A perpetual futures contract never expires, so it needs something other than delivery or rollover to keep its price close to the underlying asset. That something is the funding rate: a periodic payment between long and short position holders, sized to the gap between the contract's price and a reference spot price. Margin used, liquidation price and funding paid only mean what a risk dashboard says they mean if the mechanics behind them are understood.
The instrument
BitMEX introduced perpetual futures in 2016, and they became its most popular derivative product. A perpetual gives leveraged exposure to an underlying price without rollover or direct ownership. As the academic literature puts it, unlike fixed-maturity futures, perpetuals are not guaranteed to converge to the spot price. Venues therefore pay funding between longs and shorts at set intervals, calculated from how far the contract has drifted from an index built from spot markets.
The funding formula, as venues document it
Binance documents its funding rate as the average premium index plus a clamped interest-rate term: F equals P plus clamp(interest rate minus P, -0.05%, 0.05%), scaled to the funding interval. The default interval is eight hours, settling at 00:00, 08:00 and 16:00 UTC, with a default interest rate of 0.01% per interval. The premium index compares impact bid and ask prices, the average fill prices for a set notional on each side of the book, with a price index drawn from spot exchanges. The funding amount is the nominal value of the position, calculated at mark price, multiplied by the funding rate.
Hyperliquid, which describes itself as a layer one blockchain with fully onchain perpetual futures order books, documents the same premium-plus-clamp formula. It pays funding every hour at one eighth of the computed rate and caps funding at 4% per hour. Its payment is position size multiplied by the oracle price multiplied by the funding rate: it uses the spot oracle price, not the mark price, to convert size into notional value.
Mark price and index price are not the last traded price
Two reference prices sit behind a perpetual, and neither is simply the last trade. On Binance, the price index is a weighted average of the asset's spot price on several exchanges, each carrying a set weight. The mark price for a perpetual is the median of three values: the index adjusted for funding, the index adjusted for a 30-second moving average of the bid-ask spread, and the contract's own price. Binance uses mark price for unrealised profit and loss, which is what triggers liquidation, and says it does so to reduce unnecessary liquidations in volatile markets.
Hyperliquid's oracle price, by its own documentation, is computed by each validator as a weighted median of centralised-exchange spot prices, weighted by each exchange's liquidity. A "liquidation price" on one venue is therefore not directly comparable with one on another until you know which reference price each uses.
Initial and maintenance margin
Initial margin is the collateral required to open a position; Binance states that it is determined by the leverage selected. Maintenance margin is the minimum a position must keep. Binance's formula is notional position value multiplied by a maintenance margin rate, minus a maintenance amount, calculated by tier regardless of leverage, and "the larger the position, the higher the Maintenance Margin rate." The venue states that failing to meet the maintenance margin requirement may result in liquidation. Higher leverage means less initial margin for the same notional, and so a thinner buffer above the maintenance floor.
What happens once maintenance margin is breached
Deribit documents an incremental liquidation system with no preceding margin call. The engine first closes part of a position, trading in rounds of one per second where liquidity allows, and stops once the maintenance margin requirement falls back below the account's margin balance; positions not yet liquidated stay open. While liquidation runs, the account holder cannot place or cancel orders. Liquidation trades carry an extra fee, which goes to the insurance fund.
Insurance funds and auto-deleveraging: who absorbs the last loss
Liquidation does not always finish before an account goes below zero. Deribit's insurance fund covers the negative equity of bankrupt traders so that winning traders are still paid, and it is funded by those liquidation fees, added hourly. If the fund were ever fully depleted, further bankruptcies would be socialised among that day's winning traders; Deribit states that this has not happened to date.
Deribit places auto-deleveraging earlier in that chain, not after the fund. Its sequence runs from risk-reducing-only trading at 100% initial margin use, to liquidation at 100% maintenance margin use, to a transfer of positions to pre-approved professional support providers under its Liquidity Support Program at 100% of a lower close-out margin threshold. Only if that capacity is insufficient does ADL move the remaining positions to members holding opposing positions, pro rata to their unrealised profit on the instrument, at a mark price adjusted slightly in their favour. Any shortfall left after that falls to the insurance fund. Deribit's documentation also excludes perpetuals with an options series in the same pair, including its BTC and ETH perpetuals, from ADL. The underlying question, which pool of capital absorbs a shortfall first, also governs bad debt in onchain lending, covered in where yield in lending vaults comes from.
Centralised order books and onchain alternatives
On a centralised venue the exchange runs the matching engine, calculates mark price and margin internally and holds the collateral. Onchain designs differ from that in different ways. Hyperliquid runs its order books onchain, yet its oracle is still built from centralised-exchange spot prices, so it relies on their price formation even though orders, trades and liquidations happen on its own chain. GMX takes another route: it routes every order against its GM and GLV liquidity pools rather than matching traders with each other, uses Chainlink Data Streams oracles for pricing, and shares fees from trading, liquidations and borrowing with liquidity providers. None of these designs removes leverage risk; they change where prices are formed and who stands on the other side. How leverage and thin liquidity behaved under stress is covered separately.
Where regulators stand
Access varies by jurisdiction, and the legal status of each measure matters as much as its substance. This is a description of the rules, not legal or investment advice.
In the United States, on 29 May 2026 the CFTC issued an order permitting a designated contract market to list a perpetual contract referencing the spot price of bitcoin as a futures contract, together with a policy statement on listing perpetual contracts. The statement says the case-by-case review in CFTC Regulation 40.3 is appropriate for perpetual contracts referencing asset classes the order does not cover. The order is in effect; the policy statement sets out how the agency intends to review further listings and is not itself a rule.
In the United Kingdom, the FCA's prohibition on firms selling, marketing or distributing derivatives, including futures, that reference certain cryptoassets to retail consumers has applied since 6 January 2021. When it proposed lifting the parallel ban on crypto exchange traded notes in June 2025, the FCA said its ban on retail access to cryptoasset derivatives would remain in place, and the derivatives prohibition remains in force.
In the European Union, ESMA said on 24 February 2026 that derivatives marketed as perpetual futures or perpetual contracts are likely to fall within the existing national product intervention measures on contracts for difference. Where they do, they must meet those measures: leverage limits, risk warnings, margin close-out, negative balance protection and a ban on monetary and non-monetary benefits. The statement reminds firms of obligations that already exist, and asks them to assess what a new product actually is rather than what it is called. Regulators have faced a similar classification question with prediction markets.
What remains unresolved
The CFTC's May 2026 action covers one bitcoin-referenced product and a review route for others. It does not say which other perpetual contracts will be approved, or when. On pricing, a paper by Songrun He, Asaf Manela, Omri Ross and Victor von Wachter, last revised in August 2024, finds that deviations from no-arbitrage perpetual prices are larger in crypto than in traditional currency markets and diminish over time. A 2023 BIS working paper on crypto carry reports average carry above 10% a year, at times up to 60%, and finds that high carry predicts future price crashes. On that evidence, funding is paid for bearing a risk that tends to arrive in a crash.