In the largest flash crash in cryptocurrency history, Binance made a staggering $22 billion, while OKX netted $3 billion

❝ The 90-minute crash left 1.66 million investors with nothing. The money went into the exchange's pockets. ❞

1. What you see in the crash is not the whole truth

In the early hours of October 11, 2025, if you happened to be trading overnight, you would witness the largest liquidation in cryptocurrency history: Bitcoin plummeting from nearly $100,000, with altcoins generally dropping over 40%. You want to close your position, click confirm, it spins, timeout; try again, still timeout. The app freezes, the system crashes.

You think this is a technical accident compounded by market panic, with unlucky investors being trampled, but what you see is not the full picture.

The brutal truth behind it is that while your order channel is cut off, someone else's channel is still open, and they are specifically waiting there. That someone is the exchange itself.

2. The deliberately compressed numbers

After the incident, the crypto data platform CoinGlass released the liquidation data across the network: $19.1 billion, involving about 1.66 million accounts. This number was quoted verbatim by almost all media, becoming the official narrative anchor for the October 11 incident.

Common sense allows us to easily deduce that a liquidation amount of $19.1 billion could not have caused such a heavy blow to the cryptocurrency market; this is illogical. Therefore, there is only one truth:

This number is fake.

It is not a small deviation in statistical standards, but a systematic, magnitude-level fabrication.

▍ Restoring the real data

Hyperliquid is a decentralized exchange on-chain, with data completely open and unalterable. On October 11, its liquidation amount was $10.31 billion, with open interest (OI) of about $13.8 billion, resulting in a liquidation/OI ratio of about 75%—which is considered a normal range in extreme flash crash scenarios (50%–80%).

Using the same logic to estimate centralized exchanges:

Binance's OI before the incident: about $120 billion (the largest in the world)

OKX's OI before the incident: about $43 billion (the third largest in the world)

Assuming an extreme liquidation rate of 70% → Binance's actual liquidation scale: $84 billion; OKX: about $30 billion

However, CoinGlass reported: Binance's official liquidation amount was $2.41 billion, and OKX did not fully disclose.

This means that Binance underreported by a factor of 35.

This is not a typo, not a bug, but design.

CoinGlass later admitted: when exchanges report liquidation data to them, they only push one record per second, rather than the full amount. In extreme scenarios, the actual liquidation trigger frequency can reach hundreds of records per second. By controlling the push speed, exchanges can compress the publicly available data to one-tenth of the real scale within a compliant framework. This mechanism is referred to in the industry as "batch aggregation reporting"—in simple terms: push as much as you want the outside world to know.

ARK Invest founder Cathie Wood believes the actual deleveraging scale on October 11 was $28 billion; K33 Research provided a range of $35 to $40 billion; Wintermute CEO Evgeny Gaevoy publicly stated that the actual number "is at least between $25 and $30 billion." No institution recognizes the $19.1 billion figure.

3. How $2.2 billion and $300 million were made

When Binance announced a total compensation of $680 million for affected accounts, no one would have thought that they actually made a huge profit from this industry disaster. Their revenue composition is as follows:

First: Forced liquidation discount revenue

A head of a major exchange's original contract product provided us with a formula:

Where:

· --- Manipulation profit

· --- Nominal total amount of liquidated orders

· --- Actual acquisition price by the exchange

· --- Market fair price

The latter part of this formula is a very important coefficient, determining when the exchange intervenes to protect the market during extreme situations when a liquidity vacuum occurs, corresponding to the price at which the liquidation engine executes when user positions reach the liquidation line.

Under normal circumstances, a decent trading platform would maintain the discount rate between 0.5% and 2% to avoid the criticized phenomenon of "spike" and to uphold the image that a leading exchange should have, fulfilling the most basic responsibilities of an industry business card. Previously, a well-regarded trading platform set a KPI of 1.375% for its contract team.

However, in the extreme flash crash of October 11, where liquidity instantly dried up and market makers withdrew all orders, the situation was completely different: the exchange allowed the order book depth to plummet by 99%, and a forced liquidation order that should have been executed at $80,000 could actually be executed at $65,000. This difference—discount—was conservatively estimated to be no less than 20%, and even exceeded 30%.

More importantly: in centralized exchanges, this discount revenue belongs 100% to the exchange. No external arbitrageurs can intervene because the entire system has crashed; only the exchange itself can operate.

Binance's forced liquidation discount revenue (conservative estimate): $84 billion × 20% = $16.8 billion

Second: ADL automatic reduction revenue

When the loss from liquidation exceeds the insurance fund's capacity, the exchange activates the ADL mechanism: forcibly closing profitable positions in the market to fill the liquidation gap with the money from the winners. This part of the loss never appears in any public data, has no name, no record, and is quietly deducted from profitable accounts.

Third: Delayed execution privilege

During the system crash, users cannot place orders or add margin. However, the liquidation engine does not have to execute the liquidation immediately—it can "defer" it. Prices continue to drop, and position losses continue to expand, then being liquidated at a lower position. For users, this is no different from a normal liquidation; the difference is that the liquidation position is much lower than the appropriate price. Who pockets the extra price difference?

Calculating this way, with the three types of revenue combined, Binance's final estimated profit is between $18 billion and $22 billion, while OKX's is between $2.5 billion and $3.5 billion.

4. The caste system in the servers

To understand how the money was made, one must also understand who could place orders during the crash.

The systems of large centralized exchanges consist of multiple independent modules: order module, market data module, liquidation engine, risk control system. The so-called "system crash" refers to the order channel facing external users crashing. However, internally, there exists a priority system that has never appeared in any user agreement:

First level: Exchange's proprietary market makers—written in the same segment of code as the exchange's core system, unaffected by external crashes, and the only real sellers in the market during the crash.

Second level: Leading third-party market makers (like Wintermute, Jump Trading)—holding exclusive low-latency API agreements, able to partially operate even when ordinary users are down.

Third level: Ordinary quantitative teams—gradually losing functionality after system congestion, but recovering slightly earlier than retail investors.

Fourth level: Ordinary retail investors—always at the very end of the line.

❝ The exchange's servers have strict service priority. When everyone is down, only proprietary market makers can place orders; they are the referees and also the fastest runners in the game. — Former core management of Huobi, Xiao Daxia ❞

5. The Wintermute incident: The gentry's money returned in full

During the crash, Wintermute was also involved.

One of the largest crypto market makers globally, managing billions of dollars in assets, also suffered significant losses during October 11 due to the lack of transparency in the liquidation mechanism. Afterward, CEO Evgeny Gaevoy publicly criticized on social media: stating that Binance's liquidation data was "seriously underreported, with the actual liquidation scale being at least ten times the official data"; criticizing the priority mechanism during the crash as "extremely unfair to third-party market makers." The post was shared thousands of times, causing an uproar in the industry.

This was a rare public counterattack from the most unlikely group—those top market makers who had long profited from this system, finally spoke out because they too were cut.

About ten days later, he deleted the post. All related statements were retracted. The official statement claimed "the matter has been resolved through private channels." There has been no further public criticism since.

In the cryptocurrency industry, no one was surprised by this outcome. Wintermute got its money back—possibly even more than what was lost; Binance silenced the most credible opposing voice. A private compensation led to permanent silence, which was beneficial for both parties.

6. Jane Street: A carefully designed smokescreen

Months after October 11, a mysterious name was inexplicably brought up: Jane Street. Accompanying this was an astonishing figure: Jane Street's profits in the crypto space reached $130 billion in 2025.

This largest high-frequency market maker globally was accused in a widely circulated tweet: systematically suppressing Bitcoin prices and manipulating the market through ETF derivative structures, being the behind-the-scenes force of the crash. The narrative linked various events, including the Terra crash lawsuit, penalties in the Indian market, and the regular market crashes at the opening of U.S. stocks, creating a seemingly complete logical chain with strong emotional appeal.

The underlying fact is that Jane Street is one of the core authorized participants (AP) for Bitcoin ETFs. When large redemptions occur in ETFs like IBIT, Jane Street must sell in the spot market to hedge, causing concentrated selling pressure at the opening—this is normal market-making behavior dictated by the mechanism. Coinbase analysts tracked data over the last 60 days, finding that IBIT's opening trend was highly consistent with QQQ, and the so-called "crash" was essentially a market beta-level linkage. Moreover, manipulating Bitcoin prices makes no commercial sense for Jane Street: market makers earn from the bid-ask spread and volatility premium; directional bets only bring unnecessary one-sided risks, and the reputational loss is the most unbearable cost.

So why did this narrative gain traction? Because it diverted attention. When the market spreads the idea that "Jane Street is manipulating Bitcoin," the focus shifts away from the truth: during the period when users were unable to save themselves due to the crash, the exchanges exploited system priorities and liquidation mechanism loopholes to absorb hundreds of billions in liquidation losses.

After all, there are always those who are serious: where did the assets they lost actually go?

Jane Street is a perfect target: big enough, mysterious enough, and Wall Street enough to naturally resonate with the emotions of crypto fundamentalists. Redirecting anger towards Jane Street is the most valuable public relations gift for Binance and OKX.

7. The transformation of KPIs and the birth of the slaughterhouse

Xiao Daxia has been in cryptocurrency for over a decade, formerly part of Huobi's core management, witnessing the entire evolution of this industry from its rough beginnings to today. He attributes the root of everything to: KPIs.

In Huobi's early days, the core KPIs were only two things: no liquidations, no spikes. No liquidations was the bottom line—better to let the insurance fund bear losses than to allow user positions to be maliciously liquidated; no spikes was the integrity—price data must be real, and artificial long wicks should not be used to trigger liquidations en masse. In that era, brand value was more important than profit, and there were still idealists in the industry who believed in Satoshi Nakamoto's vision.

As the industry continued to grow, the influence of leading exchanges approached that of traditional financial giants, but the integrity and bottom line of this industry fell in inverse proportion.

When Binance's average daily trading volume exceeded $100 billion, when derivative contracts became the platform's largest source of income, and when the two exchanges held hundreds of billions in open interest, the KPI was reduced to one: maximizing liquidation revenue. Every extreme fluctuation was a massive liquidation opportunity, and every user who was liquidated was a calculable income. Protecting user positions became a proactive abandonment of revenue.

❝ The way a slaughterhouse is born is not a sudden decision to kill, but rather designing slaughter as part of the system's normal operation. ❞

8. Lawsuit

In December 2025, Xiao Daxia began systematically collecting information from victims. As of the time of this interview, the registered victims had reached 383, with total losses exceeding $303 million—this is just the portion that actively contacted for rights protection, compared to the 1.66 million affected accounts, it is but a drop in the ocean.

He hired two law firms, one in Hong Kong and one in the United States, to explore the feasibility of filing a class action lawsuit against Binance and OKX. The road is not easy: both exchanges are registered in the Cayman Islands and Seychelles, and their user agreements contain lengthy arbitration clauses, meaning any jurisdictional disputes could drag on for years. Even former industry executives face such challenges, let alone ordinary retail investors without resources, lawyers, or a voice.

But Xiao Daxia said this battle must be fought.

Data Explanation

The data cited in this article comes from ARK Invest's public reports, K33 Research reports, Wintermute CEO's public statements (now deleted), CoinGlass, CryptoQuant, and Sina Finance roundtable interviews. Binance and OKX have not publicly responded to the data and mechanisms involved in this article. All financial estimates are based on public data extrapolations and are not audited results.