HyperCore Borrow and Lend: Hyperliquid's Native Money Market Explained

Hyperliquid

There are now two completely different things on Hyperliquid that people call lending, and confusing them will cost you money.

The one most traders already know is the DeFi protocol layer: Felix and HyperLend and their peers, running as smart contracts on HyperEVM, each with its own rates, its own collateral list, and its own team.

The other one is newer and lives a layer down. HyperCore, the exchange itself, now runs its own lending book. Hyperliquid's weekly update on 7 September 2026 announced that manual borrowing on HyperCore had gone live on mainnet, following a testnet rollout. This page is about that facility: what the reserves actually look like, how the rate is set, and where it stops.

The prerequisite: you need Portfolio Margin

Nothing here is available on a standard account. HyperCore borrow and lend is a facility of Portfolio Margin, the most capital efficient of Hyperliquid's three account types, and Portfolio Margin is gated.

The eligibility rules in the Hyperliquid documentation at the time of writing are a floor and a ceiling:

• The master account needs either more than $5M in weighted volume or an account value above $10k
• Account value must stay below $25M

That ceiling has moved. Earlier alpha material capped participation at $5M of account value; the documented figure is now $25M. If you find a guide quoting the old number, it is stale.

Accounts that qualify get one unified portfolio across spot and perps instead of separate balances. Idle assets earn interest. Assets you borrow to open a position pay it, at the same rate. When a cap is hit, the account quietly falls back to non Portfolio Margin behavior rather than failing the order.

The five reserves, read from the API

A POST to https://api.hyperliquid.xyz/info with {"type":"allBorrowLendReserveStates"} at 03:21 UTC on 8 September 2026 returned five reserves and nothing else.

Converted at each reserve's own oraclePx, that is roughly $1.39 billion supplied against $263.5 million borrowed, an aggregate utilization near 19%.

Two details in that table matter more than the headline number.

The reserves do two different jobs, and the split is total. The two assets with a loan to value ratio above zero, HYPE at 0.65 and UBTC at 0.5, had exactly zero borrowed against them. The three assets people actually borrow, USDC and USDT0 and USDH, all returned an LTV of 0.0. So the book is not a symmetric money market where everything is both collateral and debt. It is collateral on one side, stablecoins on the other, and at this reading no crossover at all. If you supply HYPE here, you are posting collateral, not lending into borrower demand.

The reserve is USDT0, not USDT. The documentation's cap table says USDT. The live reserve at index 268 is USDT0, the bridged variant. They are not interchangeable when you are writing code against this.

How the rate is actually set

The documented borrow rate for the stablecoin reserves is:

borrow_APY = 0.05 + 4.75 * max(0, utilization - 0.8)

where utilization = total_borrowed_value / total_supplied_value, compounded continuously and indexed hourly to match the perp funding interval.

Read that carefully and the shape is a flat line with a cliff at the end. Below 80% utilization the rate is a constant 5%. Above it, the rate climbs by 4.75 percentage points for every point of utilization, so a reserve at 100% utilization would pay 100% APY. There is no gentle slope. There is a floor and then a wall.

That is why every reserve in the table shows the same 5.00%. None of them were close. USDC at 63.08% was the nearest, and it still had 17 points of headroom.

The supply side is the borrow side minus the protocol's cut. The docs say the protocol retains 10% of borrowed interest as a buffer for future liquidations, which implies:

supply_APY = borrow_APY * utilization * 0.9

That is a claim you can check rather than take on faith, so we did. On all five reserves it matched to ten decimal places:

The practical reading for a supplier: your yield is not the 5% headline. It is 5% scaled by how much of the pool is actually lent out, minus a tenth. USDC suppliers were earning 2.838% because 63% of the pool was working. USDH suppliers were earning 0.497% because 11% of it was. HYPE and UBTC suppliers were earning nothing at all, because nobody borrows those.

Where the caps bite

Every reserve has a global supply cap and a per user cap. The documented figures:

Set the live figures against those caps and the constraint becomes obvious. USDC supply was at 41.6% of its billion dollar cap, with plenty of room. USDT0 was at 4.0% of its 50M cap.

The collateral side is a different picture. HYPE was at 9,853,944 against a 10M cap: 98.5% full. UBTC was at 1,819.43 against a 2,000 cap: 91.0% full.

So the thing rationing this facility is not demand for loans and it is not the interest rate. It is how much collateral the protocol is currently willing to hold. Both collateral reserves were close to their ceiling at this reading, and the docs are explicit about what happens next: once caps are hit, accounts fall back to non Portfolio Margin behavior, and additional margin has to be posted in the settlement asset regardless. Caps have been raised as the rollout has proceeded, so treat these percentages as a snapshot of the ramp rather than a permanent limit.

Manual borrowing versus the automatic kind

Portfolio Margin already borrowed on your behalf. When you placed a spot or perp order with insufficient balance, the account automatically borrowed against eligible collateral, up to token_balance * borrow_oracle_price * ltv. That behavior is unchanged.

What is new on mainnet is doing it deliberately. Manual borrowing separates the credit decision from the trade: you can draw against your collateral, or supply to a reserve for yield, as its own action rather than as a side effect of hitting a button on the order form.

The difference is who sizes the loan. Under automatic borrowing your debt is a byproduct of the order you just sent, sized by whatever that order happened to need. Under manual borrowing you pick the number.

What happens when it goes wrong

Liquidation on Portfolio Margin is a generalization of cross margin: every cross margin perp position and every spot balance in the account is margined together. Sub accounts are still separate.

The trigger is the portfolio margin ratio, and the account becomes liquidatable above 0.95. The inputs worth knowing:

• liquidation_threshold(token) = 0.5 + 0.5 * LTV(token), so HYPE at 0.65 LTV has a 0.825 threshold and UBTC at 0.5 has 0.75
• min_borrow_offset is a flat 20 USDC added to the maintenance requirement
• borrow_oracle_price is a median of three sources: the Hyperliquid spot USDC price, the perp mark price adjusted by the USDT/USDC oracle, and the perp oracle price adjusted the same way

Three things about the process differ from a perp liquidation and are easy to get wrong.

There is no market phase. Portfolio Margin liquidations are taken over directly by the backstop liquidator at system address 0xbbb...b. The stated reason is that spot books have less consistent liquidity than perp books. Our liquidation explainer covers the perp path, which does have one.

The order is not deterministic. Depending on which oracle price updates first, either your perp positions or your spot borrows may go first. The docs say plainly that users should not expect a fixed sequence.

Takeover can be partial. Between the partial and full liquidation thresholds, collateral and debt are taken over in 20% slices, stopping as soon as you are healthy again. Below the full threshold, everything with positive LTV goes at once. The backstop converts collateral to the debt asset using a TWAP with a 10 minute half life, inside a slippage width of 3M USDC for HYPE and 300k USDC for BTC.

For builders: CoreWriter and the read precompiles

The weekly update's second half was a reminder that HyperEVM contracts can reach this facility. The reach is real but narrower than the phrasing suggests.

Writing. The CoreWriter system contract at 0x3333333333333333333333333333333333333333 exposes action ID 15, borrow lend operation, with fields (encodedOperation, token, wei) typed (uint8, uint64, uint64). The encoded operation accepts 0 for Supply and 1 for Withdraw. A wei of 0 applies the operation maximally, so Withdraw with zero pulls your full reserve balance. Note what is absent: the documented action carries no borrow or repay operation. A contract can be a lender here. It cannot be a borrower.

Reading. Three info endpoints cover state:

The read precompiles starting at 0x...0800 cover the wider HyperCore state that a lending contract needs anyway: order book prices, spot balances, perp positions, oracle prices. Precompile gas is 2000 + 65 * (input_len + output_len), and an invalid input consumes all gas passed into the call frame. The API guide covers authentication and rate limits for the endpoint side.

The documented design intent is worth quoting rather than paraphrasing. The Hyperliquid docs state that "Portfolio margin intentionally does not bring a full-fledged lending market to HyperCore, as that is best built by independent teams on the EVM," and note that HyperCore lending is not tokenized, though an EVM protocol could tokenize it by wrapping the CoreWriter and precompile calls in a yield bearing ERC20.

HyperCore or a HyperEVM protocol?

They are not competing for the same job.

If your capital is already sitting on Hyperliquid backing trades, the HyperCore facility stops it being idle without moving it anywhere. If you want a lending position as the trade itself, with a wider asset list and a token you can use elsewhere in DeFi, the HyperEVM protocols are built for that, and our yield guide compares the routes side by side.

To use any of this you need funds on the exchange in the first place. Deposit through app.hyperliquid.xyz or follow our USDC deposit walkthrough.

Risks worth naming

Supply yield is not guaranteed and can be zero. HYPE and UBTC suppliers earned 0.000% at this reading. Supplying an asset nobody borrows is a decision to hold that asset with extra steps.

The rate can move a long way, fast. The formula is flat to 80% and then near vertical. A reserve that drifts from 78% to 90% utilization goes from 5.00% to 52.5% APY. If you are borrowing, utilization is the variable to watch, not the current rate.

Caps can push you off the facility mid position. When a cap is hit the account falls back to standard behavior and additional margin must come from the settlement asset. That is a change in your margin requirement arriving from outside your account.

Liquidation is unfamiliar. No market phase, no guaranteed ordering between spot and perp legs, and a backstop that unwinds on a TWAP. None of that behaves like the perp liquidation most Hyperliquid traders have modeled.

Every figure on this page came from Hyperliquid's public API or documentation on the date stated, and reserve parameters are set by the protocol and change. Check the live state before sizing anything.