Home › DEX › Hyperliquid › Funding and liquidation

Funding and liquidation on Hyperliquid — when they are charged, what they cost, and how to read the numbers

By TradingPrimer · Published 6 Sep 2026

You already know what a perpetual is. What you may not know is that on Hyperliquid the two costs that decide whether a position survives are charged on different clocks and priced off different numbers: funding lands every hour and is calculated on the oracle price, while liquidation watches the mark price against a maintenance margin the asset sets and you cannot change. This guide runs both through one $10,000 BTC position, with the venue’s own formulas and nothing else.

How does funding work on Hyperliquid, and who pays whom?

One side of every perpetual pays the other, every hour, to keep the contract tethered to spot. If the contract trades above the oracle price, longs pay shorts; if it trades below, shorts pay longs. The venue takes nothing: the Funding page says the payments are “purely peer-to-peer and no fees are collected” on them.

Three details separate this from the version you have read about on a centralised venue, and each of them changes an actual number.

• The clock is hourly. The published formula computes an eight-hour rate; the venue charges one eighth of it every hour. Nothing about the annualised cost changes, but the timing does — there is no way to close five minutes before a funding stamp and skip a payment, which is a real tactic on venues that settle every eight hours.
• The price used is the oracle price, not the mark price. The Funding page states the payment is position size × oracle price × funding rate, and adds in italics that the mark price is not used. Oracle prices are computed by each validator as a liquidity-weighted median of centralised-exchange spot prices. So during a violent move, when the venue’s own book has run away from spot, your funding is still charged against the outside world’s price.
• There is a fixed floor. Like most venues, Hyperliquid bakes in an interest-rate component of 0.01% every eight hours — 0.00125% an hour, which the documentation annualises as 11.6% paid to the short side. That is the baseline a long pays in a calm market before any premium is added.

The premium on top is the part that moves. It is sampled every five seconds and averaged over the hour, and it is measured with an impact price — the average price at which a defined notional amount would actually execute on the book — rather than the best bid and offer. That choice matters: a one-lot quote pinned near the oracle cannot hold the funding rate down if the depth behind it is thin. The full expression is F = average premium index + clamp(interest rate − premium index, −0.0005, 0.0005), and the venue works it through with its own numbers: an impact bid of $10,100 against a $10,000 oracle gives a 1% premium, the clamp contributes −0.05%, and the rate comes out at 0.95%.

Finally, the ceiling: 4% per hour. The page is unusually candid about it — “much less aggressive capping than CEX counterparts” — and the cap does not vary by asset. Read plainly, that sentence says the venue permits a more extreme funding cost than a centralised exchange would. We will price that in a moment.

What does funding actually cost on a $10,000 position?

Start with the floor, because it is the only figure that is fixed. The interest component of 0.01% per eight hours is 0.00125% an hour. On $10,000 of notional that is $0.125 an hour, $3.00 a day, $9.00 over three days — which is exactly what a market-order round trip costs in trading fees — and $90.00 over thirty days. Hold a long for a month in a flat market and funding has quietly cost you ten round trips.

Now the venue’s own worked example, which is where the arithmetic stops being comfortable. The Funding page computes a rate of 0.95% from a 1% premium. Charged at one eighth per hour, that is 0.11875% an hour, or $11.88 an hour on $10,000. A 20x position has $375 of margin above its maintenance requirement; at that rate the funding bill alone consumes it in about 32 hours. The example is the venue’s, not ours; we have only divided it by eight and multiplied it by the position.

And the cap: 4% an hour is $400 an hour on $10,000. That is larger than the entire free margin of a 20x position and four fifths of the margin posted. We are not predicting that rate — in ordinary conditions the live figure is a tiny fraction of it, and when the market is short-heavy it is negative and a long is paid. The point is narrower and it holds in every market: funding is charged against the same equity that price movement is charged against, and the buffer is finite. Any statement of the form “my stop is 4% away, so I am safe” is incomplete unless it also names how long the position will be held and what funding is doing while it is.

One practical note on reading the live rate. On the public BTC screen the header shows a Funding / Countdown pair — on 5 Sep 2026 it read 0.0007% with a timer counting down under 33 minutes — and the account tables include a Funding History tab. The header figure is the coming hour’s estimate and it is re-computed continuously; it is a reading, not a forecast, and it says nothing about the next twelve hours.

When exactly does Hyperliquid liquidate you?

When account equity, including unrealised profit and loss, falls below the maintenance margin. That is the whole trigger. The number that matters is therefore the maintenance margin, and the single most useful fact about it is that you do not set it.

The Margining page defines it as half of the initial margin at the asset’s maximum leverage. BTC allows 40x, so its initial margin at maximum is 2.5% and its maintenance margin is 1.25% — and it stays 1.25% whether you open at 40x, at 5x or at 2x. Assets with lower maximums have higher maintenance rates, up to 16.7% for the ones capped at 3x.

Maintenance rates derived from the maximum-leverage bands on the venue’s Margin tiers page and the “half the initial margin at max leverage” rule on its Liquidations page, both read 6 Sep 2026. The last column is our arithmetic and applies only to a position opened at that maximum. Margin tiers also step the maximum down for very large positions — BTC drops from 40x to 20x above $150M of notional, ETH from 25x to 15x above $100M — thresholds no retail position reaches.

Two consequences follow, and the second one surprises people.

1. Switching to a smaller asset raises the level, not just the risk. A 10x-maximum altcoin carries a 5% maintenance margin. The same $10,000 of notional is closed out four times earlier in percentage terms than the same $10,000 of BTC.
2. Lowering your leverage does not lower the liquidation level; it raises the wall in front of it. On BTC the position is always closed at the point where equity hits 1.25% of notional. Trading at 5x rather than 40x simply means you posted $2,000 instead of $250, so there is more to lose on the way down.

The mechanism also uses the mark price, which the venue builds from external exchange prices combined with the state of its own book, rather than the last trade. That is a protection, not a technicality: a single thin print cannot liquidate you. The documentation is explicit that during high volatility, or on a highly leveraged position, the mark price can sit some way from the book price.

How far can price move before you are liquidated?

Far enough to compute exactly. The published formula is liq_price = price − side × margin_available / position_size / (1 − l × side), where side is +1 for a long, l is one divided by the maintenance leverage — 1.25% on BTC — and margin_available is the isolated margin (or, for a cross position, the account value) minus the maintenance margin required.

Run it across the leverage settings for one long of $10,000 on BTC, entered at $100,000 with isolated margin. The last column prices the same buffer in funding rather than in price: how many hours at the documented 4% cap would consume it on their own.

Our calculation, using the venue’s published liquidation formula and its 1.25% BTC maintenance margin, read 6 Sep 2026. Entry price of $100,000 is an example chosen for clean arithmetic, not a quote. Funding hours assume the 4% hourly cap and no price movement; the live rate is normally far below the cap.

The bottom row is the sanity check that tells you the arithmetic is right. At 1x the formula returns a liquidation price of exactly zero, which is precisely what an unleveraged long should give: you cannot be closed out of a position you have fully paid for until the asset is worthless. Any liquidation calculator that does not return zero on that input is wrong somewhere.

The row above it is the one to sit with. The form on app.hyperliquid.xyz opens at 20x on BTC, in cross margin, and it does not stop you pressing Buy / Long before you have opened the leverage modal. A trader who accepts both defaults is holding a position that a 3.80% move ends — and 3.8% is a number BTC produces on ordinary days, not on famous ones. At the maximum of 40x it is 1.27%, which is inside the range of a single hour.

What happens in the seconds after the trigger?

Not one event but a sequence, and the difference between the second stage and the fourth is the difference between keeping your remaining margin and keeping none of it.

The ordinary case is stage 2. Market orders for the full size of the position go to the book, anyone may fill them, and if enough is closed for the maintenance requirement to be met again, the remaining collateral stays with you. This is what the venue means when it says, in one line on the Liquidations page, “Unlike CEXs there is no clearance fee on liquidations.” On a large centralised venue a liquidated BTC position is charged a clearance fee of 1.25% of position value on top of the loss; here the same percentage is at stake only if the book route fails.

Stage 3 is a detail that only matters above $100,000 of notional: the first liquidation order is for 20% of the position, followed by a 30-second cooldown during which any further liquidation is for the whole thing. It is a mercy for large positions and irrelevant for retail ones.

Stage 4 is the backstop. If equity falls below two thirds of the maintenance margin without the book absorbing the position, the liquidator vault — a strategy inside HLP, the protocol vault ordinary users can deposit into — takes the position over. And stage 5 is the cost: the maintenance margin is not returned. The venue explains why without euphemism: the vault needs a buffer so that backstop liquidations are profitable on average. On our BTC position that is $125, which is 14 times a market-order round trip, gone in one event.

For a cross position the backstop is not confined to the losing trade. The page states that all of the trader’s cross positions and cross margin are transferred to the liquidator, so an account with no isolated positions ends at zero equity. An isolated backstop takes that position and its margin and leaves the rest alone. That is the entire practical argument for opening the margin-mode modal before your first order: cross is the default, and cross makes one bad position the whole account’s problem.

How do funding and liquidation feed each other, and what do you do about it?

Most guides treat these as two separate subjects. They are one subject, because they draw on the same pool of money. Funding is deducted from account equity, and account equity is what the maintenance margin is compared against. Every hour you hold, funding moves your liquidation price closer to you.

At the baseline rate the effect is small and worth knowing anyway. Our 20x position starts with $375 of free margin and a liquidation price of $96,203. After a day of baseline funding it has paid $3.00, so free margin is $372 and the liquidation price has moved up to about $96,233 — roughly $30 nearer, on a position where the whole distance is $3,797. Over a month at the same rate the drift is $911. That is why the venue warns that the liquidation price on screen “still may not be the actual liquidation price due to funding payments”.

At an elevated rate it stops being a footnote. Use the venue’s own example rate again: $11.88 an hour empties the same buffer in about 32 hours with price standing perfectly still. And there is a reason elevated funding and dangerous prices arrive together — funding is high precisely when the contract is trading rich to spot, which is when the book is crowded with leveraged longs. The hour your margin is being drained fastest is the hour a move against you is most likely. They are not independent risks and they should not be budgeted separately.

Four things follow, in order of how much they save.

1. Place a stop-loss above the liquidation price. Not near it — above it, with room for the gap between mark and book price the documentation warns about. This is the one action the venue itself recommends, in the same paragraph where it explains that a backstop keeps your maintenance margin. It removes stage 4, stage 5 and the $125.
2. Open the margin-mode modal before the first order. Cross is the default. Isolated confines a backstop to one position.
3. Set leverage deliberately, and remember it is checked once. The Margining page states plainly that “leverage is only checked upon opening a position”; after that, monitoring is your job. You can even raise the leverage of an open position without closing it. Nothing in the system will stop a position drifting into a state you would not have opened.
4. Budget funding as a holding cost, the way you would a borrowing rate. If the plan needs a week, price a week of funding into the trade before you take it, and check the rate rather than assuming the baseline.

IF YOU CAN NOW ANSWER THE THREE QUESTIONS ABOVE

Everything on this page is arithmetic you can do before you deposit anything. Two things are not arithmetic, and they are the arrangement you are accepting: you hold the keys yourself, so a lost seed phrase is lost money with no support desk to appeal to and no company that can reverse a mistake; and the venue checks your leverage once, when the position opens, and never again — the form starts at 20x cross on BTC, which is a 3.80% move from liquidation. If that is the trade you want to make, open the screen through the link, change the margin mode and the leverage before you type a size, and set the stop in the same ticket as the entry.

Hyperliquid→

Referral link — the venue pays us a share of its fees at no extra cost to you, and under its published referral rules a code gives you 4% off for your first $25M of volume. It does not change a number on this page. Education only; most retail traders lose money.

Where can this cost you?

Where it can cost you: the design is more honest than most, and three parts of it are still working against you. None of the following is hidden — all three are in the documentation — but none of them appears in the summaries people read first.

• The 4% hourly funding cap is looser than a centralised venue’s. The Funding page states this as a design choice, and it is defensible: a market-driven rate is how a perpetual holds its peg without an operator intervening. It also means the worst hour available to you here is worse than the worst hour available on a venue with tighter caps, and it is charged on the oracle price whether or not you could have traded at that price.
• “No liquidation fee” is true and incomplete. The backstop keeps 1.25% of a BTC position — the same percentage a large centralised venue charges as a clearance fee on every liquidation. The trigger differs, and in your favour; the number does not.
• The liquidation price on screen is an estimate the venue tells you not to fully trust. Before entry it depends on book liquidity; after entry it moves with funding, and for cross positions with the unrealised profit and loss of every other cross position. The documentation recommends monitoring with the exact formula instead — which is a fair thing to say and also an admission that the number in the interface, the one a beginner will rely on, is not the number.

Common mistakes

• Treating funding as an eight-hourly event. It is hourly here. There is no stamp to dodge by closing early, and a position held overnight pays eight times, not once.
• Assuming lower leverage lowers the liquidation level. It does not. On BTC the level is 1.25% of notional at every setting; lower leverage buys distance, not a different rule.
• Accepting the form’s defaults. 20x and cross, both pre-selected. That combination turns a 3.80% move into a liquidation and lets a backstop reach every other cross position you hold.
• Reading “no liquidation fee” as “liquidation is cheap”. A book liquidation returns what survives. A backstop keeps $125 on $10,000, and a cross backstop takes the account to zero equity.
• Sizing a stop from the on-screen liquidation price. It is an estimate that moves with funding and with your other cross positions. Compute it, and leave room above it.
• Ignoring funding on a multi-week hold. Thirty days at the documented baseline is $90.00 on $10,000 — ten market-order round trips — and the baseline is the floor, not the average.
• Forgetting that leverage is checked once. Nothing re-checks the position after it opens. If you would not open it at today’s effective leverage, you are still holding it.

Frequently asked questions

Sources: Hyperliquid documentation — Funding, Liquidations, Margining, Margin tiers and the support note “Why was I liquidated?”, all read 6 Sep 2026; the public trading screen at app.hyperliquid.xyz, opened 4–5 Sep 2026 without connecting a wallet. Liquidation prices, funding totals and the hours-to-liquidation column are our own arithmetic from those published formulas. Published 6 Sep 2026.