What caused the 10 October 2025 crypto crash?

1 What happened on 10 October 2025?

A trade shock reached a crypto derivatives market crowded with long leverage late on a Friday, and the selling ran hardest in the hour after 21:00 UTC. China’s Ministry of Commerce announced export controls on rare-earth-related items on 9 October 2025. The next day President Trump posted twice on Truth Social: at 14:57 UTC about those controls, and at 20:50 UTC to announce a further tariff on Chinese goods from 1 November 2025, after US equities had closed lower. Table 1 gives the evening in UTC, a source in each row.

The terminal’s Chart Desk shows the evening in prices, on a desk of three charts built for this article and set to 1 September to 15 November 2025 (Figure 2). On Coinbase, Bitcoin opened 10 October 2025 at $121,714.51, four days after the window’s high of $126,296.00, and fell to $107,000.00 in the hour from 21:00 UTC, a drawdown of 12.1% from the open; Ether fell 19.7% to $3,510.00 in the same hour. Hyperliquid’s perpetual printed $100,837 in its bar from 20:00 UTC. Exchange prints read on 28 October 2026; past results do not indicate future results.

Note: A Chart Desk desk built for this article, 1 September to 15 November 2025: Coinbase BTC-USD and ETH-USD by UTC day and Hyperliquid’s BTC perpetual in four-hour bars, stacked. To rebuild it with a free account: choose the Chart Desk’s three-chart stacked layout, set the three symbols at the intervals named here, and set the window to 1 September to 15 November 2025. Exchange data as published, read on 28 October 2026; past results do not indicate future results.Source: BlockPhi terminal, Chart Desk > COINBASE:BTC-USD / HYPERLIQUID:BTC / COINBASE:ETH-USD, captured 29 October 2026. Charts by Vela (LuxAlgo); data: Coinbase and Hyperliquid.

Free with an account·Open the Chart Desk (opens in a new tab)

The bars around 10 October 2025 on this Chart Desk desk, US dollars, times in UTC; exchange data as published, read on 28 October 2026; past results do not indicate future results.

Chart	Bar or week	Open	High	Low	Close or level
Coinbase BTC-USD, daily	9 October 2025, UTC day	123,343.24	123,822.08	119,713.65	121,714.51
Coinbase BTC-USD, daily	10 October 2025, UTC day	121,714.51	122,600.00	107,000.00	112,980.28
Coinbase BTC-USD, daily	11 October 2025, UTC day	112,976.01	113,439.99	109,683.96	110,768.89
Coinbase ETH-USD, daily	10 October 2025, UTC day	4,370.72	4,397.03	3,510.00	3,836.15
Hyperliquid BTC perpetual, 4 hours	10 October 2025, 16:00 UTC	118,985	119,429	115,804	116,632
Hyperliquid BTC perpetual, 4 hours	10 October 2025, 20:00 UTC	116,626	117,129	100,837	112,761
Hyperliquid BTC perpetual, 4 hours	11 October 2025, 00:00 UTC	112,729	113,302	110,949	112,293

2 How large was the liquidation, and why is the reported total a floor?

By CoinGlass’s count it was the largest single day of forced closures in its records: $19.1 billion across more than 1.6 million traders (The Block, 13 October 2025). CoinGlass’s 2025 annual report, on data to 20 November 2025, puts long positions at roughly 85% to 90% of the day’s total. Third-party counts as published, read on 28 October 2026; past results do not indicate future results.

The count is a floor by construction. Since 27 April 2021, according to the change log of Binance’s futures API, its public liquidation streams, which aggregators read, have pushed at most one liquidation order per contract each second. When many accounts on one contract are closed inside a second, the stream shows one, and CoinGlass said at the time that the true total was likely much higher for that reason (The Block, 13 October 2025). BlockPhi’s inference is narrow: totals compiled from that stream can count only part of Binance’s liquidations. We publish no estimate of the remainder, because no reproducible method for one exists in public.

The terminal’s Positioning page shows the regulated side of the market, where positions changed little across the week (Figure 3). In the CFTC’s weekly reports for CME’s Bitcoin future, leveraged funds were net short 14,047 contracts of 5 BTC on 7 October 2025 and 13,431 a week later, asset managers net long 8,600 and 8,043, with open interest at 29,001 and 28,503 contracts. Weekly snapshots cannot show the evening itself, and much of the funds’ short is the cash-and-carry trade, as the card’s note explains. CFTC data as published, read on 28 October 2026; past results do not indicate future results.

Note: Net positions in CME’s Bitcoin future, in the terminal’s Positioning page, weekly from September to December 2025: leveraged funds net short 14,047 contracts of 5 BTC on 7 October 2025 and 13,431 on 14 October, asset managers net long 8,600 and 8,043. CFTC data as published, read on 28 October 2026; past results do not indicate future results.Source: BlockPhi terminal, Positioning > Bitcoin futures (CME), captured 28 October 2026. Data: CFTC Traders in Financial Futures.

Free with an account·Open this chart in the terminal (opens in a new tab)

Net positions of leveraged funds and asset managers in CME’s Bitcoin future, and its open interest, contracts of 5 BTC, at each weekly report from 2 September to 30 December 2025 (positions as of Tuesday; Monday 10 November), as the terminal computes them from the CFTC’s Traders in Financial Futures report; CFTC data as published, read on 28 October 2026; past results do not indicate future results.

Positions as of	Leveraged funds, net	Asset managers, net	Open interest
2 September 2025	−15,322	8,199	26,684
9 September 2025	−14,751	7,846	26,930
16 September 2025	−16,369	8,267	28,844
23 September 2025	−14,393	7,885	27,809
30 September 2025	−13,873	8,125	26,874
7 October 2025	−14,047	8,600	29,001
14 October 2025	−13,431	8,043	28,503
21 October 2025	−14,002	7,842	28,477
28 October 2025	−17,223	7,945	28,717
4 November 2025	−15,168	6,968	27,093
10 November 2025	−15,810	7,426	27,544
18 November 2025	−13,564	6,248	27,116
25 November 2025	−13,764	6,500	26,260
2 December 2025	−14,126	6,886	24,329
9 December 2025	−14,870	7,352	24,793
16 December 2025	−12,543	6,825	24,502
23 December 2025	−11,251	6,248	21,988
30 December 2025	−10,911	6,597	20,166

3 Why did collateral prices break on one exchange?

Binance priced three collateral assets at least partly off its own spot markets, and on Binance alone they traded far below their reference values for about forty minutes. Binance compensated holders of USDe, BNSOL and WBETH collateral affected between 21:36 and 22:16 UTC on 10 October 2025, when USDe traded below $0.66 and WBETH as low as $430, about 88% under the ether price (The Block and CoinDesk, 12 October 2025). Ethena Labs’ chief executive, Guy Young, said the move was confined to one venue and was not accurately called a USDe depeg. Reported prices; past results do not indicate future results.

The record on that pricing is dated. On 6 October 2025 Binance announced that its margin price index for BNSOL and WBETH would stop drawing on Binance’s own BNSOL and WBETH spot prices and use the staking conversion ratio, to reduce the risk of a depeg; on 11 October it brought the change forward to 14:00 UTC that day. USDe was not in that announcement; afterwards Binance said it would add redemption prices to these assets’ reference prices and set a soft floor under USDe’s (The Block, 12 October 2025).

Sequence matters for a causal reading. Binance’s statement, reproduced by FX News Group on 13 October 2025, places the lowest prices between 21:20 and 21:21 UTC and the severe depegs after 21:36 UTC. On those timestamps the repricing followed the low: it deepened losses for accounts that had posted those assets as margin rather than starting the fall. Dated statements; past results do not indicate future results.

4 What is auto-deleveraging, and why did it close profitable positions?

Auto-deleveraging is a liquidation’s last step: when no buffer can absorb a bankrupt position, the venue closes part of opposing traders’ profitable positions to cover the loss. The buffers are the order book, then an insurance fund or backstop; Table 2 sets out the three cases from Binance’s and Hyperliquid’s own pages.

Tarun Chitra’s paper on the mechanism (arXiv, version of 16 February 2026) studies Hyperliquid, whose account-level data is public. It finds that auto-deleveraging closed $2.1 billion of positions there within 12 minutes on 10 October 2025 and, against benchmark allocations, that the venue’s queue overshot the smallest haircut on profitable traders’ gains that would have covered the shortfall, by $45.0 million to $51.7 million. It states that its comparison “suggests that Binance overutilized ADL far more than Hyperliquid”, while noting that Binance’s account-level data is not public. The paper’s estimates, read on 28 October 2026; past results do not indicate future results.

5 Did funding conditions tighten before the crash?

No: the funding capacity the terminal measures was flat and US money markets were calm into 10 October 2025; the strain its US pages show came in the weeks after. BlockPhi reads the evening as a leverage event inside a liquidity expansion that had slowed, not the start of a funding squeeze. Each reading below is a terminal page as a reader sees it.

5.1 Global Liquidity

The Global Liquidity Model is a composite that measures the balance-sheet capacity behind global credit: the balance sheets of five central banks plus 17 government debt markets, revalued every Friday in US dollars and multiplied by a collateral multiplier keyed to the US ten-year Treasury yield.

In the terminal’s Global Liquidity page, the growth chart at the head of this article (Figure 1) shows the model growing at 6.46% annualised over the three months to 10 October 2025 and 7.80% over twelve. That pace sat close to the median week since April 2020: an expansion neither accelerating nor reversing. Model estimate on reconstructed history, read on 28 October 2026; past results do not indicate future results.

Underneath, the collateral base barely moved. The same page’s History panel of weekly changes (Figure 4) shows it down $0.14 trillion over the four weeks to 10 October 2025: bond prices took off $0.30 trillion and currency translation $0.04 trillion, against $0.20 trillion added by balance sheets and debt. Global Liquidity fell 0.06% over those four weeks and rose 2.34% over eleven, mostly from balance sheets and debt; the Global Liquidity page explains each driver. Model estimates on reconstructed history, read on 28 October 2026; past results do not indicate future results.

Note: The Friday-to-Friday change in the collateral base, in the History panel of the terminal’s Global Liquidity page, weekly from 1 August to 17 October 2025: over the four weeks to 10 October 2025 it fell by US$0.14 trillion, mostly on bond prices, and over eleven weeks it rose by US$2.50 trillion. Reconstructed history; model estimate, read on 28 October 2026, and past results do not indicate future results.Source: BlockPhi terminal, Global Liquidity > Weekly change in the collateral base, captured 28 October 2026.

In the terminal for Premium members·Open this chart in the terminal (opens in a new tab)

On the Global Liquidity Model: the collateral base’s weekly change and its split by driver as the terminal’s Drivers view computes it, with the resulting change in Global Liquidity, US$ billions, 1 August to 17 October 2025; reconstructed history, model estimate read on 28 October 2026.

Week	Collateral base change	Balance sheets and debt	Currency translation	Bond prices	Global Liquidity change
1 August 2025	81	−22	253	−150	245
8 August 2025	−214	−25	−538	348	−646
15 August 2025	405	93	416	−103	695
22 August 2025	1,570	1,728	187	−345	2,930
29 August 2025	153	17	56	80	294
5 September 2025	129	−4	54	78	235
12 September 2025	510	18	130	362	647
19 September 2025	−176	−16	−121	−38	−436
26 September 2025	−90	10	102	−203	−126
3 October 2025	−397	−19	−376	−1	−626
10 October 2025	525	223	357	−55	1,074
17 October 2025	−585	−8	−810	233	−1,139

5.2 Bitcoin against Global Liquidity

The terminal marks the date itself. Its GL Macro Regime card (Figure 5), on the Lead & Forecast page, sets how far Bitcoin’s log price sits from its 52-week mean, in standard deviations (a z-score), against the same reading for Global Liquidity 11 weeks earlier, with a band of two standard deviations either side, and flags 10 October 2025. That week Bitcoin’s z-score was 0.93, inside a band from −0.45 to 3.55; it fell below the band in the week to 14 November 2025 and stayed below it to the year’s end. Model estimates on reconstructed history, as the card’s own rows read on 29 October 2026; past results do not indicate future results. Whether Global Liquidity leads Bitcoin at all is tested in Does Global Liquidity lead Bitcoin?

Note: The GL Macro Regime card in the terminal’s Lead & Forecast page, weekly from 1 August to 26 December 2025, with the card’s own mark on 10 October 2025: Bitcoin’s z-score of 0.93 sat inside the band from −0.45 to 3.55 that week and fell below its lower edge in the week to 14 November 2025. Reconstructed history; model estimates, as the card’s own rows read on 29 October 2026, and past results do not indicate future results.Source: BlockPhi terminal, Lead & Forecast > GL Macro Regime, captured 28 October 2026.

Free with an account·Open this chart in the terminal (opens in a new tab)

On the Global Liquidity Model: Bitcoin’s 52-week z-score of its log price against Global Liquidity’s 11 weeks earlier and the band two standard deviations either side, weekly, 1 August to 26 December 2025, as the card’s own rows read on 29 October 2026; reconstructed history, model estimates.

Week	Bitcoin z-score	Global Liquidity z-score, 11 weeks earlier	Lower band	Upper band	Bitcoin inside the band
1 August 2025	1.19	1.63	−0.37	3.63	yes
8 August 2025	1.28	2.17	0.17	4.17	yes
15 August 2025	1.28	2.36	0.36	4.36	yes
22 August 2025	1.21	2.18	0.18	4.18	yes
29 August 2025	0.82	2.09	0.09	4.09	yes
5 September 2025	0.92	2.15	0.15	4.15	yes
12 September 2025	1.16	1.99	−0.01	3.99	yes
19 September 2025	1.11	2.15	0.15	4.15	yes
26 September 2025	0.78	1.96	−0.04	3.96	yes
3 October 2025	1.43	1.67	−0.33	3.67	yes
10 October 2025	0.93	1.55	−0.45	3.55	yes
17 October 2025	0.49	1.54	−0.46	3.54	yes
24 October 2025	0.80	1.35	−0.65	3.35	yes
31 October 2025	0.68	1.43	−0.57	3.43	yes
7 November 2025	0.12	1.90	−0.10	3.90	yes
14 November 2025	−0.72	1.84	−0.16	3.84	no, below
21 November 2025	−1.64	1.79	−0.21	3.79	no, below
28 November 2025	−1.01	1.80	−0.20	3.80	no, below
5 December 2025	−1.13	1.64	−0.36	3.64	no, below
12 December 2025	−1.01	1.55	−0.45	3.55	no, below
19 December 2025	−1.19	1.38	−0.62	3.38	no, below
26 December 2025	−1.25	1.49	−0.51	3.49	no, below

5.3 US dollar funding

The US side was draining, slowly and without stress. The terminal’s US Liquidity page computes net liquidity as the Federal Reserve’s balance sheet less the Treasury’s cash account and the reverse repo facility. On 10 October 2025 it stood at $5.79 trillion, $0.41 trillion lower than 13 weeks earlier, mainly because the Treasury’s cash had risen from $0.28 trillion on 11 July to $0.79 trillion. Bank reserves averaged $2.97 trillion in the week to 1 October 2025, below $3 trillion. BlockPhi calculations on Federal Reserve and Treasury data, read on 28 October 2026; past results do not indicate future results.

The price of that money had not yet moved. On 10 October 2025 SOFR, the overnight rate on loans backed by Treasuries, printed 4.15%, level with the rate the Federal Reserve paid on reserves, and the standing repo facility lent nothing. The pressure came afterwards: SOFR traded 14 and 15 basis points above the rate on reserves on 15 and 16 October, when banks and dealers drew $6.75 billion and $8.35 billion from the facility, and 32 basis points above it on 31 October, when they drew $50.35 billion. Both are on the terminal’s US Liquidity page for Premium members. New York Fed and Federal Reserve data, read on 28 October 2026; past results do not indicate future results.

5.4 Broad financial conditions

Wider US conditions were loose and stayed so. The Chicago Fed’s National Financial Conditions Index, on the terminal’s Growth & Conditions page, read −0.52 in the week to 10 October 2025, looser than its long-run average of zero, and stayed between −0.54 and −0.51 to the year’s end. Federal Reserve Bank of Chicago data, latest vintage, read on 28 October 2026; past results do not indicate future results.

Table 3 sets the readings side by side. On every funding measure the terminal carries, conditions were not tightening into the week; on this evidence the leverage in the market, and the way venues priced collateral and closed positions, explain the size of the evening better than liquidity does. The dollar’s part in Global Liquidity’s moves is examined in Does the dollar lead Bitcoin?, and the forces behind the model in what moves it.

6 What changed after the crash?

The venues changed their rules within days, and the liquidity readings turned later in the quarter. Binance reported on 13 October 2025 that it had paid about $283 million to users liquidated because of the three depegs. Dated statements; past results do not indicate future results.

The terminal shows the slower turn. Global Liquidity’s three-month annualised growth fell from 6.46% on 10 October 2025 to −1.14% on 28 November and stayed below zero through December, while the twelve-month rate held between 7.55% and 9.65%. US banks and dealers kept drawing on the standing repo facility, $74.60 billion on 31 December 2025, and Bitcoin stayed below its band on the GL Macro Regime card. Model estimates on reconstructed history and Federal Reserve data, read on 28 October 2026; past results do not indicate future results.