Risk Research · Hyperdrive · HyperEVM
A risk-first deep dive into Hyperdrive's redemption-based lending primitive for yield-bearing collateral.
Correlated Markets are a new lending primitive designed by Hyperdrive for yield-bearing collateral. Instead of relying on external liquidators and DEX liquidity, collateral redeems directly into the borrowed asset, shifting liquidation risk from market microstructure (slippage, liquidity gaps, liquidator incentives) to collateral drawdown during the redemption window. This structure enables lending against assets that traditional liquidation mechanisms struggle to support, including liquid staking tokens, vault strategies, and tokenized real-world assets.
01 — Overview
DeFi lending has been one of crypto's great success stories. The ecosystem has grown to over $50B of outstanding borrows at its peak through largely permissionless mechanisms keeping both borrowers and lenders safe. Much of this growth is built off three core mechanics: properly pricing collateral via oracles, managing risk via appropriate loan-to-value ratios, and keeping protocols solvent through liquidations.
Source: DefiLlama
This process works well for liquid assets like ETH, SOL, and the various flavors of BTC, however, it has struggled with rapidly growing categories of on-chain assets: yield-bearing tokens, and real-world assets. These tokens share a common trait: they're redeemable for an underlying asset, but that redemption takes time. Unstaking LSTs takes days to weeks depending on the network, and withdrawing tokenized RWAs can take weeks to months.
Source: DefiLlama
The traditional lending model breaks down here for a few reasons. First, traditional DeFi lending assumes the presence of instantaneously available DEX liquidity, or more generally, the willingness of someone to quickly offboard and warehouse liquidated collateral. Maintaining deep AMM liquidity for these assets is expensive and typically requires incentive programs that bleed value from the issuer. Because secondary market liquidity for these assets is often thin and subject to fluctuations, many protocols price these assets to NAV. However, if NAV and secondary market pricing fluctuate enough, liquidators may not have the necessary incentive to facilitate liquidations.
Together, these dynamics handicap these assets, forcing them into low LTV and/or small supply-capped markets.
Correlated Markets are designed for redeemable assets with deterministic exit paths, a category that includes liquid staking tokens, certain vault strategies, and tokenized real-world assets.
02 — Mechanism
A Correlated Market is an isolated lending pair featuring a collateral asset which redeems directly into its debt asset. A borrower deposits collateral (e.g., rETH) in order to borrow the underlying (e.g., ETH), while lenders can supply the underlying to earn interest. One of the primary differences between this new primitive and existing lending markets is how liquidations happen. Instead of a liquidator seizing the collateral and repaying a user's debt in an atomic transaction, the protocol itself closes out the borrower's debt, seizes the collateral, and begins the redemption process.
For a time, while the market is redeeming the collateral asset to the debt asset, lender deposits are partly comprised of the lent asset (in this case ETH) and the redeeming collateral asset (rETH). When the redemption process finishes under normal circumstances, the amount of lent assets will be greater than the debt repaid through the liquidation, representing a liquidation bonus paid to the lenders directly.
Rather than relying on market participants to absorb collateral during stress events, the protocol itself initiates the redemption process and waits for the underlying asset to be delivered. This removes dependency on external liquidity but introduces a delay between liquidation and settlement. This shift changes the risk profile of the system in several ways:
| Dimension | Traditional Lending | Correlated Markets |
|---|---|---|
| Liquidation speed | Seconds (atomic) | Days to weeks (redemption period) |
| Liquidation mechanism | External liquidators sell on DEX | Protocol redeems collateral natively |
| Liquidity dependency | External liquidators + DEX liquidity | Native redemption pathway |
| Primary risk | Slippage, liquidity gaps | Collateral drawdown during redemption window |
| LTV constraints | Limited by DEX liquidity and slippage | Limited by redemption window volatility |
| Liquidation surplus | Goes to liquidator (MEV) | Goes to lenders (yield boost) |
In practice, this means Correlated Markets are particularly well suited for assets that have a reliable redemption mechanism into the borrowed asset, exhibit limited drawdown risk over the redemption window, and lack deep or reliable secondary market liquidity. Assets that do not meet these criteria remain better suited for traditional lending architectures.
Rather than replacing existing lending models, Correlated Markets extend the design space, enabling lending against a category of assets that traditional liquidation mechanisms struggle to support.
03 — Mechanism
When a position breaches its liquidation threshold, the protocol initiates a deterministic sequence:
Breach (Position hits liquidation threshold) → Seize (Protocol claims the borrower's collateral) → Redeem (Collateral enters the native redemption queue) → Settle (Redeemed assets repay debt) → Surplus (Excess value flows to lenders as boosted yield)
Liquidations under this format have redemption risk: lenders temporarily hold redeeming collateral while it converts to the debt asset.
A small initiation reward is paid to users who call liquidate(user) on unhealthy positions to incentivize starting the liquidation process. Upon initiation, the borrower's collateral is seized in full and enters the native redemption queue - there is no fixed liquidation bonus or discount as in traditional DeFi lending. Instead, the entirety of the seized collateral redeems to repay the debt, and any surplus is distributed to lenders.
Because Correlated Markets remove the need for rapid market liquidation and always redeem collateral at NAV, eliminating slippage as a variable, they can safely operate at higher LTVs than traditional markets for appropriate collateral. If the collateral's value declines during the redemption window, the realized surplus shrinks, and in extreme cases could turn negative, representing a loss to lenders.
04 — Mechanism
While the liquidation process does not require any secondary liquidity, winding down and voluntarily exiting a position would require a very lengthy period of withdraw, redeem, repay without it. However, healthy borrowers can access the same pool of liquidity as liquidatees via "deleveraging".
A borrower can voluntarily trigger a similar redemption process used in forced liquidations: the protocol accepts your collateral, immediately crediting that against your debt for a small fee. At that point, from the borrower's perspective the process is complete. Behind the scenes, and from a lender's perspective, the collateral is redeemed through the native mechanism.
This method of deleveraging will have a fee with two components:
The time-value component compensates depositors for losing access to their capital during the redemption window. For example, if we assume 7% APY for HLP, a 4-day redemption process (without yield) would have a cost of approximately 0.077%. For HYPED at 2.5% APY, a 7-day redemption period would cost about 0.05%.
The volatility premium compensates for the risk that collateral value drops during redemption. For HLP, using the 95th-percentile 4-day drawdown, this is approximately 0.20%. HYPED and other LSTs, which have no mechanism for their exchange rate to decrease, have lower fees.
Total base fees: approximately 0.28% for HLP and 0.10% for HYPED. An optional dynamic component scales with utilization, discouraging deleveraging when pool liquidity is tight and offering discounts when abundant. This is capped at 1% to prevent fees from spiraling at extreme utilization levels, and a utilization cap of 85% prevents deleveraging from consuming all available pool liquidity.
Deleveraging Cost Estimator
Drag the utilization slider to see how deleveraging fees scale for HLP and HYPED.
Utilization cap for deleveraging: 85%. Dynamic fee max: 1%.
05 — Mechanism
The redemption process created by the correlated market creates something interesting: a claim on a known value at a known future date. Instead of the lending pool always waiting 4 to 14 days for collateral to redeem, it can sell that redemption claim immediately at a discount. A buyer steps in, pays slightly less than face value today, and collects the full amount at maturity. The buyer earns a time-value yield; and the protocol obtains instant liquidity while capturing some of the liquidation fees.
By facilitating this process, particularly when utilization is high, the protocol can better balance liquidity and duration risk with accumulated fees for lenders.
| Scenario | Face Value | Discount | Purchase Price | Maturity | Ann. Yield |
|---|---|---|---|---|---|
| HLP calm (4-day) | $100,000 | 0.30% | $99,700 | 4 days | ~27% |
| HLP stress (4-day) | $100,000 | 2.00% | $98,000 | 4 days | ~186% |
| RWA (14-day) | $100,000 | 0.75% | $99,250 | 14 days | ~20% |
06 — Markets
Hyperdrive's first two Correlated Markets demonstrate two very different use cases this primitive can serve. Together they illustrate the flexibility of the design.
HYPED is Hyperdrive's liquid staking token for HYPE. Staking HYPE yields HYPED; unstaking returns HYPE after a 7-day unbonding period. This creates a closed loop: the collateral (HYPED) redeems directly into the debt asset (HYPE).
Under the current HyperEVM design, the exchange rate between HYPED and HYPE has no mechanism to decrease during the unstaking window - the rate can only increase as staking rewards accrue. This means the cool-down window drawdown risk effectively approaches zero for this pair. However, this assumption depends on the continued absence of slashing or other protocol-level changes to HyperVM.
HLP is a community-owned vault on the Hyperliquid perpetuals exchange that functions as a market maker and liquidity provider. The HLP strategy targets delta-neutral exposure, earning yield from trading fees and funding while hedging directional risk. Delta-neutral does not mean zero-volatility. HLP's NAV fluctuates due to directional moves, funding rate dislocations, liquidation cascades where HLP takes the opposing side, as well as tail events.
Historical analysis over 2+ years suggests HLP is appropriate for this market. Over this period, 74% of daily returns have been positive, with the worst single-day return at approximately -0.8%. But intra-day drawdowns can be steeper and faster; in March 2025, HLP lost roughly 4% in just over an hour during a liquidation event. A key metric for Correlated Markets is the 4-day rolling drawdown (the time to unstake HLP): historically, the 95th percentile sits at approximately -0.20%, and the 99th percentile at -0.86%. These are modest in absolute terms but represent real risk that lenders and borrowers must price in this market.
07 — Risk
Market risk in correlated markets is fundamentally different from traditional DeFi lending. Liquidations, the core mechanism through which protocols offboard unhealthy positions and protect against bad debt, are materially slower. Traditional DeFi markets can complete this process within a single block while correlated market liquidations take days or weeks depending on the asset.
Because HLP, unlike HYPED, can (and will) decrease relative to its underlying (USDC), its risk parameters require careful calibration. To size HLP's LTVs and liquidation buffer, scenarios were simulated across empirical and stressed volatility assumptions (1.5x observed), as well as heavy-tailed distributions designed to model extreme outcomes beyond anything observed.
Source: Allez Labs HLP Correlated Market Report
Data window: HLP daily NAV from January 2024 through February 2026 (~26 months). Sampling: Block Bootstrap uses overlapping 4-day blocks resampled with replacement to preserve autocorrelation in returns. Historical Sampling draws individual daily returns independently. Stressed scenario: Scales observed volatility by 1.5x before simulation. Cauchy: Models infinite-variance fat tails, included as a worst-case bound rather than a realistic expectation. Limitations: 26 months of data captures only one major stress regime (March 2025). Bootstrap methods preserve the autocorrelation structure observed in the sample but cannot generate correlation patterns absent from the training window.
Under base-case simulations, the probability of breaching the buffer is near zero - an 8.9% 4-day drawdown has never been observed in HLP's history. Even under stressed assumptions the breach rate stays below 0.5%, and a worst-case heavy-tailed model yields only 2.6%. The buffer is sized to absorb scenarios meaningfully worse than anything observed to date. That said, HLP's exact strategy is unknown, and past drawdown behavior may not reflect future performance - both lenders and borrowers should weigh this.
Lenders in a Correlated Market have a more direct relationship with the underlying collateral than in traditional lending. In a standard market, liquidation is someone else's problem - a third-party liquidator seizes and sells the collateral, and lenders are repaid. Here, lenders hold the redeeming collateral during the liquidation window and bear its price risk directly. This means lenders need to evaluate the collateral itself, not just the interest rate.
Because liquidations consume pool liquidity (redeemed collateral temporarily replaces lendable assets), lenders should expect more frequent periods of elevated utilization than in traditional markets, particularly during stress events when multiple positions liquidate simultaneously. During these periods, lender withdrawals may be delayed until redemptions settle. There is no empirical history of this market yet, so the frequency and severity of these episodes remains unknown.
For HLP specifically, the qualitative risks compound this: HLP's strategy is opaque and controlled by the Hyperliquid team, meaning risk models are calibrated to observed returns rather than anything mechanistic. If the strategy changes without notice, historical calibration becomes unreliable. Both launch markets (HLP and HYPED) also carry direct counterparty exposure to Hyperliquid - platform-level events (security incidents, regulatory action, downtime) could impair collateral values and the redemption pathway simultaneously.
Because the liquidation mechanism redeems collateral at NAV rather than selling into secondary markets, slippage, normally a major constraint on LTV parameters, is eliminated. This allows Correlated Markets to safely offer higher LTVs than traditional lending markets would for equivalent collateral, meaning more capital efficiency and more available leverage. Deleveraging provides a voluntary exit at a calculable cost. The tradeoff is that liquidation, once triggered, seizes the entirety of the borrower's collateral with no partial liquidations. The surplus flows to lenders, not back to the borrower. At high LTVs, borrowers operate with thin margins and need to monitor positions actively.
08 — Summary
Hyperdrive Correlated Markets replace the dependency on external liquidators, DEX liquidity, and spot-price oracles with a single mechanism: redeem the collateral and repay the debt. This trades speed for certainty and enables lending against assets that traditional protocols struggle to serve. Zero-coupon bonds extend the model by turning redemption delays into tradeable claims. The protocol is in early stages with conservative supply caps, and its risk models need validation through real market cycles before broader conclusions can be drawn.
This article is for informational purposes only and does not constitute financial advice. Parameters, fees, risk conditions and models described here are subject to change. For full terms, see allez.xyz/tos.