How is a PT (principal Token) different from an ordinary stable asset? Why can its price drop 3% on its own even when the underlying asset never loses its peg?
It's intuitive to treat a token like PT-reUSD as roughly equivalent to a Stablecoin, since it does redeem for $1 at maturity. But the difference lies in the period before maturity: a PT's market price isn't simply tracking the underlying asset — it's set by the supply and demand dynamics of the PT/YT pair inside the Pendle market. If someone buys YT aggressively (effectively betting future yield will be higher), PT's price passively falls, and that drop says nothing about whether the underlying reUSD has any actual problem.
That's exactly why the official statements in this incident specifically emphasized that reUSD itself never lost its peg — the drop in PT-reUSD's price was the normal functioning of the pre-maturity market pricing mechanism, which is a completely different question from whether the asset is actually backed. The closer to maturity, the more a PT's price gets pulled back toward $1 (since redemption certainty rises); the further from maturity — like the December-dated PT in this incident — the more room the market price has to swing.
Why does recursive Leverage looping (borrow, buy more, borrow again) exist as a strategy? If the risk is this large, why do people still run it?
The logic of recursive leverage is rational arithmetic on its own: a PT is guaranteed to accrete to $1 at maturity, so if held to term, the return is fixed and predictable. Borrowing at a lower rate to amplify that fixed spread can, in theory, turn a modest yield into a more attractive number — which is exactly what draws capital into this strategy. The problem isn't the endpoint of "holding to maturity"; it's the path getting there. Mid-way through, the position is marked against a live Oracle price, and leverage compresses how much valuation deviation you can tolerate — deviation that can have nothing to do with whether the underlying asset is actually maturing safely.
In other words, recursive leverage itself isn't the flaw — the flaw is conflating "certainty at maturity" with "price stability before maturity." That's exactly why user positions got forcibly liquidated in this incident: not because reUSD actually had a problem, but because PT's market quote experienced a one-time technical dislocation before maturity, and recursive leverage had compressed users' tolerance for that dislocation down to nearly zero.
Concretely, how do I judge whether a PT lending market's Oracle design is problematic? Are there steps I can actually check myself?
Three pieces of public information can be checked directly. First, find the Pendle pool page for the relevant PT and note total liquidity. Second, find the Morpho (or other lending protocol) market page where that PT is used as collateral, and note current total collateral and total borrows. Third, divide total borrows by pool liquidity — if that ratio exceeds roughly 2x, the lending market's scale has clearly outgrown what the pricing pool can absorb. In this incident, that ratio was roughly 5.8x, meaning a single trade that moved the pool's quote could drag the entire lending market with it.
Beyond the size comparison, check the oracle's averaging window length. A 15-minute Time-Weighted Average Price (TWAP) means an attacker — or ordinary market volatility — only needs to hold a distorted quote for 15 minutes to trigger liquidations; the shorter the window, the cheaper it is to trigger an attack or an accidental cascade. Some protocols use longer one-to-four-hour windows specifically to make it far more costly to move the quote in a short burst — this is a concrete, checkable parameter you can compare before deciding whether to enter a given PT market, not an abstract risk rating.
If I'm not running recursive Leverage myself and just deposited funds into a vault marketed as "automatically optimizing yield," am I still exposed to this kind of event?
Yes — and that's precisely the cohort most easily overlooked in this incident. Before the event, a governance proposal on Morpho's forum had already called for continuous disclosure of collateral transparency, leverage-loop exposure, and liquidity-deterioration warnings for vaults, explicitly noting that passive vault depositors don't choose to take on recursive leverage risk themselves, yet may be indirectly exposed to the leverage structure of the same lending market without realizing it. In this incident, passive depositors ultimately didn't suffer real losses — borrowers' collateral was seized, lenders were made whole — but the exposure gap flagged in that proposal was real; it simply wasn't triggered this time.
Practically, what you can check: does the vault's page disclose which specific lending markets it allocates funds into, what collateral types those markets use, and whether the vault itself or other participants sharing the same market are running recursive leverage. If a vault's description only says "automatically finds the best yield" without disclosing the underlying allocation, that opacity itself is a signal — not a guarantee something will go wrong, but a sign that you can't verify the risk yourself the way the four checks in this article let you do for a market you'd enter directly.
In the early hours of August 25, 2026, a roughly $320,000 trade triggered more than $36 million in on-chain liquidations within 14 minutes. No hack occurred, no Smart Contract was exploited, and the underlying stable asset never lost its peg — every Liquidation executed exactly according to protocol rules. That's precisely what makes this incident worth dissecting: it demonstrates that "clean execution" and "a safe outcome" are two different things. Every line of code ran correctly, and a cohort of users still got forcibly wiped out.
The mechanism sits inside Pendle. Pendle splits a yield-bearing asset into two components: a principal token (PT), similar to a zero-coupon bond that redeems for $1 at maturity, and a yield token (YT), representing the variable yield stream until that date. The two prices move like a seesaw — the more the market bids up YT, the higher the implied annual yield, and the more the paired PT gets priced down correspondingly.
Between 04:28 and 04:37 UTC, one wallet ran 11 consecutive trades, converting roughly $320,000 of SY-reUSD into more than 9.5 million YT-reUSD. That burst pushed the implied annual yield from around 11% to above 20%, dragging PT-reUSD down about 3%. This wasn't an attack on Pendle's mechanism — it was the mechanism repricing exactly as designed.
The real problem wasn't the 3% — it was the Leverage structure of the Morpho lending market pricing off it. At the time, the Pendle pool for PT-reUSD held roughly $8.97 million in liquidity, while that same PT was backing roughly $67.5 million in collateral and $52.2 million in borrows on Morpho — a lending market nearly six times the size of the pool setting its price. Pricing a much larger lending market off a shallow pool is a structural risk in itself, not something that appeared out of nowhere that morning.
Stacked on top of that was leverage: users deposited PT-reUSD as collateral, borrowed stablecoins against it, used the borrowed funds to buy more PT-reUSD, redeposited, and repeated — a recursive loop that pushed positions right up against a 91.5% liquidation loan-to-value (LLTV). On-chain analysts found at least one borrower sitting at 90.9% LTV against that 91.5% threshold — a health-factor buffer of less than 0.7 percentage points from forced liquidation. A 3% price move is lethal to a position with less than 1% of buffer remaining.
First, correlated positioning: everyone running the recursive loop held the same collateral, borrowed the same Stablecoin, and clustered near the same LLTV ceiling — a single price update could hit the entire cohort simultaneously. Second, the buffer was deliberately minimized: recursive leverage strategies are optimized to reduce idle capital, so the more efficient the loop, the closer the Health Factor sits to 1.0 and the smaller the tolerable price deviation becomes. Third, the liquidation incentive wasn't large — Morpho's Liquidation Bonus at 91.5% LLTV works out to roughly 2.6% — but it was enough to clear the entire eligible book within minutes.
If you're running — or considering — a recursive leverage loop on any PT-style token, this incident hands you four checkable parameters. First, the Oracle's averaging window: a 15-minute TWAP on a shallow pool is a cost an attacker only has to sustain for a few minutes. Second, the pricing pool's depth relative to the lending market's total borrows — if total borrows exceed pool liquidity by more than roughly 2x, the oracle is structurally undersized. Third, whether the liquidation LTV is appropriately calibrated against time to maturity — the further out the maturity date, the more room a PT's implied yield has to reprice, and the wider the buffer should be. Fourth, your own health-factor buffer against the liquidation line — measured not against "an average day's volatility" but against the largest single trade the pool could realistically absorb. All four numbers are publicly checkable in about ten minutes — far cheaper than discovering after the fact that your position was already sitting on the edge.