⬡ Federation Research Log — Week 03
--
The recursive leverage stack Three protocols, one loop, and what historical leverage cascades teach us about looking through it Originally published
2026–05–06 on Paragraph · Galactic Federation of Finance
Aave V3 currently holds more than four billion dollars of Pendle Principal Token collateral. That single fact is the start of a story that almost no DeFi user can see clearly, because the protocols involved are usually analyzed one at a time.
This essay analyzes them together. Three protocols — Ethena, Pendle, Aave — composed into a single loop that synthesizes leverage out of yield. None of the three protocols is broken. The loop they create when stacked is the largest latent structural exposure in 2026 DeFi, and it is almost entirely invisible to the indexers, dashboards, and TVL charts that most users rely on.
The Federation does not predict what happens next. The Federation reads the architecture, names the mechanism, and publishes the analysis before the cascade clarifies it for everyone.
Each protocol is mature. Each has been audited extensively. Each has a clean operational record at scale. Read separately, each is a credible piece of DeFi infrastructure.
- 1. A user mints USDe by depositing collateral with Ethena. They stake it, receiving sUSDe. The position now earns Ethena’s basis-trade yield.
- 2. They deposit sUSDe into Pendle and receive a PT-sUSDe token at a discount to the principal.
- 3. They post that PT-sUSDe as collateral on Aave V3, where it has been added to a stablecoin eMode at high LTV.
- 4. They borrow USDC against the PT collateral.
- 5. The borrowed USDC is used to mint more USDe.
- 6. Return to step 2.
Each pass through the loop creates additional USDe supply, additional Pendle PT supply, and additional Aave debt. The same underlying capital is recycled through three distinct accounting layers, each of which sees the position as fully collateralized and structurally sound.
The result, by mid-2026, is approximately $4B of PT-sUSDe collateral inside Aave V3 alone, the majority of which is the second, third, or fourth pass of capital that originated as a much smaller starting position. Pendle reports that 75% of its TVL is sUSDe-derived. Ethena reports that the majority of recent USDe demand growth comes from Pendle PT positions rather than direct holders. Aave reports rising PT collateral concentration in its quarterly risk reviews.
Each protocol’s metrics are accurate. The composite picture is not visible from any single protocol’s dashboard.
Standalone leverage is dangerous when the collateral falls. Reflexive leverage is dangerous when the yield source falls.
The loop is sustained by Ethena’s basis-trade yield being meaningfully positive. When perp funding rates are positive, longs pay shorts, and Ethena collects. When funding rates compress toward zero or invert, the yield disappears.
If sUSDe yield falls below the borrowing cost on Aave, the loop’s economics reverse. Holders of looped positions become structurally short their own carry trade. Some unwind voluntarily — they sell PT back to the market, return USDC to Aave, redeem USDe, exit. Each unwinder pushes PT prices down marginally, raising effective borrow costs marginally, and pulling the next marginal holder closer to unwind.
In stable conditions, the loop holds. In conditions where funding rates compress for a sustained period — as occurred briefly in March 2026, and as occurs reliably in any meaningful crypto bear cycle — the loop tightens from yield-positive to yield-negative across all three layers simultaneously. Aave’s PT-collateral pools see utilization spike. Pendle’s PT secondary markets see depth thin. Ethena’s USDe sees redemption pressure on the underlying basis trade.
Whether that pressure clears gracefully or cascades depends on how concentrated the unwinders are, how deep PT secondary liquidity remains, and how quickly Aave’s risk parameters can adapt. None of those variables is observable in real time from public dashboards.
Recursive yield stacks are not new. They have a recognizable shape:
In all three cases, the protocols involved were not technically broken. They functioned exactly as designed. What broke was the composite structure when the yield assumption underneath it changed faster than the unwinding mechanism could absorb.
The Ethena-Pendle-Aave stack is a sophisticated variant of the same shape. The components are better-audited and better-capitalized than their predecessors. The mechanism is identical.
Three observations the Federation considers important to document publicly.
The right disposition is awareness, not alarm. Users with positional exposure to any of the three protocols benefit from understanding what their position is part of. Users without exposure benefit from understanding why TVL charts may not be telling them what they appear to be telling them.
The Federation’s published positions on lending design are recorded in the liquidation specification, currently at v0.2. Section 3 of that specification names a principle that applies directly to the loop analyzed above: the protocol must surface the actual position to the user, not the abstraction the user thinks they hold.
That principle is hard to implement when the position spans three protocols and the user is interacting with only one of them at a time. It is the kind of design problem that gets harder, not easier, as DeFi composes deeper.
The Federation will continue researching this design space publicly, in this Research Log and in revisions to the liquidation specification, before any code is written. The next revision is expected to address Section 7 — non-standard collateral types — which is where this kind of recursive composition would be governed if a Federation lending product ever shipped.
Until then: research first. Always.
- Admiral Zoran Voss Galactic Federation of Finance
The Federation Research Log publishes weekly on Wednesdays. No paywalls. No price predictions. No roadmap announcements. Analytical, not promotional. ⬡ Sources referenced: Ethena Labs documentation; Pendle Finance protocol documentation; Aave V3 risk reports (Q1 2026); DeFiLlama protocol-level TVL methodology; Anchor Protocol post-mortem (Terraform Labs, 2022); Curve Finance governance forum archives (2022 incident reviews).
Originally published at https://paragraph.com.
