Skip to content

Research notes

Hyperliquid Funding Rate vs Funding Payment

An eight-hour formula and hourly payments use different periods. Follow a native ETH position through three invented funding events and reconcile its signed USDC entries.

Ivory segmented metering wheel releasing a lime droplet into a small ceramic bowl on a black background.
Illustration for research notes; examples in this article are educational.

Hyperliquid’s funding rate describes a percentage for a defined period; a funding payment is the cash amount booked to an account. Its documented formula uses an eight-hour scale, while payments occur hourly at one eighth of that formula output, subject to the hourly cap. Reconcile a native perpetual’s payment using the applicable hourly rate, position size and oracle price, then check the account funding record.

A decimal can be correct and still answer the wrong question. An eight-hour formula result, an hourly percentage on a screen and a USDC debit belong in different columns. Combining them can produce an eightfold error before position sizing enters the calculation.

This guide uses Hyperliquid’s documentation reviewed on 29 October 2026. The worked ledger covers the native ETH perpetual on the first perp DEX, with USDC collateral. All example rates, prices, fills and payments are hypothetical. The contract specifications describe the native crypto contracts’ USDC margining and generally USDT-denominated oracle prices, with no USDC/USDT conversion applied. A USDC accounting amount consequently does not establish an actual USDT transfer or a stablecoin’s market value.

Convert the formula scale to an hourly rate once

Hyperliquid’s funding methodology gives an eight-hour formula using the hour’s average premium, P, and an interest component, I, of 0.0001. The interest adjustment is clamped between −0.0005 and +0.0005. Clamping means restricting a value to those bounds.

Adjustment = clamp(I − P, −0.0005, +0.0005).

Eight-hour formula result = P + adjustment.

Hourly rate = eight-hour formula result ÷ 8.

The same methodology specifies a 4% hourly funding cap. The three hypothetical rates below remain well within it.

That last division converts this formula output. It does not authorize dividing every number labeled “funding” by eight. Establish whether a display or dataset already gives the hourly rate. A field name alone does not tell you which transformations an exporter applied.

For example, an invented eight-hour result of 0.0008 converts to an hourly decimal rate of 0.0001. Displayed as a percentage, that hourly rate is 0.010%. These are three expressions of related values, with different units and periods.

Preserve the original decimal string, period and source field before formatting. Converting 0.010% to a multiplier requires dividing by 100, giving 0.0001. The API decimal-precision guide explains how to keep those values intact through parsing and arithmetic. Exact decimal arithmetic cannot rescue an incorrectly labeled period.

Separate market rates from account payments

Hyperliquid exposes market context, historical rates and account funding through separate info requests. The perpetuals API reference documents the following distinctions:

RecordRelevant fieldsWhat it helps establish
Current market contextfunding, oraclePx, markPxThe market observation you retrieved
fundingHistorycoin, fundingRate, timeA historical market rate for the identified event
userFundingdelta, hash, timeThe account’s booked funding entry

The native account example places coin, fundingRate, szi and usdc inside delta. It pairs a positive position size and rate with a negative USDC amount. Keep the raw response before translating it into a report that might display expenses as positive.

The distinction also appears in the Python SDK at commit 2fdb18f. Its funding_history method sends the market’s coin and time range with request type fundingHistory; user_funding_history sends an account address and time range with type userFunding. These wrappers retrieve different evidence. They do not reveal the matching engine’s internal ordering or calculate a user’s payment themselves.

A current observation cannot establish an earlier settled rate. A public historical rate cannot establish that your account held a particular quantity. The broader perpetual funding-cost ledger develops that evidence boundary across venues; the reconciliation below applies Hyperliquid’s specific period and price rules.

A checked three-hour Hyperliquid funding ledger

Assume a 1.20-ETH long is already open at 09:00 UTC and remains through a fictional 10:00 payment. A completed reduction of 0.45 ETH at 10:20 leaves 0.75 ETH for the 11:00 and 12:00 payments. Assume that quantity is definitively included for each listed event, with no other fills or boundary uncertainty. The timeline is entirely hypothetical.

First construct three hourly rates from invented average premiums. Values in the middle columns are decimal rates; the final column is a percentage.

Event, UTCAverage premium PEight-hour formula resultHourly percentage
A, 10:000.00070.0002+0.0025%
B, 11:000.00130.0008+0.0100%
C, 12:00−0.0009−0.0004−0.0050%

For B, the unclamped adjustment is 0.0001 − 0.0013 = −0.0012. Clamping gives −0.0005, so the formula result is 0.0013 − 0.0005 = 0.0008. Dividing by eight gives 0.0001 per hour.

For C, the unclamped adjustment is 0.0010, which clamps to +0.0005. Adding that to −0.0009 gives −0.0004; dividing by eight gives −0.00005. The change of sign changes which side pays.

Now calculate cash. Hyperliquid’s payment formula uses the oracle price to convert position size to funding notional. Its price-index documentation distinguishes that oracle from the mark used for margining and unrealized P&L. A convenient mark-price column is therefore not interchangeable with the required input.

For this worksheet, signed quantity is positive for a long and negative for a short. Positive cash means a receipt; negative cash means a payment. Under these conventions:

Account cash change = −signed quantity × oracle price × hourly decimal rate.

EventIncluded long sizeOracle, USDT/ETHModeled USDC cash change
A1.20 ETH2,500−0.075
B0.75 ETH2,520−0.189
C0.75 ETH2,480+0.093
NetThree separate eventsUse each event’s price−0.171

B’s oracle notional is 0.75 × 2,520 = 1,890 USDT. Its payment is 1,890 × 0.0001 = 0.189 USDC under the stated native contract accounting, which applies no USDC/USDT exchange conversion. At C, the negative rate gives the long a 0.093-USDC receipt. Adding signed cash changes gives −0.075 − 0.189 + 0.093 = −0.171 USDC: a net payment of 0.171 USDC.

For comparison, substituting a short quantity of −0.75 ETH at B gives a +0.189-USDC receipt. The positive rate makes the long pay and the short receive under these assumptions. Preserve position direction when computing a signed amount.

These payments omit trading P&L, execution fees, transfers and collateral requirements. The partial close can generate other account changes; those do not belong in this funding subtotal. Likewise, funding is calculated on the specified exposure, rather than by multiplying a collateral allocation by the rate. Changing leverage while keeping quantity, oracle price and rate fixed leaves this payment arithmetic unchanged. The cross versus isolated margin guide explains the separate question of which collateral supports that exposure.

Use the discrepancy to find the mistaken input

Suppose the hypothetical account entry for B contains szi = "0.75", fundingRate = "0.0001" and usdc = "-0.189". Several plausible-looking calculations fail to match it:

Alternative calculation for BModeled debitError exposed
Apply 0.0008 as the hourly rate1.512 USDCEight-hour formula output used directly
Divide the hourly 0.0001 by eight again0.023625 USDCPeriod conversion applied twice
Substitute an invented mark of 2,5500.19125 USDCWrong price input
Retain the old 1.20-ETH quantity0.3024 USDCCompleted reduction omitted

Each difference is a diagnostic clue. It does not prove the actual account followed that mistake. A reconciliation should identify the first unsupported input, preserve both calculated and booked amounts, and investigate the original record.

Do not silently substitute a later position snapshot for the event’s included quantity. If fills near the payment boundary make event ordering uncertain, retain that uncertainty. Hyperliquid’s public funding description is not a guarantee that a local click before a countdown reaches zero establishes exclusion. The funding timestamp worksheet provides a general reconstruction method; its venue-specific timing caveats must stay with the named venue.

If planned exits complicate the size history, consult the Hyperliquid normal TP/SL versus position TP/SL comparison and retrieve the executed fills. The payment review needs evidence of the quantity reduction, not merely the existence of an exit instruction.

Build a funding review that another person can reproduce

Start with the exact market and account. Keep the perp DEX, returned coin name, collateral context and master or subaccount address. Hyperliquid’s info-endpoint documentation explicitly requires the actual account address for account queries and warns that an agent wallet address can produce an empty result. The spot, perp and HIP-3 asset-ID guide helps resolve market identity before matching similarly named assets.

For each event, retain the raw rate and its period, event timestamp, retrieval time, signed size, oracle-price evidence and booked funding row. Compute an expected cash amount with a stated sign convention and compare it with the preserved native amount. Round the presentation after the comparison; keep unexplained differences visible.

Check history completeness too. The info reference limits time-range responses and documents pagination from the last returned timestamp. Because time bounds can include their endpoints, an overlapping row needs recognition when merging pages. Record the ranges covered instead of equating a short response with proof of a complete account history.

If a live userFundings stream feeds the ledger, its first message can contain previously recorded entries. Hyperliquid documents isSnapshot: true for that initial snapshot and false for subsequent updates in its WebSocket subscription reference. Reconcile overlapping history before adding amounts; reconnecting must not turn an old payment into a second expense.

Keep this native-market example scoped. Current funding documentation specifies a different premium formula for HIP-3 markets and refers to deployer funding parameters. Reusing the native premium calculation for every builder-deployed market would require evidence for that market’s configuration.

Close an event review when the period, included exposure, price input and signed account amount agree, or record precisely which evidence is missing. A remaining difference should lead to a specific check: the raw unit, an omitted fill, a mismatched event or an incomplete history. Carry that explanation alongside the funding total so the next review can reproduce it.

Keep the context with the example.

Each worked example is illustrative, with its sources and assumptions in the article. Read our editorial standards for authorship, verification and corrections.