BitMEX invented the perpetual swap. On September 23, it turns itself off.

BitMEX closes September 23. BitMart closes January 31. Three exchanges stopped trading in twenty-five days — two are returning funds on a published schedule, and one froze withdrawals and may not. We cover the exact deadlines, the three separate forces closing exchanges in 2026, what the history actually says about closures marking a bottom, and what to do with your funds before the cutoff.

BitMEX invented the perpetual swap. It is the single most-traded product type in crypto, copied by every exchange that came after, and on September 23, 2026, the company that created it will switch itself off.⁰¹ Three days after BitMEX said so, BitMart announced its own wind-down.⁰² Both are solvent and staged: a month's notice, positions closed in sequence, funds returned.

AscendEX is the other kind. It stopped trading on July 1 and did not publish a notice until July 6, by which point automated withdrawals were frozen and every request had gone to manual review. It told users it was "not in a position to give assurances about timing or amounts," warned that some may not recover full balances, and said plainly that a formal insolvency process might follow. On-chain investigators found its reserves close to empty.³⁷

Three exchanges stopped trading in twenty-five days. Two are handing the money back on a published schedule and one may not be able to. That contrast is the whole story, and it has a date on it: two withdrawal deadlines are now running against real people's money.

§ 01 What exactly did BitMEX and BitMart announce?

Both are voluntary wind-downs with staged deadlines. Neither is a bankruptcy, a hack, or a regulator pulling a license.

BitMEX announced on July 23, 2026 that HDR Global Trading Limited — its Seychelles-registered owner — would close the exchange following "a strategic review of the business." New registrations stopped that day. On August 26 at 04:00 UTC the platform goes reduce-only: you can close positions, not open them. On September 23 at 04:00 UTC anything still open is force-closed and trading ends. The site stays up afterward for viewing balances and withdrawing.⁰¹

There is a detail in that notice worth reading twice. Balances left after the closure are charged "an account fee of USD50 equivalent or 1% per annum (whichever is greater), charged monthly" — and BitMEX says that fee could rise with notice. The floor is what bites. One percent sounds mild; a flat $50 does not, on anything small. Below roughly $5,000 the floor dominates, so from September 23 a $600 account pays 8.3% a year to sit still. Leaving funds parked is not a neutral choice.

BitMart's notice landed on July 26, 2026 at roughly 01:30 UTC. Registrations, deposits and new orders halted immediately; futures went reduce-only. All trading ends August 26, 2026 at 01:00 UTC, with withdrawal requests recommended by 05:00 UTC that day. The platform closes permanently on January 31, 2027 at 15:59 UTC. Its stated reason: "After a careful evaluation of the Company's operating conditions, market environment, and future strategic direction, BitMart has made the difficult decision to commence an orderly wind-down of its trading platform operations."⁰²

Two asymmetries between the notices are worth flagging, because they are the kind of thing that matters later. BitMEX asserted that "all assets exceed liabilities as stated on our Proof of Reserves and Liabilities." BitMart's notice contains no equivalent solvency statement, and BitMart has never published an audited proof of reserves. And where BitMEX's founders and executives spoke publicly, no named BitMart executive is quoted anywhere on the wind-down — every outlet attributes the statement to "BitMart" the entity. BMX, BitMart's exchange token, fell by roughly half on the news; outlets logged somewhere between 46% and 63% depending on the snapshot they took.¹³

Both notices carry the same warning, and it is the correct one: there is no expedited withdrawal channel, and anyone offering you priority processing for a fee is running a scam.

§ 02 Why are crypto exchanges shutting down in 2026?

Because three unrelated forces are squeezing at once, and the popular framing of "the industry is dying" collapses them into one story they do not belong in.

Regulatory cost. MiCA's non-extendable July 1 deadline culled the EU provider population by roughly four in five and pushed Binance out of the bloc entirely — I covered that in full when the deadline hit, and the numbers have not improved since.¹⁶ What matters here is the mechanism: compliance became a large fixed cost, and fixed costs kill the venues with the least volume to spread them across. This force is what closed AscendEX, which held no MiCA authorization and had a liquidity deal fall through at the same moment.³⁷ It is not what closed BitMEX or BitMart. Neither was an EU business.

Sanctions and enforcement. A separate thread with no MiCA connection: Garantex seized by US and European authorities in March 2025, HTX designated by the UK in May 2026 and hit with an EU transaction ban effective August 23, 2026, KuCoin permanently barred from the US by CFTC consent order in March 2026 after a $297 million DOJ settlement,¹⁹ EXMO winding down after a UK sanctions listing. KuCoin's EU arm shows how fine the line is — it was licensed in November 2025 and barred from onboarding three months later.¹⁸ These are actions against specific conduct. They are not evidence that the exchange model is broken.

Ordinary business failure. This is the BitMEX and BitMart bucket, and it is the honest one. Simon Dedic of Moonrock Capital put the mechanism plainly: the mid-tier exchange model "has a 'fatal flaw' and needs a constant flow of new users, and when that dries up, the business fails."¹⁴

BitMEX is the clearest case in crypto's history of a first mover losing everything it built. At its peak it did around $8 billion a day, cleared over $1 trillion in 2019, and held roughly 57% of the global crypto derivatives market.¹² By the end, Kaiko put it near $400,000 a day and under 0.01% of the market. That is not a decline. That is an evaporation. It carried $200 million in fines across two actions — a $100 million CFTC and FinCEN settlement in August 2021,⁰⁴ and a further $100 million criminal fine in January 2025⁰³ — tried and failed to sell itself through Broadhaven Capital Partners at around a $1 billion valuation from February 2025, and lost its CEO, CFO and chief growth officer in the same week of June 2026.³⁸ Then it turned off the product it invented, having watched perpetual-swap volume migrate to venues like Hyperliquid that never asked anyone for a passport.

One caution on BitMEX: a proposed class action filed the same day as the announcement alleges an internal desk with privileged access to customer liquidation data. Those are unproven allegations, BitMEX had not responded publicly at the time of writing, and no source establishes any link between the suit and the decision to close.¹⁵ Do not read it as the reason.

§ 03 Is crypto at the bottom?

Honest answer: nobody knows, the data genuinely does not resolve it, and the exchange closures are close to useless as evidence either way.

Start with where the market actually is. Bitcoin is around $64,729, against an all-time high of $126,198.07 set on October 6, 2025 — a drawdown of 48.7%, running 293 days.⁰⁶ Ether is near $1,857 against a $4,946 high, down 62.5%. Total market cap sits near $2.27 trillion against roughly $4.27 trillion at the peak. The Fear & Greed Index reads 26 — "Fear" — having been at 13 a month ago.⁰⁷ Futures open interest is down about 10.6% month over month and funding is close to neutral. Spot Bitcoin ETFs just closed a barely-positive week at +$33.8 million after an eight-week stretch that drained roughly $8 billion.⁰⁸

That is a market that has bled out leverage and sentiment. It is equally consistent with a base forming and with a market that has not finished falling.

Now the part that matters, because it is the claim being made loudest right now. Does an exchange shutting down tell you the bottom is in? Look at what actually happened each time:

• Mt. Gox went offline February 24, 2014. The cycle bottom came January 14, 2015 — 324 days later, nearly eleven months. Bitcoin fell roughly another 60% from the crash level after the "signal" fired.²¹
• Celsius filed Chapter 11 on July 13, 2022 with Bitcoin near $20,000. That looked like capitulation. Price fell another 22% and did not bottom until 131 days later.²⁵
• FTX filed November 11, 2022. The bottom came November 21 — 10 days later. The signal was nearly perfect.²⁴
• QuadrigaCX ran the other way entirely. The December 15, 2018 bottom happened before the collapse became public on January 14, 2019. Waiting for the exchange to visibly fail would have made you miss the low.⁰⁵

Ten days to nearly eleven months, and one case where the failure arrived after the bottom had already passed. The timing relationship is real but wildly inconsistent — which is another way of saying it is not tradeable.

Four data points is not a base rate, and the honest reading of them is exactly that. The people whose job is to make this call are not agreeing either. Samson Mow said in late June that "the bottom is in."²⁸ Markus Thielen of 10x Research revised his target down on July 1, to $46,000–$47,000, with a base case low in Q4 and a stated plan to buy below $50,000: "There's no real buyer in the market right now. That's why we're still in this liquidation period from the ETFs."³⁹ Arthur Hayes — BitMEX's own co-founder — went from "if I had $1 to invest right now, would I be putting it into Bitcoin? No" in March to "we most likely bottomed at 60,000" in May.³⁰ Glassnode's framing is the most defensible: "The conditions for a bottoming process are in place across onchain, offchain and derivatives data, but confirmation has not arrived."⁰⁹

And there is a specific reason this closure wave is weaker evidence than 2022's, not stronger. More projects have wound down in 2026 than in 2022 by raw count — roughly 95 by one tally. But 2022's closures were fraud-driven insolvencies that forced cascading liquidations. These are mostly businesses choosing to stop.²⁰ More closures with less systemic shock is a different signal, and conflating the two is a modeling error.

§ 04 What do the closures actually tell you?

An exchange shutting down is an unreliable signal about price and a highly reliable signal about counterparty risk. The first one you cannot act on. The second one you can act on today.

BitMEX and BitMart are the good version. AscendEX is the reminder that the good version is a choice the operator makes, not a property of the model. And when it goes the other way, the cost is measured in years:

• Mt. Gox collapsed in February 2014. The first repayments went out in July 2024 — ten years later. The distribution deadline has been extended three times and now runs to October 31, 2026.²²
• QuadrigaCX collapsed in January 2019. Creditors were approved for roughly 13% of their claim, valued at April 2019 prices, paid out in 2023.²³
• Celsius froze withdrawals in July 2022. Cumulative recovery reached about 65% by the third distribution in August 2025.²⁷
• FTX is the case people cite as the happy ending, and it needs its asterisk. Recoveries of 103–120% sound like creditors did well — but claims are valued at the November 11, 2022 petition-date price, when Bitcoin was near $16,871. Anyone who would simply have held their coins came out far behind in real terms, and the first distributions did not arrive until 2025.²⁶

The percentage varies. The time does not. Every one of these took years, and in the worst case more than a decade, before people got access to their own money. Nobody in self-custody has ever waited on a bankruptcy court to unlock their wallet — and that does not depend on whether Bitcoin bottoms this quarter or next year.

§ 05 How do I withdraw before the deadlines?

If you hold funds on either exchange, the actions are boring and time-bound. Do them now rather than near the cutoff, when queues and review backlogs are worst.

1. Withdraw first, decide later. Get assets out to a wallet you control, then take your time choosing what to do with them. Do not let a deadline force an investment decision.
2. Start identity checks immediately. Both platforms subject wind-down withdrawals to enhanced review — KYC, device and address checks, source-of-funds, sanctions and travel-rule screening. Submitting a request is not the same as completing a review. This is where people run out of runway.
3. Close positions early. BitMEX goes reduce-only on August 26 and force-closes anything open on September 23, explicitly disclaiming responsibility for losses caused by its own forced wind-down. Choose your own exit price.
4. Do not leave a dormant balance. From September 23, BitMEX charges the greater of $50 or 1% a year, billed monthly. On small balances the flat floor does the damage.
5. Expect the scams. Both notices say it directly: no expedited withdrawal channel exists, and no legitimate party charges a fee to "unfreeze" your funds. Never share a seed phrase or 2FA code, and never trust a support number found through a search ad.
6. Handle low-liquidity tokens first. BitMart listed well over a thousand assets. Thin tokens may have no withdrawal route to a chain you can use, and no other venue that lists them — sort those before the majors, because they take longest to solve.

We keep a running tracker of every live exchange withdrawal deadline, which now includes both of these.

One inference the two notices support and neither states. If the mid-tier model needs constant new-user growth and that growth is gone, the constraint is not specific to BitMEX or BitMart — it applies to every venue in that tier, and the next notice will name one this article does not. You cannot know which. You can know that "would I be fine if this exchange posted a notice tomorrow" has a cheap answer and an expensive one.

§ 06 Should you leave centralized exchanges?

Disclosure before the argument: I founded LeoDex, which competes structurally with every exchange named above. Read what follows as a position, not an audit.

The structural answer to "my exchange might close" is not "find a better exchange." It is to stop having a balance that someone else can put into withdrawal-only mode. But I am not going to pretend that trade is free.

What you genuinely give up. There is no fiat on-ramp in self-custody — converting dollars to crypto still touches a regulated intermediary somewhere, so "leave centralized exchanges entirely" is not an achievable end state for most people. There is no password reset; losing a seed phrase is permanent and there is nobody to appeal to. On-chain swaps pay gas and slippage that an order-book match does not. MEV is a real, quiet tax: research on Ethereum has put sandwich attacks at roughly 1% of DEX trades, and the cost lands on ordinary users. Bridges have lost more than $2.5 billion to exploits, frequently with no entity obliged to make anyone whole.³¹ And tax reporting is harder, not easier: the 1099-DA regime reports gross proceeds without cost basis, and most DeFi activity generates no form at all, pushing the entire reconciliation burden onto you.³²

What you get. No account to close, no jurisdiction to be excluded from, no withdrawal-only mode, and no KYC dossier sitting on a server waiting to leak — Coinbase had 69,461 customers' data exposed via insider bribery in 2025;³³ Ledger leaked around a million email addresses in 2020.³⁴ That exposure is permanent in a way a balance is not. It is also worth knowing why exchanges restrict funds in the first place, because the mechanics are the same whether the trigger is a wind-down or a risk flag.

Where the evidence sits: the ratio of DEX to CEX spot volume has roughly tripled since 2021,³⁶ running near 17% in January 2026 against an all-time high of 21.75% set in June 2025 — real structural share migration, though it is a maturing trend rather than one currently accelerating to fresh highs.¹⁰ Bitcoin held on centralized exchanges was down to about 2.7 million BTC as of March 2026, the lowest since November 2018.¹¹ People are already voting.

That is the lane LeoDex works in: non-custodial crosschain swaps across 60+ chains, routing across THORChain, Chainflip, MAYAChain, NEAR Intents and others, comparing multiple routes per swap, including native Bitcoin — with no account, no email and no KYC, and settlement on-chain into your own wallet.³⁵ This matters specifically for someone leaving an exchange, because a CEX quietly unified assets across a dozen chains under one login, and the moment that login disappears you need something that can still route Bitcoin and Solana and an L2 token in one place. LeoDex is one of several crosschain aggregators that can. I built it, I think it routes well, and you should compare quotes rather than take my word for it.

I will be straight about the limits. LeoDex does not solve the fiat on-ramp, cannot recover your keys, and does not make swaps tax-free — crypto-to-crypto trades are taxable in nearly every jurisdiction and no-KYC has never meant no-tax. And since it is the subject of this article: LeoDex is a business, and businesses post notices. Ours could too. The difference is what happens next — a front-end going dark costs you a bookmark, because the assets were never on it. That is the entire argument, and it applies to me.

Two exchanges just posted that notice. A third could not even promise the money back. The bottom may or may not be in — I genuinely do not know, and the four cases above are why I do not believe anyone who says they do. But the custody question was never a market call, and it is the one you can answer this week.

Survive > maximise. Withdraw before the deadline, not on it. Keep your keys.