Volver
Risk8 min de lectura

The fund is a multiple, not a balance

A futures insurance fund balance is not the loss-absorption budget. Binance defines each fund's Maximum Takeover Capacity as a confidential multiple of its balance, calibrated to a 99.9% confidence interval that the venue explicitly says nothing about how often ADL happens. Coin-margined contracts share one fund per collateral asset, which makes them smaller and more exposed.

Aún no está en tu idioma — mostrando inglés.

Every major perpetual venue publishes a number in the interface labelled insurance fund. It looks like the answer to a simple question: how much can this thing absorb before it starts taking money from people who were right?

It is not the answer. The number on that screen is a balance. The number that governs you is a multiple of that balance, and the multiplier is not published anywhere.

Capacity is a multiple, not a balance

Binance is the only venue in this set that documents a capacity concept at all, and the definition is unusually direct:

"Each Futures Insurance Fund has a Maximum Takeover Capacity, defined as a multiple of its current balance. This capacity represents the maximum notional value of bankrupt positions that the fund can absorb at any given time. The multiplier for each insurance fund is pre-determined and confidential due to risk management considerations."

The example that follows is explicitly an illustration, not a parameter:

"Example: If the fund balance is $1 million and its pre-set multiple is 3x, its Maximum Takeover Capacity would be $3 million. This mechanism ensures the fund can handle a cluster of liquidations without being immediately exhausted, preserving its ability to protect the market from subsequent events."

Three things fall out of that wording, and all three are load-bearing.

The multiplier is confidential and per-fund. The 3x is a teaching example. No venue publishes its actual multiplier. Anyone quoting you a real takeover multiple is guessing.

Capacity is measured in notional, not in loss. The sentence says "the maximum notional value of bankrupt positions," while the fund's actual payout covers only the gap between bankruptcy price and execution price. Read together, that means two positions of the same notional size consume the same capacity whether they clear at a small shortfall or a large one. That is an inference from the documented wording rather than a sentence Binance writes down, but it is the reading the two sentences support, and it is the opposite of how most traders intuitively size a buffer.

A rising balance does not mean rising capacity in a durable sense. Capacity is a live multiple of a live balance. If the balance is deployed downward, capacity falls with it, immediately and without announcement.

Two separate limits, and they are not the same instrument

Binance's ADL page states the balance-based condition one way:

"When the unrealized loss of Bankrupt Positions taken over by the liquidation engine is larger than the available balance of the relevant Futures Insurance Fund that is allocated to that type of contract, then the liquidation engine can no longer trade that contract type and the Auto-Deleveraging (ADL) mechanism will be engaged."

The insurance fund page states the capacity-based condition separately:

"If the bankrupt position has reached the limit of Futures Insurance Fund maximum take over capacity, then Binance will be unable to take over such Bankrupt Positions."

One limit is denominated in losses against an allocated balance. The other is denominated in notional against a confidential multiple. Both exist. Neither is a substitute for the other, and a fund can pass one and fail the other.

What the 99.9% figure does and does not mean

Binance also publishes a sizing method, which is rarer than it should be:

"The required minimum size of each Futures Insurance Fund is calibrated to cover the losses arising from Bankrupt Positions (for the category of futures covered by the Futures Insurance Fund). The approach used to determine the required minimum size is similar to methods adopted in traditional finance futures markets (and expected by regulators of such futures markets). This approach is based on a 99.9% confidence interval and historical stressed scenarios reflecting extreme but plausible market conditions."

And then it removes the most natural misreading, in a sentence worth more than the one above it:

"Note that the 99.9% confidence interval is not related to the occurrence of ADL and does not mean that ADL will be, or is expected to be, avoided in 99.9% of liquidation scenarios."

Read that carefully. It is a sizing claim about the required minimum relative to a modelled loss distribution. It is not an availability claim about how often you will be left alone. In 999 of 1000 modelled scenarios, the calibrated minimum is expected to cover the losses. The tail scenario where it does not is precisely the scenario the tail exists for, and that is the scenario where capacity is exceeded and the mechanism engages.

The venue is telling you the fund is not designed to make deleveraging impossible. It is designed to size how often deleveraging is needed.

The calibration is reviewed on a fixed cadence: "The calibration of the required minimum size of each Futures Insurance Fund will be reviewed by Binance on a quarterly basis and at times when Binance considers it appropriate in its sole discretion."

OKX, dYdX and Hyperliquid publish no confidence level, no loss distribution, and no calibration interval. OKX's own framing of its trigger is looser, describing the mechanism as "currently defined as a 30% straight-line decline from its peak within 8 hours but this threshold may be adjusted based on market conditions." Hyperliquid publishes no capacity concept at all in either the liquidations or the vaults documentation.

Where the money comes from

Four documented Binance channels, and two of them run in opposite directions.

The fee channel is real but the routing is discretionary. "When a position is liquidated, a portion of the assets that have been made available to maintain the position will be deducted and paid to Binance as a Liquidation Clearance Fee, unless the position is a Bankrupt Position following liquidation." Note the direction: that fee is paid to the exchange. Whether any of it reaches a fund is a separate decision: "All or part of the Liquidation Clearance Fees paid by traders with respect to non-bankrupt positions subject to liquidation may be allocated to maintain the Futures Insurance Funds, as Binance considers appropriate." The word doing the work is "may."

The second channel closes the loop: any surplus arising when the fund takes over a bankrupt position is credited back to that same fund. The third is the exchange itself: "The size of each Futures Insurance Fund will be monitored frequently. If the size of any Futures Insurance Fund is less than the required minimum, Binance will contribute additional assets to the Futures Insurance Fund."

The fourth channel is the one that caps it:

"If the size of the Futures Insurance Funds exceeds the required minimum determined in a periodic review, any funds in excess of the required minimum may be deployed by Binance for other purposes as it considers appropriate in its sole discretion."
"To the extent that Binance considers necessary or appropriate, assets may be rebalanced between the various Futures Insurance Funds to ensure that each Futures Insurance Fund is at least equal to the required minimum."

A stated floor and a discretionary ceiling, in the same document. That combination has a direct consequence for anyone reading the balance as a strength signal: a high balance is partly a signal that excess may be withdrawn.

OKX runs the same structure with the discretion placed on size rather than routing. Its fund "is self-sustaining and is funded primarily by collecting profits generated during the liquidation process," and the mechanism is a surplus rather than a fee: "If the position is closed out at a price better than the bankruptcy price, a surplus is generated."

OKX adds that "OKX contributes its own profits from time to time to maintain an industry-appropriate level of funding for the Security Fund," settles forced-reduction outcomes on a fixed clock ("Every day at 16:00 (HKT), the platform settles profits and losses from forced reductions and liquidations from the past 24 hours and updates the risk reserve accordingly"), and keeps reserves segmented: "Risk reserves are kept separate by business line (margin, futures, perpetual swaps, and options), and also by futures and crypto within each line." Its reservation is explicit: "OKX retains absolute and sole discretion over the establishment, maintenance, size, composition, and utilization (or non-utilization) of the Security Fund."

dYdX has the cleanest funding rule of the four and the vaguest refill. Its governance documentation sets "The maximum liquidation fee, in parts-per-million. 100% of this fee goes to the Insurance Fund," and the liquidations page states that "up to the entire remaining value of the account may be taken as penalty and transferred to an insurance fund," with a default Maximum Liquidation Penalty of 1.5% that "will be subject to adjustments by the applicable Governance Community." Seeding was a one-time community act: "the applicable Governance Community needs to initially fund the insurance fund from the applicable community treasury." There is no documented automatic top-up, so replenishment after a draw is a governance decision.

Hyperliquid names no insurance fund at all. The backstop is the liquidator vault, "a component strategy of HLP," and two structural things differ from the exchange model. There is no fee skim: "Unlike CEXs there is no clearance fee on liquidations." And the vault takes a position rather than covering a gap: "When a cross position is backstop liquidated, the trader's cross positions and cross margin are all transferred to the liquidator." The buffer comes out of the liquidated account: "During backstop liquidation, the maintenance margin is not returned to the user. This is because the liquidator vault requires a buffer to make sure backstop liquidations are profitable on average." The vault is also the exit for the flow itself: "On Hyperliquid, the pnl stream from liquidations go entirely to the community through HLP."

Pooling is why coin-margined is more exposed

Capacity is a multiple of a balance. So the size of the pool that balance is shared across becomes a direct input to how much the fund can absorb.

Binance names the rule on both sides of it. On the composition:

"Binance also offers coin-margined contracts. All coin-margined contracts which use the same cryptocurrency asset as collateral will share one single Futures Insurance Fund. For example, BTC coin-margined perpetuals and delivery contracts will share the same Futures Insurance Fund."

And the consequence, stated plainly on the ADL page:

"Note that coin-margined contracts are more likely to be subject to ADL than USD-margined contracts. This is because all coin-margined contracts which use the same cryptocurrency asset as collateral will share one Futures Insurance Fund, resulting in smaller Futures Insurance Funds."

The chain is two documented links. Sharing produces smaller funds. A smaller fund means a smaller capacity multiple. Binance states the first half and the multiplier definition; nobody publishes the second half as a number.

The same dilution applies to USD-margined pools. Binance publishes the USDⓈ-M composition: BTC, ETH and BNB USDT-margined contracts share one fund; DOT, LINK, XMR, ADA, BCH, EOS, ETC, LTC, TRX, XLM and XRP USDT-margined contracts share another; USDC-margined contracts share a fund; and other USDT-margined contracts sit in other pools. Membership is not permanent, by the venue's own statement: "Binance may allocate futures contracts to different Futures Insurance Fund accounts and establish new Futures Insurance Fund accounts from time-to-time, as it considers it appropriate in its sole discretion."

That list is a snapshot of an allocation decision, not a structural guarantee.

The numbers, with the dates they were read

Every figure below is a read from a public page or a public endpoint on the date shown. Balances tick continuously, so treat each as a point sample rather than a level.

| Venue | Fund or pool | Balance read | Read at | |---|---|---|---| | Binance USDⓈ-M | BTC, ETH, BNB USDT-margined (dated row) | 1,284,002,434.0309 USD, plus 67.6540 BTC and 0.0000 BNB | 2026-10-04 04:00:00 UTC | | Binance USDⓈ-M | Same pool, live header | about 1,289,773,xxx USD | 2026-10-04 approx. 18:2x UTC | | Binance USDⓈ-M | DOT, LINK, LTC, ADA, XRP, TRX USDT-margined (dated row) | 64,050,787.1832 USD | 2026-10-04 04:00:00 UTC | | Binance USDⓈ-M | USDC-margined shared fund, BTCUSDC and ETHUSDC | 91,927,132.60 and 91,927,132.58 | 2026-10-04 approx. 18:2x UTC | | OKX | BTC-USDT swap | 89,831,586.6053 USDT | 2026-10-04 18:29-18:33 UTC | | OKX | ETH-USDT swap | 59,845,504.9261 USDT | 2026-10-04 18:29-18:33 UTC | | OKX | DOGE-USDT swap | 53,550,391.0993 USDT | 2026-10-04 18:29-18:33 UTC | | OKX | XRP-USDT swap | 27,555,115.9652 USDT | 2026-10-04 18:29-18:33 UTC | | OKX | SOL-USDT swap | 16,350,117.7309 USDT | 2026-10-04 18:29-18:33 UTC | | OKX | BTC-USD swap | 349.5837 BTC | 2026-10-04 18:29-18:33 UTC | | dYdX | Cross-margin insurance fund address, on-chain | 8,075,470.762990 USDC | 2026-10-04 18:26:42Z and 18:33:49Z | | dYdX | Same fund, stated balance in governance proposal | approximately $17M USDC before a 10M USDC transfer | proposal submitted 2026-03-30, voting ended 2026-04-02 | | Hyperliquid | HLP liquidator vault | not obtained | see note below |

The dYdX address dydx1c7ptc87hkd54e3r7zjy92q29xkq7t79w64slrq is public, and the proposal states it "holds the cross-margin insurance fund exclusively. It backstops the 109 cross-margin markets currently listed on dYdX Chain." The same proposal is the only place in this set where a fund is deliberately reduced: it sought 10M USDC for DAO funding and reported that against roughly $60M of open interest, "the fund represents a 28.3% coverage ratio (11.7% after the $10M withdrawal)." One caveat the proposal itself flags: its open-interest denominator is total OI, so "the effective coverage ratio for cross-margin positions is higher than presented." Isolated-market open interest sits in the denominator without drawing on this fund.

No Hyperliquid HLP balance appears in this article. vaultDetails returned null or a deserialization error on repeated attempts from the host used for this research, while other calls to the same endpoint succeeded, and the HLP vault address itself was not confirmed from Hyperliquid documentation. A missing figure is not a small figure.

What a backstop is not

The disclaimers are consistent across all four venues and they are the part most worth reading.

Binance: "Notwithstanding the use of the term 'insurance', the Futures Insurance Funds are not insurance products, and Binance is not an insurer." On scope: "The Futures Insurance Funds are not used to cover traders' losses. Its only purpose is to cover any difference between the bankruptcy price and the execution price of the liquidation order. Traders will not get any of their losses back from a Futures Insurance Fund."

OKX: "The Security Fund is not an insurance policy, a custodial protection fund, or a guarantee against user losses. OKX is not an insurer." And on ownership: "The Security Fund is the sole property of OKX. Users have no beneficial interest, ownership rights, legal claim, or ability to direct the use of any assets held within the Security Fund."

A backstop covers the engine's shortfall between bankruptcy price and execution price. It is not a claim on your losses, and on these venues it is not a claim on the fund either.

The part that is actually actionable

Check three things, in this order.

Find out which pool your contract belongs to, and whether that pool is shared. If it is shared with other underlyings or with delivery contracts, the fund backing your position is smaller than a per-contract reading suggests, and its capacity is a smaller multiple of a smaller balance.

Read the balance as a snapshot with a date, not as a solvency statement. Binance publishes the history page explicitly, and OKX exposes per-underlying balances through its public interface. Neither figure carries a multiplier, and neither tells you whether the current period's balance is above or below the required minimum, because that minimum is confidential.

Assume the fund can run out. On Binance the excess above the required minimum may be deployed elsewhere; on OKX the size is at the exchange's sole discretion; on dYdX a draw is followed by governance, not an automatic refill; on Hyperliquid the vault is a trading strategy that has to average out ahead in order to keep absorbing. In every case the ladder above the fund still ends somewhere.

Sources

  • Binance: Introduction to Futures Insurance Funds: the Maximum Takeover Capacity definition and example, the 99.9% calibration method and its caveat, quarterly review cadence, all four funding channels, the pool composition and coin-margined sharing rule, and the ADL-versus-capacity conditions. Re-fetched and quoted verbatim 2026-10-05.
  • Binance: What Is Auto-Deleveraging (ADL) and How Does It Work?: the coin-margined ADL-proneness statement and its shared-fund cause. Re-fetched and quoted verbatim 2026-10-05.
  • Binance Futures: Insurance Fund History: the USDⓈ-M fund balances and dated rows in the table, read 2026-10-04. Binance's public API endpoint was geo-blocked from the host used here, so these are rendered-page reads with the selected contract confirmed before each value was recorded.
  • OKX: Understanding OKX's Security Fund: funding by liquidation surplus, OKX contributions, absolute discretion over size and utilization, and the not-an-insurance disclaimers. Re-fetched and quoted verbatim 2026-10-05.
  • OKX: forced liquidation FAQ: daily 16:00 HKT settlement, risk-reserve segmentation by business line, and the 30%-in-8-hours ADL definition. Re-fetched and quoted verbatim 2026-10-05.
  • OKX ADL introduction: the trigger conditions, including "drops significantly lower than a certain preset threshold," where the preset number is not disclosed.
  • OKX public API: GET /api/v5/public/insurance-fund: per-underlying fund balances in the table, read 2026-10-04.
  • dYdX: Liquidations: the penalty-to-insurance-fund rule and the default 1.5% Maximum Liquidation Penalty. Re-fetched and quoted verbatim 2026-10-05.
  • dYdX: Contract Loss Mechanisms: "Isolated markets are markets that have segregated pools of collateral and their own insurance fund."
  • dYdX: Governance Functionalities: the governance-set levers, including Max Insurance Fund quantums, the parts-per-million liquidation fee routed 100% to the fund, and per-position and per-subaccount per-block caps. Re-fetched and quoted verbatim 2026-10-05.
  • dYdX DRC: Utilise Insurance Fund for DAO Funding: the stated balance, accrual rate, coverage ratios, market count, per-subaccount per-block payout cap, and the deleveraging fallback. Re-fetched and quoted verbatim 2026-10-05.
  • dYdX chain: cross-margin insurance fund bank balance: the on-chain figure in the table, read twice on 2026-10-04 and re-read 2026-10-05.
  • Hyperliquid: Liquidations: the 2/3 maintenance-margin backstop trigger, the no-clearance-fee statement, the maintenance-margin buffer, the cross and isolated transfer rules, and the liquidation flow routing to the community through HLP. Re-fetched and quoted verbatim 2026-10-05.
  • Hyperliquid: Protocol vaults: the HLP description. Re-fetched 2026-10-05. This page states what HLP is and does not publish a vault balance or a capacity parameter.

Risk note: cryptocurrency trading, leveraged perpetual futures, and automated algorithmic strategies carry significant risk of rapid and total financial loss. Never risk funds you cannot afford to lose completely. Insurance funds and backstop vaults are not insurance products, do not cover trader losses, and are disclosed by every venue cited here as the exchange's own property with no assured size, multiplier, or payout. A backstop that is exhausted transfers the loss to winning counterparties through auto-deleveraging. Nothing in this article is investment advice, a recommendation, or an offer to sell any product.