There is a new vault on Usual. Markov Labs curates it. The yield comes from two of DeFi's deepest markets: ETH perpetual funding on Hyperliquid, and ETH credit on Aave. You deposit USDC. The strategy holds one leg in each market, nets its ETH exposure to zero, and collects the net carry between the two.

The roles are distinct: Aave provides the credit market, Hyperliquid the perpetual market, Markov the model that ties them together, Usual the access. What is new is the packaging. The trade itself is well known; it had never been offered as a curated vault arbitraging lending and borrowing markets.

Why this yield exists

Perpetual futures are where most of crypto's trading happens. When more traders want to be long than short, longs pay shorts a recurring fee called funding. On ETH, that fee has been positive most of the time: demand for leverage has run ahead of the capital willing to take the other side.

Collecting it sounds simple. Short the perpetual, hold the asset, pocket the fee. In practice the trade spans two venues with independent margin systems. It needs continuous hedging, margin moved between the legs to keep either side away from liquidation, and the discipline to sit through stretches where the fee turns negative. Most holders will not run that operation. That is why the spread has persisted.

Access has been the other problem. For most, collecting funding has meant an outright perpetual position: the funding is visible, but the position is a directional bet. The dominant packaged alternative hedges the direction but discloses its collateral and venue mix in aggregate, not rate by rate. This vault packages the hedge and leaves both markets' rates in view.

What the vault does

The construction has two legs, sized so that ETH exposure nets out. Delta-neutral, in trading terms.

Half the capital goes to Aave, the largest on-chain credit market. It is converted to ETH, supplied as collateral, and doubled through borrowing: the vault borrows USDC against the ETH and adds to the position until it holds ETH collateral equal to the full deposit, with borrowed USDC worth half of it outstanding. A 50% loan-to-value. The other half sits on Hyperliquid as margin behind a short ETH perpetual, sized to match the collateral. One unit long, one unit short. Exposure nets to zero by construction; as prices move it drifts, and rebalancing pulls it back.

Three flows remain. The vault receives funding on the short. It earns supply yield on the Aave collateral. It pays interest on the borrowed USDC, at half the notional. Net carry = funding rate + ETH supply yield, minus half the borrow rate.

Before costs, that formula is the whole engine. Trading fees, gas, and rebalancing slippage reduce what is realized, and they grow when volatility does. The rates themselves are public: all three, on Aave and on Hyperliquid, at any time.

The rate model

Two forces drive this trade, and they behave differently. Markov's rate model treats that asymmetry as the core design input.

Funding is volatile and mean-reverting. Fitted to realized Hyperliquid data, its long-run average is positive and its half-life is measured in days. In that history, dips below zero were shallow and short-lived. So the vault does not unwind on negative funding: selling into a trough turns what the model treats as a temporary drag into a realized loss and gives up the expected recovery.

Borrow cost is the persistent side. Aave's USDC rate moves with utilization and can stay elevated for extended periods. A sustained borrow spike is what the model treats as structural rather than transient. That is the unwind trigger, and it is published in advance. Because it is published, a depositor can check the exit condition rather than rely on the manager's judgment.

What the backtest shows

Markov replayed twelve months of realized rates, July 2025 to July 2026, on a modeled $10 million: Hyperliquid's funding history, and Aave's supply and borrow rates read directly from the contracts.

The simulated result, gross of execution costs: 7.14% over the period, 7.12% annualized, with 0.48% annualized volatility and a maximum drawdown of 0.23%. Over the same window, sUSDe, the closest packaged delta-neutral carry product, returned 4.89% net of its fees. SOFR, the dollar risk-free reference, returned 3.99%. A full net-of-cost accounting is in progress and will be published. A backtest is a reconstruction of the past, not a forecast.

Carry was negative on 16.9% of days, and all of those episodes together cost 23 basis points. The low volatility follows from the construction: the P&L tracks a rate differential, not the price of ETH.

What is new here

Most packaged yield in DeFi shows you an APY, not the rates behind it.

The vault introduces no new infrastructure. Aave has priced on-chain credit for years. Hyperliquid runs the deepest perpetual book in DeFi. Enzyme's Onyx provides the vault rails. What did not exist was the assembly: a curated, on-chain-financed, single-asset carry vault where a depositor can see the rates that actually generate the yield. No one had packaged that before, and none of these protocols could have offered it alone.

Markov curates it the way it curates credit markets across Morpho, Euler, and Fira: the methodology, the backtest, the rate model, and the unwind policy are published. Markov's standard is that every parameter is the output of a model you can read. The name comes from Markov chains, processes where the future depends only on the present state.

The terms

No performance fee and no management fee for initial depositors; both will be revised once deposits scale. A 0.15% fee on exit. Deposits enter a queue and are approved by the manager, so both legs deploy together and the position is neutral at entry. There is no idle cash buffer: all capital sits in the position, which is why there are no instant redemptions. Exits settle in USDC on a T+1 basis. No epoch lockups, no maturity dates.

The risks

Market-neutral is not riskless. Funding can sit below the borrow drag for a stretch, and a sustained Aave borrow spike inverts the carry: that is the designated exit. Both legs carry liquidation exposure in opposite directions, which is why the position needs active margin management between venues. Neutrality drifts between rebalances and has to be swept back. Execution costs rise exactly when markets get volatile. The Aave supply leg is yield from utilization, not a guarantee. Oracles can misprice the book. Regulation around offshore perpetuals can move. Hyperliquid is a younger venue than the credit protocols behind Markov's other products. The vault deposits, borrows, and settles in USDC, so a material depeg would hit both sides of the book. Everything here is smart contracts: the vault, Aave, Hyperliquid. And curation itself is a risk: the model, the execution, and the discipline are Markov's to get right. The full list of risk factors is in the documentation. Read it before you deposit.

Where to find it

The Markov Funding Rate Carry Vault runs on Ethereum mainnet under the ticker mkFRA, at contract 0x64423193CFdB25bF87b0fc63aFEC2575c61f6bc0, on Enzyme's Onyx vault infrastructure. You reach it from the Usual app, where it is listed with Hyperliquid as the underlying protocol and Markov as curator, or through Markov's vault page. The methodology, backtest, and rate analysis are at docs.markovlabs.xyz.

The position, the rates, and the exit rule are public. Own your money.

Backtested performance is simulated, not realized, and does not predict future returns. Digital asset strategies can lose money. Read the vault documentation before depositing.