Pendle V2: A Permissionless Yield-Trading Protocol
Executive summary
Pendle V2 is the most successful yield-tokenisation protocol in DeFi history, and its whitepaper is the cleanest specification of the PT/YT split primitive. The paper formalises a Standardized Yield (SY) wrapper that normalises yield-bearing assets, splits them into Principal Tokens and Yield Tokens, and prices YT through a custom AMM whose invariant is designed for time-decay assets. The math is the strongest part of the paper — the Notional-style implied-rate AMM is genuinely novel — and the governance design (vePENDLE plus gauge voting) is competently specified. The paper systematically underestimates how much its success would depend on a single 2024 catalyst (LRT-points-farming) that the design did not anticipate. DI rates V2 as the right design at the right time: a mechanism built before the market existed, that captured the right market when it arrived. The paper deserves credit for solving a genuinely hard pricing problem; it deserves criticism for not engaging with the LP economics of YT volatility, which became the dominant operational concern post-2024.
Background and motivation
Yield-bearing assets had become a dominant share of DeFi total value locked by 2021-22 — Lido's stETH, Aave's aTokens, Compound's cTokens, Curve LP tokens, GMX GLP — but the market had no efficient way to trade the yield component separately from the principal. A user holding stETH was implicitly long both ETH price exposure and the Lido staking yield curve. A user wanting only the staking yield (e.g., to hedge against ETH price risk) had no clean instrument; a user wanting only the principal exposure with no yield exposure was equally underserved. Traditional finance solves this with strip securities (zero-coupon Treasuries, MBS interest-only and principal-only strips), but DeFi had no equivalent until Element Finance's 2021 protocol introduced a primitive PT/YT split. Element's design used a balancer-style weighted pool to price PT — the YT side was almost an afterthought — and the protocol struggled to attract liquidity. Pendle's V1 (also 2021) used a similar approach with Sushi-style pools. The V2 redesign explicitly identifies the V1 limitation: passive AMM pools price PT and YT poorly because their value drifts deterministically toward maturity, and constant-product pools penalise LPs for not actively rebalancing against this drift. The intellectual lineage on the math side is more academic. The Yieldspace AMM design (Yield Protocol, 2020) had specified an AMM invariant that prices implied rates rather than nominal token amounts, with the property that LP positions are insulated from the deterministic drift. The Pendle V2 paper adapts this Yieldspace-style invariant for PT pricing and adds an original twist for YT: rather than pricing YT directly, the protocol prices YT off the PT pool using the no-arbitrage relationship PT + YT = SY. This off-ledger pricing is what makes V2 tractable — separately liquid PT and YT pools would fragment liquidity unworkably. The motivating question is precise: how do you build an AMM for time-decay assets that does not punish LPs for the drift, while supporting both PT and YT in a single composable system?
Core technical contributions
Section 3 specifies the SY (Standardized Yield) wrapper. SY tokens are a generalised interface that turns any yield-bearing asset into a uniform format with three operations: deposit (in underlying or yield-bearing), redeem, and exchangeRateCurrent. The wrapper is technically a thin ERC-5115 contract, and Pendle was the proximate cause of ERC-5115's standardisation. Section 4 specifies the PT/YT split. A user deposits SY and receives equal quantities of PT and YT. PT redeems for one underlying-equivalent at maturity. YT accrues all yield generated by the SY between split and maturity, then expires worthless. The decomposition is mathematically clean: at any time t before maturity T, value(PT_t) + value(YT_t) = value(SY_t), so any two of the three prices determine the third. Section 5 is the AMM. Pendle V2 trades PT against SY (not against the underlying) using a Yieldspace-derived invariant of the form: x = SY balance, y = PT balance, t = time-to-maturity, and the invariant is x^(1-t) + y^(1-t) = k. As t approaches zero (maturity approaches), the invariant approaches a 1:1 swap, reflecting that PT-at-maturity equals SY exactly. As t approaches 1 (very long-dated), the invariant approaches constant-product, reflecting maximum uncertainty. The mathematical innovation is that an LP's position value is invariant to the deterministic drift toward maturity, so LPs do not experience drift-IL. This is qualitatively different from Uniswap V3 ranges and is the technical core of the paper. YT pricing is derived: given a PT/SY pool spot price and a fixed exchange-rate function for SY, the implied YT price is computed by no-arbitrage. Trading YT triggers an internal flash-loan-style flow: buy SY, split into PT+YT, sell PT into the pool, pocket YT minus the slippage cost. The paper specifies this flow precisely. Section 7 specifies vePENDLE governance: PENDLE locks for up to two years, voting weight scales with lock duration, and weekly gauge votes direct PENDLE emissions to specific markets. This mirrors veCRV but adapts the gauge voting to maturity-based markets.
Methodology and rigor
The paper's mathematical specification is unusually rigorous for a DeFi protocol whitepaper. The AMM invariant is derived, not asserted; the no-arbitrage YT pricing is proved; the LP drift-invariance property is shown explicitly with calculus. This is closer to academic rigor than the typical whitepaper format and reflects that the Pendle team includes researchers who engage with the underlying math seriously. The paper does have rigor gaps. The LP economics under YT-volatility shocks are not analysed. PT pricing is robust to small drift, but YT prices can move substantially within a single block during yield-rate spikes (as occurred during the 2024 LRT points farming wave), and the LP exposure to this volatility was not modelled in the paper. The implied-rate framing assumes the underlying SY exchange rate evolves smoothly; SY assets with discontinuous exchange-rate updates (e.g., emergency-paused Lido, USDC depegs propagating to aUSDC) violate this assumption and the paper does not analyse the consequences. The paper also does not analyse the interaction between vePENDLE gauge voting and bribe markets, which became a substantial fraction of protocol revenue post-Curve-Wars but with poorly-understood incentive distortions. Pendle's actual gauge market today operates more like Convex's veCRV proxy than the paper anticipates. Empirically, the paper does not present any backtests, simulations, or deployed-protocol data; V2 was a forward-looking specification at publication. Three years of production data have largely validated the math, with the LP-volatility caveat noted.
Strengths
The strongest contribution is the AMM invariant itself. The Yieldspace-derived invariant is genuinely well-suited to time-decay assets, and Pendle's adaptation for the PT-versus-SY trading axis is correct and elegant. The fact that LPs do not experience drift-IL is a real and material property that has supported Pendle TVL through multiple market regimes. Second, the SY abstraction was the right architectural choice. By normalising yield-bearing assets behind a uniform interface, V2 became composable with any new yield-bearing primitive (LSTs, LRTs, perp-DEX vault tokens, RWA tokens) without protocol upgrades. The same V2 contract code that traded stETH yields in 2023 trades sUSDe and rsETH yields in 2026 with no modifications. Third, the maturity-based market lifecycle is operationally clean. Markets are launched with a fixed expiry, principal redeems automatically at maturity, and new markets relaunch on the same SY. This produces a predictable rotation rather than the perpetual-position complexity of most DeFi primitives, and it gives users clear lifecycle expectations. Fourth, the permissionless market creation has aged extraordinarily well. By April 2026, Pendle has hosted markets for over 200 distinct SY assets across nine chains, and the permissionless listing primitive has been the operational engine of growth. Fifth, the paper is precise enough to be implementable from the spec alone, and the V2 codebase is essentially a faithful realisation of the paper. Sixth, the V2 design captured the LRT-points-farming meta of 2024 cleanly: the YT-of-LRT pattern (where buying YT-eETH gave leveraged exposure to EigenLayer points) generated the largest single-product TVL spike in DeFi history, taking Pendle TVL from $200M to over $7B in six months.
Weaknesses and limitations
The paper's most significant limitation is its under-analysis of LP economics under YT-volatility shocks. When YT prices spike (as during LRT points wave), LPs in the underlying PT/SY pool can experience adverse selection from sophisticated YT traders without realising it. The realised LP yields are visibly lower than naive fee-based projections, and a meaningful fraction of LP TVL is now operated by Pendle itself or specialist managers rather than passive depositors. The paper does not anticipate this distributional outcome. Second, the SY assumption of smooth exchange-rate updates is violated by an increasing number of yield-bearing assets, particularly those with reset events, emergency pauses, or oracle-driven rate updates. The protocol has had to add per-asset adapter logic for non-trivial SY assets, which complicates the originally-clean abstraction. Third, the maturity-rollover process generates user-experience friction: positions held through maturity require either manual redemption or auto-rollover, and the mechanism for rolling positions across maturities is not specified in the paper — it has been built ad-hoc by Pendle and ecosystem participants. Fourth, the vePENDLE gauge voting market has been substantially captured by Penpie/Equilibria-style proxy protocols (mirroring the Convex/veCRV pattern), which the paper did not anticipate or design against. The bribe-revenue-to-vePENDLE-locker flow is now the dominant value-accrual mechanism, but its dynamics are out-of-paper. Fifth, the paper does not analyse the systemic risk of large LRT-PT positions during de-pegging events. The October 2024 sUSDe partial depeg produced a brief crisis in PT-sUSDe markets that the protocol survived but only with manual intervention. Sixth, the paper does not engage with the interest-rate-derivatives literature seriously; it could have benefited from a more explicit comparison to Notional, Element, Sense, and Voltz, all of which made adjacent design choices.
Subsequent influence
Pendle V2 is now the canonical yield-trading primitive in DeFi. Spectra (formerly APwine), Sense Finance, and a handful of newer protocols compete on adjacent designs but Pendle dominates volume and TVL by a substantial margin. The SY standard influenced ERC-5115 and indirectly shaped how new yield-bearing assets are designed for composability. The PT/YT split has been adopted as a primitive by lending protocols (Morpho's PT-collateral pools, Aave's PT-eMode integration), points-trading platforms (Whales Market, Resolv), and structured-product protocols (Origami, Spectra). The vePENDLE gauge model has been imitated more selectively, with a meaningful refinement being Penpie/Equilibria-style proxy protocols that capture vePENDLE governance the way Convex captured veCRV. The PT-as-collateral pattern in particular has had outsized influence: by 2026 over $4B of PT tokens are deposited as collateral in lending protocols, creating a synthetic fixed-income layer that did not exist before V2. The LRT-points-farming meta of 2024 was almost entirely a Pendle phenomenon; protocols like ether.fi, Renzo, and Kelp DAO designed their points programmes assuming Pendle YT-trading was available, and the Pendle-mediated points TVL inflows substantially exceeded what would have happened without yield-tokenisation. The paper has seeded broader interest in DeFi-native fixed-income markets.
How it holds up in 2026
Pendle V2 is one of the cleaner success stories in DeFi protocol design. Three years post-launch, the V2 contracts are still in production (with V3 in late-stage testing), the SY standard has propagated, and the PT/YT primitive has become a load-bearing component of the broader DeFi credit and structured-product stack. Pendle TVL stabilised in the $4-5B range after the LRT-points wave receded, with a more durable composition of stablecoin yield markets (sUSDe, USR, sUSDS) that look likely to persist. The math has held up. The AMM invariant has not produced any solvency or fairness issues; the SY abstraction has scaled to assets the paper's authors would not have anticipated; the maturity-based lifecycle operates cleanly. Where the paper has aged less well: LP economics, gauge bribery, and rollover UX are all out-of-paper concerns that have required substantial post-deployment engineering. V3 (in development) explicitly addresses some of these — concentrated-liquidity-style range provision for LPs, integrated rollover flows, and refined gauge voting — and is the largest in-progress refinement of the V2 design. The fundamental insight of the paper, that yield-bearing assets have a separable yield component that can be priced and traded independently, has been validated by market adoption: nearly every meaningful DeFi yield-bearing asset in 2026 has a Pendle market. DI's holistic assessment: V2 is the rare DeFi protocol whose research artefact and product execution are both excellent, and the paper has earned its place as the canonical reference for yield tokenisation.
Strengths
- Yieldspace-derived AMM invariant is correct and well-suited to time-decay assets
- SY abstraction has scaled cleanly to 200+ asset types across nine chains
- Maturity-based market lifecycle is operationally clean and predictable
- Permissionless market creation has been the operational engine of growth
- Mathematical specification is rigorous enough to implement from the paper alone
- Captured the LRT-points-farming meta cleanly, taking TVL from $200M to over $7B in 2024
Weaknesses
- LP economics under YT-volatility shocks not modelled — passive LPs underperform expected returns
- SY assumption of smooth exchange rates violated by an increasing share of real-world assets
- Maturity-rollover UX not specified in paper; built ad-hoc post-deployment
- vePENDLE gauge market captured by proxy protocols (Penpie, Equilibria) not anticipated in design
- Systemic risk from large LRT-PT positions during depeg events not analysed
- Limited engagement with adjacent yield-tokenisation literature (Notional, Element, Sense, Voltz)
Key contributions
- Standardized Yield (SY): unified wrapper interface for diverse yield-bearing assets (stETH, aUSDC, etc.)
- Principal/Yield Token split: PT pays $1 at maturity, YT accrues all underlying yield until maturity
- Time-decay AMM with implied-rate pricing — the first AMM designed for assets with deterministic price drift
- vePENDLE governance: vote-escrow tokenomics with gauge-style emissions to active liquidity pools
- Permissionless market creation: any SY can be split into PT/YT and listed without governance vote
- Maturity-based market lifecycle: markets expire, principal redeems, and new markets re-launch — clean state
TL;DR
Pendle V2's whitepaper specifies the cleanest yield-tokenisation primitive in DeFi: SY wrapper, PT/YT split, and a Yieldspace-derived AMM. Validated by three years of production and the LRT-points wave of 2024, with caveats around LP economics that the paper did not anticipate.
Get DeFi Intel research in your inbox
Weekly long-form coverage of papers, incidents, jurisdictions, chains, tokens and the people building them. Free tier covers headlines; Pro adds the analyst-grade breakdowns.
Sources
External references gathered from the body of this brief. Last reviewed 2026-05-03.