Crypto failures have hit exchanges, tokens and everyday investors. This article pulls together data from three research papers and one summary report to show how often these failures happen and who pays for them.
The figures cover exchange defaults, market value losses, retail trading results and models that try to predict failure. Where a number comes from one study, the article says so.
The sources use different time periods and methods, so the numbers should not be added together. Each section stands on its own.
Crypto Bankruptcy Statistics at a Glance
A study by Niranjan Sapkota looked at 845 crypto exchanges around the world, using data up to February 2023. Of these, 366 had defaulted and 479 were still running. That means 43.3% of the exchanges in the sample had failed.
A default in this study means the exchange stopped operating, faced financial trouble or was declared bankrupt. The data came from Coinmarketcap.com, Cryptowisser.com and Coingecko.com.
The author says that since the Mt. Gox collapse in 2014, roughly 60% of crypto exchanges have faced defaults. Causes include fraud, cyberattacks and government rules.
| Statistic | Figure | Source |
|---|---|---|
| Exchanges in the study | 845 | Sapkota (2025) |
| Defaulted exchanges | 366 (43.3%) | Sapkota (2025) |
| Running exchanges | 479 (56.7%) | Sapkota (2025) |
| Crypto value lost, late 2021 to end of 2022 | Over $1.8 trillion | BIS (2023) |
| Median retail loss, as estimated | $431 of $900 invested | BIS (2023) |
| Bitcoin bankruptcy probability, short term | 17% to 23% | Kozlovskyi et al. (2022) |
| Bitcoin bankruptcy probability, long term | 13% to 16% | Kozlovskyi et al. (2022) |
| Tokens that failed since 2021 | 3.7 million (52.7%) | Summary report |
Failure is much rarer in stock markets. The Sapkota study cites research showing that less than 1% of NYSE-AMEX firms and just over 1% of NASDAQ firms go bankrupt in a given year.
Another check shows how many exchanges are idle. As of April 2023, CoinMarketCap listed 609 exchanges, but only 237 had traded at least $1 in the previous 24 hours.
Coins fail at a high rate too. A 2020 study cited by Sapkota found that about 60% of cryptocurrencies defaulted between 2014 and 2018 because of scams, hacks and business failures.
The Sapkota study also points out that failure rates in crypto are close to an even split. In stock markets, bankrupt firms make up about 1% of the data, which makes them hard to model. In this crypto sample, defaults and survivors are nearly balanced, so no data tricks were needed to study them.
The $1.8 Trillion Market Value Wipeout
The Bank for International Settlements (BIS) tracked the 2022 crash in a bulletin published in February 2023. Prices peaked in November 2021. After that, the prices of many coins fell by around 75% during 2022, including bitcoin and ether.
Over $1.8 trillion of crypto value dissolved as prices dropped. Two events caused much of the damage.
In May 2022, the algorithmic stablecoin TerraUSD fell from $1 to almost zero in a few days. Over $450 billion in crypto and DeFi value was destroyed between May and June.
In November 2022, the exchange FTX collapsed after a run sparked by concerns about its cash and solvency. Another $200 billion was lost, and bitcoin, ether and other coins fell by over 20% in a matter of days.
Together, those two events account for about $650 billion, or roughly 36% of the total loss.
| Event | Date | Value lost | Key detail |
|---|---|---|---|
| Price peak | November 2021 | Not applicable | Bitcoin reached about $69,000 |
| Terra/Luna collapse | May 2022 | Over $450 billion | UST fell from $1 to near zero |
| FTX bankruptcy | November 2022 | About $200 billion | Major coins fell over 20% in days |
| Full 2022 decline | 2022 | Over $1.8 trillion | Many coins fell about 75% |
Trading rose during both crashes. Daily active users on Binance, Coinbase and FTX jumped after news of the troubles broke.
The BIS found that large holders, which it calls whales (wallets with more than 1,000 bitcoin), reduced their holdings in the days after each shock. Small holders with less than one bitcoin increased theirs.
The BIS said this pattern suggests large investors sold to smaller ones before the steep price drops. It also found that the crash had little visible effect on broader financial markets outside crypto.
Major Exchange Failures and Where They Happen
Exchange failures follow a few repeating patterns. Some shut down after legal action, some are hacked and some simply disappear.
The Sapkota study lists several high-profile cases. The table below sets out the main ones.
| Exchange | Year | What happened |
|---|---|---|
| Mt. Gox | 2014 | Collapse that began a long run of exchange defaults |
| BTC-E | 2017 | Forced to stop after U.S. money laundering allegations |
| BTC-China | 2017 | Shut down during China’s crackdown on crypto |
| QuadrigaCX | 2019 | Bankruptcy after the CEO died; $190 million became inaccessible |
| Thodex | 2021 | Froze when the founder left with all funds |
| FTX | 2022 | Collapsed with an $8 billion shortfall on its balance sheet |
Between 2014 and February 2023, business problems, regulatory trouble and scams caused failures at about the same rate. After 2018, most exchanges that failed simply vanished without a stated reason, leaving clients with no answers.
Location matters less than many expect. Singapore had the most defaulted exchanges at 41, but it also had 55 running ones. Hong Kong had 29 defaults.
Asia had the highest number of running exchanges at 129, while Europe had the most defaults at 60. The United Kingdom, the United States and Australia each had an equal number of running and defaulted exchanges. These are counts, not failure rates, so they say more about where exchanges are based than where they are riskiest.
Exchanges that hold client money in custody carry a particular risk. The QuadrigaCX case shows how one person holding the only keys to the wallets can lock thousands of customers out of their funds.
Stock exchanges rarely fail because they do not hold client money in the same way and operate under strict rules. Many crypto exchanges operate in a legal gray area, which makes it hard to hold them to account.
A separate summary report puts the total at no fewer than 17 major exchange bankruptcies since 2009. It says more than five exchanges collapsed in 2022 alone, after the fall of TerraUSD and FTX.
The same report puts investor losses from exchange bankruptcies at well over $30 billion. Its headline range is $30 billion to $50 billion.
Retail Investor Losses: Most Users Lost Money
The BIS built a database of crypto exchange app use in 95 countries from August 2015 to mid-December 2022. It drew on more than 200 apps, using downloads as a stand-in for new users.
Between August 2015 and November 2021, the price of bitcoin rose from $250 to $69,000. Over the same period, the monthly average of daily active users grew from around 100,000 to more than 30 million worldwide.
New users tended to arrive after prices had already risen. User numbers lagged price gains by about two months on average.
Almost three quarters of users downloaded a crypto app when bitcoin traded above $20,000. That put many buyers in at high prices just before the 2022 drop.
To estimate results, the BIS assumed each new user bought $100 of bitcoin in the month of their first download and every month after. Under that setup, the median investor would have lost $431 by December 2022. That is almost half of the $900 put in.
In nearly all of the countries studied, a majority of users probably lost money. The share was even higher in emerging markets such as Brazil, India, Pakistan, Thailand and Turkey.
If users kept investing every month, over four fifths would have lost money. The BIS notes these are estimates based on app data, not account records.
App use also varied widely. Some places had fewer than one monthly user per 100,000 people, while others had 1,500 or more. The top countries for downloads included Australia, Canada, the Netherlands, the United Kingdom, the United States, Singapore, Turkey and the United Arab Emirates.
The BIS also tested what drives people to join. Bitcoin’s price was a much stronger predictor of new users than stock market results, gold prices or global uncertainty.
In plain terms, people tend to enter when prices are high and rising. That is the point when the risk of a fall is greatest.
Bitcoin Bankruptcy Probability
A 2022 study by Serhii Kozlovskyi and colleagues asked a different question. How likely is an investor to go bankrupt by holding bitcoin?
The team used the Monte Carlo method, which runs many random simulations, on data from September 1, 2014 to July 1, 2022. They chose bitcoin because it is the first decentralized cryptocurrency and has the longest price history.
| Investment strategy | Probability of bankruptcy |
|---|---|
| Short term | 17% to 23% |
| Long term | 13% to 16% |
The long-term range sits below the short-term range. The authors say this runs against a common belief, and that investors who want to avoid bankruptcy should prefer a long-term approach.
They also suggest that alternating long and short investment periods is the best way for crypto investors to protect themselves.
The study gives some context on the size of bitcoin. There are around 25 million bitcoin holders, who represent 42.2% of the crypto market, and more than 10,000 types of cryptocurrency exist. The authors report that almost all coins have been shown to follow bitcoin.
The study’s wording matters. It treats bankruptcy as the result of tail events, which are rare but severe price moves that can wipe out an investor’s funds.
The authors also point out that crypto is not regulated like other assets. People who invest do not have the same protection they would have with stocks or bonds, so each investor must pick a strategy that fits their goals and limits.
What Separates Running Exchanges From Failed Ones
The Sapkota study compared the average features of running and defaulted exchanges. Some gaps were large, while others were small enough to be chance.
Running exchanges listed more coins, had higher ratings and had been open longer. Defaulted exchanges lacked certification and referral programs far more often.
| Measure | Running exchanges | Defaulted exchanges |
|---|---|---|
| Average coins listed | 63 | 27 |
| Median coins listed | 22 | 12 |
| Average age in years | 5.6 | 3.0 |
| Average rating (scale of 0 to 3) | 0.51 | 0.06 |
| Certified | 23.0% | 3.8% |
| Has an affiliation program | 26.7% | 11.5% |
| Supports wire transfers | 59.5% | 47.1% |
| Allows U.S. customers | 61.4% | 70.1% |
Age stood out most. Running exchanges had operated for about 5.6 years on average, against 3.0 years for those that failed.
Fees told a different story. Average maker and taker fees did not differ in a way that could be told apart from chance, and the same was true of the average home-country transparency score.
Risk Factors Behind Exchange Defaults
Averages can hide overlap, so the study also ran logit and probit models. These estimate how each factor changes the odds of default while holding the others fixed.
Both models agreed on the main results. The odds of default fell with more years of operation and with more coins listed.
| Factor | Direction | Change in odds of default |
|---|---|---|
| Decentralized exchange | Lower | 72.5% lower than centralized |
| More coins listed | Lower | 47.1% lower per one-unit rise in the log of coins |
| Allows U.S. customers | Higher | 52.1% higher |
| Higher home-country transparency | Higher | 1.5% higher per index point |
| Higher withdrawal fees | Higher | Small rise, statistically clear |
| Better rating, affiliation program | Lower | Weak evidence, only at the 10% level |
Some results need care. The author notes that a simple comparison found no clear gap between decentralized and centralized exchanges. Of the defaulted exchanges, 12% were decentralized, against 18% of those still running.
The U.S. and transparency results may reflect scrutiny. Exchanges that serve U.S. customers, or sit in countries with strong disclosure rules, may have their failures spotted and recorded more often.
In a breakdown by cause, U.S. access was linked to defaults from scams only. The author suggests that open, well-regulated markets can also attract fraud.
How Well Can Models Predict Exchange Failure?
The study tested whether public data can tell failing exchanges from healthy ones. The logit model reached 81.18% accuracy and the probit model 81.42%.
The author then tried machine learning methods. Results depended on how much data was used to train each model.
| Model | 60% training | 70% training | 80% training | 90% training |
|---|---|---|---|---|
| Random forest | 92.8% | 95.2% | 98.8% | 81.9% |
| Support vector machine | 82.8% | 81.9% | 79.9% | 78.8% |
| Gradient boosting | 82.5% | 82.6% | 82.2% | 82.1% |
| Neural networks | 77.4% | 73.8% | 78.0% | 79.5% |
| Naive Bayes | 62.9% | 61.9% | 61.3% | 59.0% |
Random forest scored highest, but its accuracy dropped sharply when 90% of the data was used for training. The author reads this as a sign that the result depends on the data split.
A five-fold cross-validation test gave a steadier picture. Random forest scored 81.7%, the support vector machine 80.5% and gradient boosting 78.9%.
One method, XGBoost, hit 100% accuracy. The author treated that as a sign of memorizing the data, since the sample of 845 exchanges is small. Naive Bayes scored lowest, at 59% to 63%.
A stacked model that combined all five methods scored between 81.7% and 83.4%. Most of its power came from the gradient boosting piece, which suggests the extra methods added little.
In short, traditional models at about 81% held up well against more complex tools. Public data can flag risk, but it cannot predict every collapse.
Failed Tokens and What the Data Means
Exchanges are only one part of the picture. A summary report on crypto failures counts the tokens that stopped trading or were abandoned.
It says that out of nearly 7 million cryptocurrencies listed since 2021, 3.7 million have failed. That equals 52.7% of all projects.
The report says 1.8 million failures came in the first quarter of 2025 alone. It lists the causes as lack of use, poor security, rug pulls and scams, market swings and regulatory pressure.
It also gives dollar estimates. It puts the Q1 2025 token failures at nearly $250 billion and total investor losses above $2 trillion.
These figures come from a summary that does not show its method, so readers should treat them with more caution than the peer-reviewed numbers above. An earlier study cited by Sapkota found that newer coins are more likely to fail than older ones.
Taken together, the sources point the same way. Failure is common in crypto, and the cost has fallen hard on retail users.
Exchange age, coin variety and certification all lined up with survival in the data. The BIS also found that the 2022 crash stayed mostly inside crypto, with little visible effect on wider markets.
The data has limits. It relies on open-source records, covers exchanges only up to February 2023, and cannot prove cause and effect.
