The PYUSD reward subsidy, and what Sky actually has to do with PYUSD

TL;DR — the decision rule

Size a PYUSD lending position on its in-NAV yield only, and treat the reward leg as a one-week option that is being deliberately wound down. On the Morpho “Paypal USD Main” vault, 3.09 of the 4.98 percentage points of headline yield (62%) is an off-NAV Merkl distribution funded a week at a time by a 1-of-1 Safe whose own budget traces to freshly-issued PYUSD. The campaign is a top-up-to-target design — it buys a specific headline number — and that target has been ratcheted 6.0% → 5.7% → 5.5% over the last four weeks. The rewards are real (top holders have claimed 98–99.7% of everything accrued), pre-funded, and land within hours; they are simply not durable, and they are not principal. Separately: there is no PYUSD peg-stability module with Sky, and there never was. The Paxos↔Maker PSM in the governance record is PSM-PAX-A, whose gem is USDP — a different Paxos token — and it sits at a zero debt ceiling. The real Sky↔PYUSD linkage runs through Spark’s balance sheet, which is discretionary and can be withdrawn, not a 1:1 mint/redeem backstop. Do not credit PYUSD with a Sky-backed peg floor it does not have.

Question

Two things the standard eval routine does not check, both load-bearing for the morpho-paypal-usd-main position:

1. The reward leg. Most of the quoted yield is an external distribution. Before sizing on it — or discounting it — we need to know who pays, from what budget, on what commitment horizon, whether it actually reaches depositors, and where it is heading.
2. The Sky/Maker PSM. A PYUSD peg-stability module would be a hard peg-defense and exit rail: 1:1 convertibility into DAI/USDS at scale, independent of Paxos’s own redemption desk. If it exists, PYUSD’s tail is much shorter than it looks. If it does not, the belief that it does is a mispricing we should not inherit.

Both generalise: every issuer-subsidised stablecoin venue we look at (Aave PYUSD, Euler PYUSD, the Sei USDC campaign) has the same structure, and “protocol X has a PSM with Maker” is a claim that gets repeated far more often than it is true.

Methodology

# reward campaign metadata
curl -s https://api.merkl.xyz/v4/opportunities/1-ERC20LOGPROCESSOR-0xb576765fB15505433aF24FEe2c0325895C559FB2/campaigns

# the funding transaction (gross vs net, Merkl's cut)
cast receipt 0xfa728dd50685fe0cff98fbe06f1bdc494fadfad882f4ab1e7aaf8a73eb12e5c2

# Sky: does a PYUSD ilk exist at all?
cast call 0x5a464C28D19848f44199D003BeF5ecc87d090F87 'list()(bytes32[])'
cast call 0x5a464C28D19848f44199D003BeF5ecc87d090F87 'gem(bytes32)(address)' $(cast format-bytes32-string 'PSM-PAX-A')

Findings

1. The reward is an issuer-side subsidy, renewed weekly

0x4307…1609 — a 1-of-1 Safe (getThreshold() = 1, sole owner the EOA 0xfc05…5e87) — has created 30 PYUSD campaigns against this vault since 2025-12-23, plus 2 MORPHO campaigns from a separate 3-of-7 Safe that stopped on 2026-03-31. Total budgeted across all 32: 5,813,886 PYUSD.

Each PYUSD campaign is exactly 7 days. There is no multi-week commitment anywhere in the record.

The most recent funding transaction, 0xfa728dd5… (block 25,589,256, 2026-07-22 15:47 UTC), shows the full mechanism in one place: the Safe sent 187,150 PYUSD to the Merkl Distributor (0x3Ef3…D9Ae) and 2,850 PYUSD to Merkl’s fee recipient (0xeAc6…7DAb) — a 1.5% protocol fee on 190,000 gross. Four sibling campaigns for other Sentora vaults were funded in the same transaction on the same 1.5% terms, for 368,000 PYUSD gross per week in total, of which this vault takes 51.6%.

Tracing the budget upstream: the funder Safe has received 14,247,133 PYUSD in 148 transfers, ~54% from its own owner EOA and ~42% from a second EOA (0x1e30…7282). Those two wallets hold only PYUSD — no USDC, USDP, PAXG or USDG — and between them received 14,710,070 PYUSD directly from the zero address, i.e. by primary issuance. Only Paxos can mint PYUSD. The subsidy therefore originates on the issuer/distributor side, not from Sentora’s treasury and not from protocol revenue. We could not confirm the legal entity behind the wallets from public sources and do not assert one.

Visible remaining budget: 0xfc05…5e87 holds 6,640,560 PYUSD and 0x1e30…7282 holds 369,553 PYUSD — 7,010,113 PYUSD combined. At 368,000/week gross that is ~19 weeks (to roughly 2026-12), absent further top-ups.

2. The campaign buys a headline number, and the number is being cut

Every campaign since 2026-07-02 carries an explicit distributionMethod: ERC4626_APR with mode: MAX_APR and a targetAPR. This is a top-up design: Merkl pays whatever is needed to lift total APR to the target, capped by the week’s budget. Two consequences follow. Reward APR rises when borrower interest falls, which flatters stability; and the headline compresses as TVL grows against a fixed budget.

The target has been walked down every week it has been visible:

The current week is budget-constrained, not target-constrained: Merkl reports dailyRewards 26,735.71 against a maxDailyRewards of 31,234.76 — the budget is fully spent and still lands ~0.5pp short of the 5.5% target. Peak weekly spend was 295,500 PYUSD in April 2026; it is now 36% below that.

Composition today: netApy 4.979% = netApyExcludingRewards 1.888% (in-NAV, net of the 1% management fee) + 3.092pp of Merkl distribution. 62% of the headline never touches share price.

3. The rewards do actually get paid — this is not a phantom yield

The failure mode worth ruling out is a campaign that accrues on a dashboard and never settles. It does not apply here. Cumulative accrual versus on-chain claims for the three largest depositors:

The residual in each case is the current epoch’s accrual. The Merkl Distributor updated its root three times in the ~10 hours before our snapshot (blocks 25,620,588 / 25,622,486 / 25,622,974), so settlement latency is hours, and the week’s 187,150 PYUSD is already sitting in the Distributor before the week starts. Pre-funded, promptly claimable, and one week long. Claiming is a manual pull — rewards do not compound into the vault and do not arrive unless someone calls claim.

4. The capital beside us has already proved it is fast

If most of the yield is rented, the question is what the co-investors do when it stops. There is a clean natural experiment in the April 2026 stress window. Reading the vault directly at historical blocks:

−63.4% of the vault redeemed in ~82 hours, and price-per-share rose throughout. This was pure flight, not a markdown: no bad debt, no haircut, every redemption honoured at NAV. Three months later the vault stands at $315.7M — 74.8% of the April peak. Read both ways: the exit machinery worked under real load, and the money next to us leaves in days rather than weeks.

Against the playbook’s mercenary-capital tests this scores badly on (a) incentive-attracted intent — 62% of yield is subsidy — and (c) exit-at-first-drawdown. It scores well on (d): the yield source is legible and the fees are visible on-chain.

5. There is no PYUSD peg-stability module with Sky

Enumerated from Sky’s IlkRegistry (0x5a46…0F87, count() = 35), the complete collateral set contains no PYUSD ilk of any kind. The PSMs that exist:

The Paxos↔Maker PSM that appears in the governance record — including the 2023 Trident Digital/Paxos proposal offering MakerDAO a 2.25% monthly payment on the notional of PSM-PAX-A — is USDP (Pax Dollar), a different Paxos-issued token from PayPal USD. It is offboarded, at a zero debt ceiling and $1 of dust. PYUSD has never had a Sky PSM. Anyone carrying “PYUSD has a Maker PSM” in their model is conflating two Paxos tokens.

6. What the Sky↔PYUSD relationship actually is: Spark’s balance sheet

The linkage is real but structurally different — it runs through Spark, Sky’s capital-allocation Star, and it is discretionary allocation rather than convertibility:

• Spark ALM Proxy (0x1601…347E) holds 237,453,670 PYUSD outright — the Spark Liquidity Layer’s PYUSD inventory, and by itself ~13% of PYUSD’s Ethereum float.
• SparkLend lists PYUSD as reserve id 17: spPYUSD supply 100,000,448 PYUSD, variableDebtPYUSD 8,859,243, so ~8.9% utilisation with 91,141,920 PYUSD sitting idle in the aToken. Supply APR 0.29%, borrow APR 3.65%.
• Spark Savings PYUSD (spPYUSD, 0x8012…D354) is an ERC-4626 whose TAKER_ROLE is held by the Spark ALM Proxy; it pays a Vault Savings Rate of 3.52% APY. Despite the $1B ambition reported at its December 2025 launch, it holds 31,936 PYUSD — and is draining fast: $797,525 (07-03) → $591,648 (07-10) → $526,866 (07-17) → $198,760 (07-24) → $31,936 (07-27), a −96% month.
• Sky’s own sUSDS runs a savings rate of 3.52%, identical to the Spark PYUSD VSR — the anchor is visible in the numbers.

So Sky/Spark is a very large discretionary lender to and holder of PYUSD, not a peg backstop. Both legs — a $237M inventory and a $100M SparkLend supply that is 91% unutilised — are balance-sheet decisions Spark governance can reverse, and neither creates a right to convert PYUSD into USDS at par.

One forward-looking tell sits inside our own vault: the curator raised the sUSDS collateral cap to type(uint128).max on 2026-05-18 (block 25,123,205), in the same batch that opened every other collateral cap. No sUSDS market cap exists, so nothing can be allocated there today — but the collateral layer is pre-authorised, and adding the market needs only a curator submit plus the 3-day timelock.

Decision framework

1. Size subsidised stablecoin venues on netApyExcludingRewards. Ask the API for the ex-rewards figure explicitly; a blended headline is not a number you can hold.
2. Pull the campaign’s targetAPR history before trusting the reward leg. A MAX_APR campaign is buying a headline. The trajectory of the target — not the current APR — is the forward signal, and here it is falling.
3. Price the commitment horizon, not the rate. Weekly campaigns are a one-week option. Compute runway as visible funder balance ÷ observed weekly gross spend and state it as a floor.
4. Verify claims, not accruals. Cumulative-versus-claimed for the top three holders separates a real distribution from a dashboard number, and takes one API call.
5. Never accept a “PSM with Maker/Sky” claim without enumerating IlkRegistry.list(). Check the gem() symbol — USDP, GUSD, USDC and PYUSD are four different tokens and only one of them has a live PSM.
6. Distinguish convertibility from allocation. A PSM is a mechanism; a protocol treasury holding your asset is a decision. Only the first survives the counterparty deciding it would rather not.

This stops applying when: the funder moves to multi-month campaigns or an on-chain streaming commitment; or the reward token stops being the deposit asset (a volatile reward token adds a price leg this analysis ignores); or Sky onboards an actual PYUSD ilk, at which point re-run finding 5 from the registry.

Caveats

We did not identify the legal entity behind 0x4307…1609 or its funding EOAs. The on-chain trace establishes that the budget arrives as primary-issued PYUSD, which only Paxos can create — it does not establish who instructed the issuance, and we make no claim beyond that. The 19-week runway assumes no top-up, which contradicts the observed pattern of routine top-ups; treat it as a floor on visible budget rather than an expiry date. We measured retention through the single April 2026 window; one crisis is one observation, and the vault was four months old at the time. We did not attempt to value the Merkl campaign’s forwardingEnabled: true setting, which allows rewards on vault shares held through a wrapper to be routed to the end holder — relevant if we ever hold this position indirectly.

The smallest fact that would falsify the core conclusion: a single campaign with a duration longer than 7 days, or a targetAPR that stops declining for two consecutive months.