Do Perpetual Futures Expire? Not Until the Exchange Closes

A perpetual futures contract has no expiry date. Your position still has an end date, and it was never printed on the contract.

On 23 July 2026, BitMEX notified its users that the exchange closes permanently on 23 September at 04:00 UTC, ending an eleven-year run. Four weeks before that, on 26 August, every account switches to reduce-only: no opening, no adding, closing only. Between those two dates, according to CoinDesk, operators "will systematically force close all remaining open contracts to ensure the market shuts down in an orderly manner." The venue that carried the crypto perpetual to a mass audience has handed the instrument the one thing it was built never to have: a date.

Nothing in the contract was changed to make that possible. No expiry clause was activated, because there is no expiry clause to activate. A company finished a strategic review, and every open position on its book inherited the outcome.

This walkthrough follows Lilith, Kodex's twenty-year cybersecurity hand, who reads any system by asking who holds the keys and what breaks when the infrastructure does. Across the desk is Nina, twenty-three, eleven months into her first brokerage account, in the habit of fact-checking the marketing before she believes it.

Nina opens with the sentence every exchange academy prints, and she is not wrong about it.

"Perpetual means it never expires," she says. "That is the definition. That is the entire product. So how did one of them end up with a date?"

Lilith does not correct her. "You have read it correctly," she says. "Now tell me what that sentence is a promise about."

Perpetual futures have no settlement date. Your holding period is a different question.

The mechanics are exactly as advertised. A perpetual is a futures contract with no delivery date. Per the reference description of the instrument, it can be held indefinitely, with no rolling into a new contract as expiry approaches. Economist Robert Shiller proposed the structure in 1992 as a way to build derivatives markets for illiquid assets. The inverse version appeared on a small exchange in 2011. BitMEX carried it to scale in 2016, which is why the instrument and the venue have been spoken of together ever since.

What replaces the expiry date is the funding rate. It is a periodic payment between the long side and the short side, and it pulls the contract price back toward spot whenever the two drift apart. That single mechanism is why the contract can run forever without settling, and it is what the label actually protects.

So Nina's reading holds. There is no clause that ends the contract, and nothing in the product is lying to her.

Lilith wants the boundary of the claim, though, not the claim itself. A dated future ends because the paperwork says so on a specific day, agreed in advance by both sides, enforceable by a third party. A perpetual has no such line. That absence is a fact about the terms of the contract. It is not a fact about how long you get to hold one.

The contract has no clock. The company running the book does.

A perpetual has no clearinghouse, so your position has nowhere to go

"Then move it," Nina says. "If BitMEX is closing, take the position somewhere else. That is what you would do with a house or a phone number. Why is a position different?"

This is the question the whole story turns on, and the answer sits one layer below anything printed on a trading screen.

In regulated dated futures, a clearing house stands in the middle. When two firms agree a trade, it is submitted for clearing. The original bilateral agreement is then legally replaced through novation: a trade between firm A and firm B becomes two trades, A to the clearing house and the clearing house to B. Neither side is exposed to the other any more. Both are exposed to the clearing house, which collects margin from both and maintains a guarantee fund. If a member fails, its positions and collateral are liquidated in an orderly procedure, and the fund absorbs what the collateral does not cover. The structure exists because counterparties have failed before and someone had to still be standing afterward.

A crypto perpetual has none of that. Your position is a book entry inside one company's ledger, collateralized by margin you posted to that company, backstopped by that company's insurance fund. When the fund runs short, the venue reaches for auto-deleveraging, which closes profitable positions on the winning side to cover the shortfall the losing side could not. Every one of those parts is internal. There is no outside party holding your claim, which is precisely why no outside party can carry it anywhere.

Auto-deleveraging is worth sitting with, because it is the wind-down in miniature. A losing position blows through its margin faster than the market can absorb it. The insurance fund cannot cover the gap. So the venue reaches across to the profitable side of the same contract and closes those positions instead, ranked by how much profit and size they carry. You did nothing wrong. Your trade was working. The book needed your position more than you did, and the terms you accepted when you opened it settle that argument in the book's favour. It is a solvency backstop for violent days, which is how it is meant to work. It is also a standing demonstration that a position here can be ended by another account's failure, on a timetable set entirely inside one building.

Nina keeps circling the comparison, so Lilith lays the three cases side by side.

The middle column is the strange one, not the outer two. It is the only case where nothing sets an end date, and it holds that property only for as long as the third column stays hypothetical. BitMEX just moved from the middle column to the right one, and what moved it was a board decision.

Positions do not migrate. They close.

What reduce-only does to a book that can only shrink

It is 04:00 UTC on 26 August, the reduce-only switch lands, and a position you have held for months can now only be made smaller. It is easy to read as a courtesy: a month of warning, plenty of time to get out.

"So I get four weeks," Nina says. "That sounds fair, honestly. I have closed a position in four minutes."

Lilith lets the number sit before she takes it apart. "You get four weeks in which the only permitted action is the one action every other account is also restricted to. Ask what that does to the price you close at."

That is the part the calendar hides. Reduce-only does not simply limit what you can do. It removes the market's ability to form the other side of what you are doing. A book works because someone is willing to open a position opposite yours. Ban opening, and the only available counterparty to your exit is another holder unwinding in the opposite direction. Market makers, whose business is warehousing risk they expect to lay off later, have no reason to quote size into a venue with a published closing date.

A book where no account may open is not a market. It is a queue.

What leaves first is the quoting. A market maker earns by holding inventory briefly and offsetting it, and offsetting requires a venue that will still be there tomorrow. Give that business a closing date and the arithmetic inverts. Whatever inventory it takes on has to be unwound inside a shrinking window, against a counterparty pool that is shrinking at the same time. Spreads widen before volume falls. That widening is not information about Bitcoin, it is information about the building.

And the queue is not something you get to join at your own pace. The force-close runs in stages across those four weeks, with the operators working through the remaining contracts as they go. The choice of when you leave stops being yours somewhere in the middle of that window, at a moment that is not published in advance.

Nina, who reads terms carefully because she has been caught by them before, notices the shape of it. The announcement reads like an extension. The mechanism reads like a countdown.

Why does the price drift when open interest drains?

The countdown reaches the funding rate too. Funding was the answer to how a contract with no end date stays tied to reality, and it is worth being precise about who pays it, because the precision is the whole point here.

Funding is not a subsidy from the exchange. It is a transfer between the long side and the short side of that specific book. When the contract trades above spot, longs pay shorts, and the payment makes holding the expensive side costly enough that the gap tends to close. The mechanism works because both sides exist and both sides care about the cost.

Drain one side and the machinery loses its grip. The rate can move to an extreme and still fail to pull the mark back, because a payment only disciplines a position that someone is still willing to hold. On a book heading toward closure, that is exactly the condition. The tether between the contract and spot weakens in the same weeks that leaving becomes compulsory.

"So the number I have been checking to judge the cost of holding stops telling me anything?"

"It tells you what the last two accounts agreed to pay each other," Lilith says. "On a full book, that is information. On a draining one, it is an echo."

What survives as a signal is open interest, the count of contracts still outstanding. A rate moving while open interest falls is describing a thinner and thinner agreement. A rate moving while open interest holds is describing a real disagreement about price. Reading the second number is what keeps the first one honest.

When perpetual futures settle in coin, your exit moves with the price

Both numbers still leave out what you are actually handed when the position closes, and that depends on one more layer sitting on the BitMEX contract.

XBTUSD is an inverse contract. It is quoted in dollars, but margined and settled in Bitcoin. Your collateral is the same asset you are taking a view on, which makes your profit and loss nonlinear in price rather than a straight multiple of it. That axis is covered in full under settlement currency. It deserves naming here for one reason: when the force-close lands, what arrives in your account is BTC, valued at whatever the book prints that day.

This is the part Nina expected to be the good news. "I would have assumed settling in Bitcoin was the friendlier version. It is the thing I wanted to hold anyway."

"It is," Lilith says. "Right up to the moment you stopped choosing the timing."

The collateral and the exposure move together, so a bad exit price does not just decide the size of the result. It decides the value of what you are handed to walk away with.

What happens to my position if an exchange shuts down?

Strip the specific venue out and the general answer is narrower than either fear or reassurance suggests.

You do not lose the position, and you do not keep it. You close it, on their schedule, at whatever the book offers by then. That is the full shape of it, and it holds whether the venue is winding down entirely or simply retiring one contract. Exchanges delist individual perpetuals regularly while continuing to operate normally. Same mechanism, smaller blast radius, and no farewell coverage to tell you it is happening.

Funds are a separate question from positions, and it is worth keeping them separate. BitMEX stays reachable for withdrawals after the closing date, so balances are not confiscated. They are, however, charged for staying: users who leave assets behind face a maintenance fee of $50 a month, or an annualized levy of 1% on what remains. A position ends by force. A balance ends by attrition.

The sequence generalizes cleanly enough to carry to the next venue. An announcement, then a reduce-only date, then a window in which the operator closes whatever is left, then a withdrawal-only period with a fee attached to ignoring it. Each step is defensible on its own, and each one is disclosed. Together they move every decision that mattered out of the holder's hands, in order, on dates that did not exist when the position was opened.

The on-chain version rhymes rather than repeating. On a decentralized perpetual venue the standing counterparty is often a liquidity vault instead of a company. That changes who absorbs the loss and where the failure shows up, as the Ostium exploit demonstrated in detail: the vault is the house. Different address, same question. Something specific is standing behind your position, and it is worth knowing what.

The question to ask before the funding rate

Nina has been building toward the practical version of this for the whole conversation, and she finally puts it plainly.

"So what am I supposed to check? I cannot read a company's board minutes. I do not know which exchange announces a closing date next year."

"You are not being asked to predict it," Lilith says. "You are being asked to know what stands behind the position, and whether anything does."

That is a smaller question than forecasting a shutdown, and it is answerable before you click. A dated future on a cleared market has a third party contractually obliged to still be there. A perpetual on an exchange has the exchange. Say the only thing standing behind your position is the venue itself. Then the venue's remaining life is your maximum holding period, and every other number on the screen is quoting a cost inside a window whose edge is not shown.

Nina had it right at the start, which is the part worth keeping. The contract does not expire. She was simply reading a promise about the paperwork as though it were a promise about her.

BitMEX has a date on a calendar now. Every other venue has one too. Theirs has not been announced yet.

You cannot rehearse a wind-down in a paper account, but you can rehearse the part that stings: closing at a moment you did not pick. Take a position in the $5,000 simulator account, set an exit time before you enter, and close there regardless of what the chart is doing. The distance between that price and the one you wanted is what a forced exit costs.

Start free →