚠️ Roughly 4.2× the entire $5,250,790 surplus — a move impairing Ethena by a quarter erases the cushion
`yzPRIME` (Yuzu’s own)	4.8%	⚠️ Recursive — the reserve is partly backed by the issuer’s other liability, and that product runs a $4,557 surplus on $7,577,108. It has no cushion of its own to lend
`yzPP` (junior tranche)	~90% of the surplus	⚠️ Circular — first-loss capital whose value is a claim on the book it cushions, and third-party money redeemable on a 30-day window

Read the last two rows together and the cushion looks different from 8.98%. ⚠️ Ex-junior the ratio is 100.89%, not 108.98% on the underlying’s own published alternative basis — because the surplus is largely capital that can ask to leave.

Diversification is real but shallow: roughly a dozen strategies across nine chains, with the top name at more than a third. A reserve spread across many venues that all lend against the same few assets is one exposure wearing several names.

5 · Contract & Admin

⚠️ The authority is split, and the weaker half holds the value.

layer	owner	delay
yzUSD token (Plasma)	`0x21304575…b7cfbc` — an OZ `TimelockController`	2 days (`getMinDelay()` = 172,800)
syzUSD vault + ProxyAdmin (Plasma)	`0xa2a97004…` — a 4-of-5 Safe	none stated
syzUSD bridge (Monad)	`0x4ea00dc0…4a89ae` — a bare EOA	none

The syzUSD vault and its ProxyAdmin are both owned by 0xa2a97004…, verified live as a 4-of-5 Safe — not a bare key. The bare EOA 0x4ea00dc0… owns the bridge. ⚠️ So the single-key exposure is real, and its blast radius is the roughly 10.9M mirrored shares rather than the 99.1% of value sitting in the vault.

This axis scores 4.5 — the highest on this report — and the two-day delay is why. Both the yzUSD token owner and the proxy admin sit behind that single 48-hour TimelockController on Plasma (0x2130457539612ae9838e0314cda1a8abbdb7cfbc), so no bare key controls the token. Its role table was read on-chain: one 4-of-5 Safe, 0xa2a9700407934e913c840556b3d29f19cf6f203d, holds the proposer, canceller and executor roles, and the timelock is its own only administrator. The delay therefore provides a public reaction window but no independent veto — the four signatures that schedule a change can also cancel it and execute it. That same Safe is the upgrade owner for yzPP and for syzUSD’s Plasma vault, so the three Yuzu assets share one authority surface rather than three; anything treating them as separate legs is counting diversification that does not exist. MINIMUM_DELAY() and MIN_DELAY() both revert, so 48 hours is a configured setting rather than a code-enforced floor, though any reduction must itself wait out the current delay. And on Monad the axis drops to 2.0, because the mirror’s authority is the bare key with no delay.

6 · Issuer

The issuer is a named legal entity — YUZU FINTECH LLC, a Virtual Asset Service Provider registered under the Kutaisi Free Industrial Zone rules in the Republic of Georgia, named as contracting party in its own terms. ⚠️ That is an accountability anchor rather than prudential supervision.

⚠️ The operators are unnamed. There is no disclosed CEO or founder, and the people running a roughly 70%-levered book day to day remain anonymous. The seed round was $30M FDV with backers including Sam Kazemian (Frax), Michael Bentley (Euler), Wintermute and Hypernative — backers are not operators.

⚠️ The attestation channel is related-party: Accountable’s founder is an investor in Yuzu. Every backing figure on this page comes from that one source. It does not make the numbers wrong — and nothing here is a second pair of eyes. ⚠️ Two other Accountable hostnames fail DNS, so a check that stops at those returns nothing; the live host is the one under axis 2.

⚠️ Governance is disclosed with its denominator missing. The published policy states a quorum of 4 of an unstated total — 4-of-5 and 4-of-25 are indistinguishable from what is published — alongside a default action of block that does not say what is blocked. Establishing the denominator needs the signer set, which the disclosure does not carry.

Who should avoid this

Anyone who reads a 108.98% CR as an 8.98% cushion. About 70% of the reserve is levered, Ethena alone is roughly 4.2× the surplus, and ex-junior the ratio is 100.89% (axes 2 and 4).
Anyone uncomfortable with an issuer’s reserve holding the issuer’s own product. 4.8% is yzPRIME, which runs a $4,557 surplus on $7.58M of its own liabilities (axis 4).
Anyone treating the two-day timelock as protection. It governs the token, not the wrapper where the value sits (axis 5).
Anyone who needs a primary exit. Redemption is KYC-gated and best-effort, and has never been probed (axis 3).
Anyone reading the three-chain deployment as diversification. Two of the three are empty.

What to watch

Whether the Reserve Fund grows with issuance. $503K against $58.5M — 0.86% — and the ratio is what to track, not the balance (axis 2).
The loop share and the Ethena concentration. Those two, not the headline CR, are what would move this score (axis 4).
The 4-of-5 Safe’s owner set and threshold, and any MinDelayChange on the Plasma timelock — one quorum stands behind yzUSD, yzPP and syzUSD (axis 5).
Whether the wrapper’s bare EOA is ever replaced by a multisig or timelock. Unchanged across 82 days (axis 5).
Whether the operators are ever named (axis 6).

Revision history: 2026-09-22 — Contract & Admin 4.5; Overall held at 4.0. The Plasma timelock’s role table was read: one 4-of-5 Safe holds proposer, canceller and executor, the timelock administers itself, and the same Safe is the upgrade owner for yzPP and syzUSD’s Plasma vault. The delay is a reaction window without an independent veto. Nothing else re-read; last_verified stays 2026-06-08. 2026-08-29 — first measurement pass since 2026-06-08; no score change. Supply on Plasma is 58,497,074, up 38.9% over the gap, while Monad (23.77) and Ethereum (0.10) remain a stub and a placeholder. ⚠️ 99.1% of all yzUSD sits inside a single syzUSD vault, so this is a wrapper input rather than a circulating stablecoin, and every ratio quoted against supply describes a locked quantity. Backing is published and verifiable — CR 106.92% as measured 2026-08-29, $62,546,859.02 against $58,497,074.10 (⚠️ the reserve moves in steps; the body carries the current read) — with about 70% of the reserve in levered loop positions, Ethena at 34.7% of backing, then roughly five times a $4.05M surplus, and 4.8% in the issuer’s own yzPRIME, whose book runs a $4,557 surplus on $7.58M. The Reserve Fund holds $503,428.89, 0.86% of supply, almost all aMonUSDT0 on Monad. Authority is split across three layers: the yzUSD token owner is an OZ TimelockController with a 2-day delay, the syzUSD vault and its ProxyAdmin are a 4-of-5 Safe, and the bridge is a bare EOA with no delay. Scores held at 4.0, because the observations point in both directions and a re-score needs a basis the composition figures do not settle. last_verified stays 2026-06-08: only the on-chain layer was re-measured, and the strategy and redemption material has not been re-read.

Open live dashboard