syrupUSDC as Morpho collateral — the shared node between our Maple and PYUSD positions

TL;DR — the decision rule

Holding Maple syrupUSDC and holding the Sentora PYUSD vault are not two independent positions — they are the same credit, entered from opposite sides. $118.1M of syrupUSDC (10.0% of its supply) sits inside the Morpho Blue singleton as collateral for $111.1M of stablecoin borrowing across four markets, every one of them at 91.5% LLTV, and $67.5M of that — 61% — is the single market our PYUSD vault supplies. Every one of those markets prices syrupUSDC off Pool.convertToAssets(), i.e. Maple’s own mark-to-model NAV, and two of the four apply no price feed at all beyond it. That NAV has only ever risen, so the collateral leg is structurally incapable of signalling stress ahead of Maple: the first honest downward mark is the liquidation event, arriving simultaneously in four markets against borrowers at 91.5% LTV. The rule: treat a syrupUSDC position and any PYUSD/AUSD/RLUSD-vault position lending against it as one exposure for sizing, and cap the combined line at what you would accept as direct unsecured Maple credit — because on a write-down that is exactly what it becomes. This stops applying if any of these markets migrates to an oracle with an independent syrupUSDC price leg, or if LLTVs come down materially from 91.5%.

Question

The July-2026 refresh of maple-syrupusdc surfaced a holder we had not recorded: the Morpho Blue singleton (0xBBBB…FFCb) is the #3 syrupUSDC holder at 10.0% of supply. That is not a depositor — it is syrupUSDC pledged as collateral. Separately, morpho-paypal-usd-main Risk #2 already flags ~21% of that vault lent against syrupUSDC. The two evals had been describing the same node from opposite ends without either saying so.

The question generalises past these two venues: whenever a yield-bearing vault share becomes collateral elsewhere, who bears the write-down, in what order, and does the collateral oracle give any warning? The playbook’s contagion lens (steps 5 and 7 — shared dependencies, and reuse-as-collateral) asks exactly this, so the answer belongs here once rather than in each eval.

Methodology

Findings

1. The surface: four markets, one LLTV, one curator cluster

Two things stand out. Every market is 91.5% LLTV — there is no conservative tier anywhere on this surface; the whole $111M is levered to within 8.5 points of the collateral mark. And the curator set is not diversified: Steakhouse appears in three of the four markets, Sentora in two, Grove × Steakhouse in two. A single curator’s judgement about syrupUSDC is load-bearing across most of the surface.

2. The oracle: every market prices syrupUSDC off Maple’s own NAV

All four set BASE_VAULT to the Maple pool itself, so the collateral price is Pool.convertToAssets() — the share price Maple’s own accounting publishes — optionally multiplied by a peg feed. The AUSD and USDC markets carry no feed leg whatsoever: syrupUSDC is marked at its ERC-4626 NAV and the loan asset is assumed to be exactly $1.

There is no independent price for syrupUSDC anywhere on this surface. That is not an oversight — syrupUSDC has essentially no secondary market (~$15M of Arrakis Uniswap V4 depth, last verified ~10 months ago), so there is no feed to point at. But the consequence is concrete: convertToAssets() has risen monotonically since launch (1.124 → 1.174533, Sep-2025 → Jul-2026, on-chain) and by construction is a deferred mark — Maple books credit losses when Maple Direct recognises them off-chain, not when a borrower’s collateral gaps.

3. What that means for loss order

Maple’s own exit is FIFO-queued and pays out at NAV — depositors ahead in the queue leave whole, so a write-down lands hardest on whoever is still holding. The collateral leg inverts the timing:

1. Maple marks down convertToAssets() in a single step (there is no gradual path; the NAV is published, not traded).
2. All four Morpho markets reprice in the same block. At 91.5% LLTV, a write-down larger than ~8.5% puts every borrower on the surface underwater simultaneously.
3. Liquidators must buy syrupUSDC and exit it through Maple’s FIFO queue — the same queue the write-down just made everyone else want to use. Documented worst case there is 30 days; the standing pool buffer was $10.6M (0.90% of TVL) on 2026-07-21.
4. Whatever liquidators cannot clear becomes bad debt to the supplying vaults — i.e. to us, on the PYUSD side.

So the same event hits a syrupUSDC holding (deferred NAV haircut, queue behind $527M of top-holder) and a PYUSD-vault holding (bad debt from unliquidatable collateral) at once. They do not diversify each other; they compound.

4. Scale check — this is a live concentration, not a hypothetical

The surface grew into what it is recently. Morpho Blue’s syrupUSDC holding over the refresh window (on-chain, balanceOf at weekly blocks): 132.0M shares (2026-06-05) → 110.2M (06-23) → 120.3M (07-07) → 100.6M (07-21) — i.e. it has run between roughly $118M and $154M all quarter, consistently ~10–13% of syrupUSDC supply. Our PYUSD vault’s own leg has held at ~$67.5M throughout.

Decision framework

Treat the pair as one line. When sizing either Maple syrupUSDC directly or a Morpho vault that lends against it, add the two exposures and size the total. The economically honest question is “how much unsecured Maple Direct credit do we want?” — because at 91.5% LLTV against a deferred, self-published mark with no secondary price, the collateralisation converts to unsecured on the only event that matters.

Read the collateral leg as offering no early warning. Do not treat “the Morpho market is healthy” as independent evidence about Maple. It cannot be: the market’s view of Maple is Maple’s view of Maple. The genuine early signals live on the Maple side — pool buffer, queue depth, top-holder redemptions — which is where both evals’ triggers should point.

This stops applying when: any of these markets adopts an oracle with an independent syrupUSDC price leg (a real secondary feed, or a market-derived one); LLTVs drop materially below 91.5%; or the surface shrinks below ~$25M, at which point liquidation could plausibly clear through Maple’s buffer without queueing.

Caveats

• Point-in-time. One block (25,582,684). Section 4’s series is four weekly samples of the aggregate only — enough to show the concentration is persistent, not enough to characterise its volatility.
• We did not model the liquidation cascade. No sizing of how much bad debt a given write-down produces — that needs a per-borrower position pull and an assumption about liquidator appetite for a queue-gated asset. That is the next-most-useful follow-up.
• The Morpho API is trusted for enumeration only. If it omits a market, we would miss it; every market it did return was verified on-chain.
• 0xb6dd7ae2…15a2, Maple’s 44.8% top holder, is unattributed. Provenance is traced but identity is not established, so “who redeems ahead of a liquidator” is answered structurally, not by name.
• Smallest fact that would falsify the conclusion: an independent, liquid syrupUSDC price source that any of these markets could switch to. We found none — if one exists, the “no early warning” claim weakens considerably.