Decline in Crypto Loans: How Is It Different from 2022? – BitPlanet

Over three quarters, there has been a decrease of 10%, 5%, and 17%, with deleveraging occurring without a chain of bankruptcies.

[Block Media = BitPlanet] The balance of loans secured by crypto assets decreased by nearly 17% in just the last two quarters. This article examines the background of this decline in loans.

EXECUTIVE SUMMARY

Loan Balances Have Decreased for Three Consecutive Quarters, but No Chain Bankruptcies Occurred

Key Points

The balance of loans secured by crypto assets decreased by $11.33 billion (-16.78%) to $56.16 billion in Q2 2026. Compared to the peak in Q3 2025, this represents a decrease of about 28.6% (self-calculated). This is the first time since Q4 2022 that the collateral amounts for CeFi, DeFi, and CDP have all decreased. During the same period, Bitcoin reached an all-time high of $126,080 on October 6, 2025 (according to CoinGecko) and fell to around $64,500 by August 17, 2026.

Core Significance

This decline is different from that of 2022. In Q2 2022, the loan balance dropped by 55% in just one quarter, whereas this time it has decreased gradually by about 10%, 5%, and 17% over three quarters. So far, there have been no confirmed chain bankruptcies of major lenders that could lead to a credit crunch across the market.

The decrease in balance reflects voluntary repayments, liquidations due to price drops, and a decrease in the dollar equivalent of loans borrowed in coins. It is difficult to distinguish the contributions of each factor based solely on public data. However, it can be confirmed that the speed and manner of the decrease in loan balances differ from those in 2022.

However, the decrease in balance and the risk of remaining debt are separate issues. According to a filtered Aave V3 core loan ledger, about half of the total debt is concentrated in E-mode loans, which have a collateralization ratio (LTV) of about 90%.

Key Monitoring Indicators

The first indicator to watch is whether the balance of DeFi loans continues to rebound. The balance, which was $20.43 billion at the end of Q2, increased by about 7.4% to $21.94 billion on July 21 (self-calculated). It is necessary to confirm whether this rebound is a temporary movement or leads to a quarterly increase.

The second indicator is the gap between the weighted average borrowing rate of stablecoins and the federal funds rate. After the end of the quarter, the weighted borrowing rate for stablecoins was 3.88%, while the federal funds rate was maintained at 3.50-3.75% as of August 20. The key question is whether this gap narrows further.

Finally, we need to observe whether the quarterly loan balance drops by about 29% or more, similar to Q4 2022, or whether there are suspensions of withdrawals or bankruptcies of major lenders within one or two quarters after such a drop. If such signals appear, the current decline may need to be re-evaluated as a different form of adjustment compared to 2022.

01. Crypto Collateral Loan Balance Decreased to 71% of Peak Level After Three Quarters

Galaxy Research released its Q2 leverage report on August 17, 2026. According to the report, the balance of crypto collateral loans is $56.16 billion. Here, crypto collateral loans refer to loans taken out using cryptocurrencies like Bitcoin as collateral, for which stablecoins or fiat currencies are borrowed and not yet repaid.

The loan balance for Q2 decreased by $11.33 billion (-16.78%) compared to the previous quarter. Compared to the peak of $78.69 billion in Q3 2025, this represents a decrease of $22.53 billion, or about 28.6% (self-calculated, based on quarter-end data).

Galaxy Research indicated a decrease rate of 40.13% from the peak. However, this figure essentially matches the value obtained by dividing the decrease amount of $22.53 billion by the current balance of $56.16 billion rather than the peak balance (i.e., $22.53 billion ÷ $56.16 billion ≈ 40.1%, self-calculated). This report uses the conventional decrease rate of about 28.6% based on the peak balance. Specific formulas and reasons for application are summarized in Appendices 1 and 2. The peak balance was initially reported as $73.59 billion but was later revised to $78.69 billion.

The loan balance is aggregated in dollar terms, but the way asset price declines are reflected in the balance varies depending on the loan structure. A significant portion of crypto collateral loans is borrowed in stablecoins or fiat currencies, and CDP debts are also expressed in dollar terms. Therefore, a drop in the price of Bitcoin used as collateral does not automatically reduce the loan balance itself.

Instead, if the collateral value falls, loans may be forcibly repaid through liquidation, or borrowers may repay debts in advance to avoid liquidation, thus reducing the balance. A decrease in dollar-equivalent balance without separate repayments occurs in loans borrowed in cryptocurrencies like Ethereum. If the price of the borrowed coin falls, the dollar-equivalent amount of the same quantity of debt also decreases.

Bitcoin reached an all-time high of $126,080 on October 6, 2025 (according to CoinGecko) and was trading at around $64,500 on August 17, 2026. Based on these two points, this represents a decrease of about 49% (self-calculated). The three quarters in which the loan balance decreased coincide with this price decline period.

Thus, this balance decrease likely includes voluntary repayments, liquidations due to price drops, and a decrease in the dollar equivalent of loans borrowed in coins. However, public data alone cannot distinguish the proportion of each factor in the total decrease. Therefore, this analysis focuses on the speed and form of the decrease in loan balances rather than definitively identifying the causes of the decrease.

By category, the decrease in on-chain loans was the largest. The outstanding borrowing balance of DeFi loan applications decreased by $7.79 billion (-27.61%) to $20.43 billion, marking a decrease for three consecutive quarters. The CeFi loan balance decreased by $2.45 billion (-9.62%) to $22.98 billion. The collateral loan balance of CDP stablecoins issued against crypto assets also decreased by $1.09 billion (-7.86%).

The CeFi loan balance has surpassed that of DeFi loan applications for the first time since Q3 2023. As of June 30, 2026, the market shares are CeFi 40.93% (+324bp), DeFi applications 36.37% (-544bp), and CDP collateral 22.70% (+220bp).

However, this should not be interpreted as DeFi funds moving to CeFi. The CeFi balance has also decreased, and the DeFi balance has been decreasing relatively faster, resulting in a reversal of market share.

02. In 2022, there was a 55% drop in one quarter, while this time there has been a 10%, 5%, and 17% decrease over three quarters {#rps-4}

During this adjustment period, there were also large-scale liquidations. On October 10, 2025, approximately $19 billion worth of positions were forcibly liquidated in the perpetual futures market. This was the largest scale on a single trading day [4]. However, the scale of liquidations in the futures market and the crypto collateral loan balance are different indicators. In fact, the collateral loan balance in Q4 2025 only decreased by about 10%, and the large-scale liquidations in the futures market did not lead to a rapid decline in the loan balance due to a chain reaction of credit tightening [1].

2022 was different. At that time, the crypto collateral loan balance dropped by more than 55% in just the second quarter, followed by decreases of -9% in the third quarter and -29% in the fourth quarter [1]. The collapse of Luna triggered defaults among large borrowers like 3AC (Three Arrows Capital), and the liquidity crises and bankruptcies of major lenders such as Celsius, Voyager, and BlockFi occurred in a chain reaction over several months [5].

In contrast, this time, from Q4 2025, the balance decreased by approximately 10%, 5%, and 17% over three quarters. So far, there have been no confirmed chain bankruptcies of major CeFi lenders that could cause a widespread credit tightening in the market [1]. However, since in 2022 the loan balance dropped sharply first, followed by bankruptcies after a certain lag, the possibility of future bankruptcies cannot be ruled out.

One of the key factors that exacerbated the credit tightening in 2022 was the accumulation of unsecured credit among institutions [5]. Since then, the market structure has changed. CeFi lenders have tightened collateral requirements, and the weight of the loan market has shifted to an on-chain market where collateral and debt status can be verified on public ledgers [4]. In on-chain lending, if the collateral ratio falls below a certain threshold, automatic liquidation occurs. Unlike in 2022, where unsecured credit accumulated and collapsed all at once, this time, a significant portion of the debt is secured by collateral, and the process of reducing debt can also be verified on-chain.

Galaxy Research interpreted this phase not as a collapse led by forced liquidations or the bankruptcies of trading counterparties, but as a deleveraging process where market participants gradually reduce their risks [1]. Deleveraging refers to the process of repaying loans to reduce the scale of debt. However, Galaxy also added a caveat in the original text stating “in our view” regarding this interpretation [1]. This is merely the interpretation of the original research publisher, and it is noted that Galaxy is directly engaged in the crypto lending business, which will be examined separately in Chapter 5. Regardless of the interpretation, the fact that the balances have gradually decreased over three quarters without a chain of bankruptcies among major lenders is confirmed by data [1].

Looking solely at DeFi, the decrease is much larger. The borrowing balance of DeFi lending applications reached an all-time high of $47.13 billion on September 19, 2025 [1]. Subsequently, on July 21, 2026, it dropped to $21.94 billion, a decrease of $25.19 billion (-53.45%) from the peak [1].

Recently, there have been signs of a partial reversal in the downward trend. The DeFi loan balance, which was $20.43 billion at the end of Q2 2026, increased by about 7.4% to $21.94 billion on July 21 (self-calculated) [1]. It remains to be seen whether this rebound is a temporary movement or will lead to a real quarterly increase.

During the second quarter, there was also a major exploit. On April 18, the rsETH bridge path of Kelp was attacked in the LayerZero V2 Unichain segment connecting to Ethereum, resulting in the outflow of 116,500 rsETH, worth approximately $292 million [6][7]. An exploit refers to an attack that uses vulnerabilities in the code to steal assets.

The attacker borrowed about $190 million from Aave using the stolen rsETH as collateral, leaving behind debt that effectively had no collateral value [6][7]. However, Aave’s own contract was not attacked, and Aave subsequently froze the rsETH market [6].

Half of the Remaining Debt in Aave Concentrated in E-mode with LTV of About 90% {#rps-5}

Just because the overall loan balance has decreased does not mean that the risks of the remaining debt have also diminished.

E-mode is a lending method that allows for a higher loan-to-value (LTV) ratio for collateral assets and borrowed assets that have a high correlation in price movements. Galaxy Research analyzed the loan structure of the Aave V3 core market as of August 7, 2026 [1].

Among the 19,073 outstanding loans that applied the conditions of a minimum debt of $100 and a health factor below 50, E-mode loans accounted for only 8.91% in terms of number of loans, but represented half of the debt amount [1]. The following figures are all based on the Aave V3 core market ledger under these conditions and cannot be generalized to the entire on-chain lending market.

The average collateral recognition ratio (LTV) of the debts in this ledger is about 90%, and the health factor is 1.06 [1]. If the health factor falls below 1, the position becomes subject to liquidation. 66.2% of the collateral assets are concentrated in Ethereum staking derivative tokens such as weETH, rsETH, and wstETH [1].

Ultimately, half of the remaining debt in the analyzed ledger was concentrated in high-LTV borrowing using Ethereum staking assets. While overall market leverage has decreased, this indicates that a significant amount of positions still remain based on high collateral utilization in some lending ledgers.

03. First Time Since Q4 2022 That All Three Categories Have Decreased Together {#rps-6}

Another characteristic of this decrease is that the CeFi, DeFi, and CDP collateral segments have all decreased simultaneously. This is the first time since Q4 2022 that all three categories have decreased in a single quarter [1]. This means that the issue is not limited to a specific channel, but rather that the borrowing scale has decreased across the entire market.

[Interpretation] While the fact that all three categories have decreased simultaneously does not definitively indicate the cause, it suggests that this adjustment cannot be easily explained by individual credit events such as the insolvency of specific lenders. If the problem of a specific operator were the main cause, the decrease would likely have been more concentrated in that channel.

However, this only indicates the possibility of common factors affecting multiple markets simultaneously, and does not prove causality. There is also a possibility that some loans were double-counted during the aggregation process between CeFi, DeFi, and CDP, and the impacts of this will be discussed separately in Chapter 5. It cannot be ruled out that some of the observed decline in the three categories stemmed from the aggregation method.

Common factors could include the decline in asset prices and changes in the borrowing yield environment. However, it is difficult to separate the extent to which each of these two factors has influenced the situation based solely on publicly available data, so this report does not separately estimate the contributions of these two factors.

A trend of decreasing leverage has also emerged in the derivatives market. As of the end of the quarter, the open interest in futures was $103.2 billion, a decrease of 3.08% from the previous quarter. Open interest refers to the total amount of derivative contracts that have not yet been settled or paid.

However, the extent of the decrease varied by asset. The open interest in Bitcoin futures decreased by 6.24% to $45.04 billion, while Ethereum saw a more significant drop of 26.31% to $21.99 billion. Although the overall decrease in open interest was not large, a relatively stronger reduction in leverage was observed in the Ethereum market.

04. Stablecoin Deposit Rates of Major Apps Lower than Traditional Financial Accounts as of April 7 {#rps-7}

The interest rate is more of a condition that makes it difficult for the withdrawn funds to return rather than a cause that triggered the decrease in loan balances.

As of April 7, 2026, the deposit yield for USDC on Aave was 2.61% per annum. On the same day, the interest rate applied by Interactive Brokers on idle cash was 3.14%, which was 0.53 percentage points higher than Aave (self-calculated). However, 3.14% is the tiered interest rate applied to amounts exceeding a balance of $10,000, and since it reflects the figures as of April 7, it may change thereafter.

At the same time, the deposit yield for USDT on Aave was 1.84%, and Compound’s USDC was 2.55%. Of course, there were products within Aave that offered yields in the 4-6% range, and other protocols had even higher yields. Therefore, the comparison here is limited to representative stablecoin deposit products of major protocols.

On-chain deposits carry additional risks compared to traditional financial accounts. In the first half of 2025, the total amount of assets stolen from the cryptocurrency market reached $2.47 billion. However, this figure includes the total damage from wallet and exchange breaches and does not solely refer to losses incurred from DeFi deposits. Nevertheless, it is noteworthy that the yield on on-chain deposits, which require taking on additional technical and protocol risks, was lower than that of traditional financial accounts at that time.

However, it is difficult to consider this yield inversion as the starting point for the decrease in loan balances. Loan balances began to decline as early as the fourth quarter of 2025, and the comparison reflects figures from April 2026, two quarters later. Additionally, the yield on on-chain deposits tends to decrease when borrowing demand diminishes, so the decline in yield itself may also be a result of the decrease in loan balances.

[Interpretation] Even if it is challenging to clearly delineate the direction of cause and effect, the impact on the market is evident. If the yield on on-chain deposits, which require taking on additional risks, remains lower than that of traditional financial accounts, the incentive for withdrawn funds to flow back in and for borrowing demand to recover weakens.

Borrowing costs have not sufficiently decreased either. The stablecoin-weighted borrowing rate is an indicator that averages the borrowing rates of major lending protocols and the CDP issuance fees weighted by the outstanding borrowing balances. Recently, on-chain borrowing costs, including this indicator, have effectively moved with the federal funds rate as a lower bound.

The target range for the federal funds rate has been held steady at 3.50-3.75% for five consecutive meetings. At the meeting on July 29, the decision to maintain rates was made with a vote of 9 to 3, with the three dissenters advocating for a 25 basis point increase.

After the end of the quarter, the stablecoin-weighted borrowing rate rose to 3.88%. However, the original text does not specify the exact observation date. This is 38 basis points higher than the lower bound of the federal funds rate target range of 3.50% and 13 basis points higher than the upper bound of 3.75% (self-calculated).

In other words, even though borrowers are borrowing amounts less than the value of the collateral, the borrowing costs are still forming at levels higher than the short-term policy interest rate. If the deposit yields are low and borrowing costs are high, it will be difficult to expect a rapid recovery in the lending market.

05. Reduction of Tether’s Collateral Loans and Increase in Six Other Lenders {#rps-8}

Not all businesses have reduced lending. In the CeFi market, the largest player, Tether, led the overall balance decrease by reducing its collateral loans. In contrast, six companies including Galaxy, Coinbase, Ledn, Arch, Sygnum, and Milo increased their loan balances during the quarter.

Tether’s market share in the CeFi sector fell by 371 basis points to 58.54% compared to the previous quarter. The combined market share of the top three players, including Maple (8.91%) and Nexo (7.51%), reached 74.96%. The reasons for Tether’s reduction in collateral loans are not confirmed in publicly available materials.

The balances and proportions cited in this report are based on data directly verified from Galaxy’s original text. However, the characterization of this phase as “orderly deleveraging” is an interpretation by Galaxy, the data publisher. The data itself has some limitations. Some CeFi figures were provided by unverified third parties, and there may be some overlapping aggregation between CeFi lending and DeFi borrowing, as well as between CeFi lending and CDP issuance.

Additionally, Galaxy is not just a simple data provider but also operates lending businesses in the aggregated market and has increased its own loan balances during this quarter. The legal notice in the original text also states that Galaxy affiliates are lending funds to the protocols covered in the report and that conflicts of interest may arise during the collateral liquidation process. Therefore, it is necessary to distinguish between the data and Galaxy’s interpretation.

Note: The Debt of Digital Asset Treasury Companies is a Different Form of Leverage than Collateral Loans {#rps-9}

The figures below are not included in the previously examined crypto collateral loan statistics. The debt used in the digital asset treasury strategy, which involves purchasing and holding crypto assets with funds raised by companies, amounts to $16.1 billion. This is more centered around corporate-issued debt rather than loans borrowed against crypto assets.

As the $1.5 billion debt purchase completed by Strategy in May 2026 is reflected, the related debt balance has decreased to the level seen in July 2025. The total crypto-related debt in the industry, combining the debts of digital asset treasury companies and crypto collateral loans, stands at $73.2 billion, marking a decline for the third consecutive quarter.

Bitplanet, the issuing organization of this report, also holds Bitcoin as a financial asset. However, as of the semi-annual report in 2026, there are no records of providing collateral based on the held crypto assets or borrowing against them.

Conclusion: The decrease in loan balances is progressing differently than in 2022

Based on the data available so far, it is difficult to view the current decrease in loan balances as a repetition of 2022. The fact that the balances have not collapsed all at once but have decreased over three quarters, and that the collateral segments of CeFi, DeFi, and CDP have all reduced in the same quarter, can be directly confirmed from Galaxy’s time series data.

In contrast, there has been no confirmation of a chain reaction of defaults and bankruptcies among lenders, or a situation where loan balances rapidly decrease in a short period, similar to a bank run, during these three quarters. Of course, this judgment is limited to the situations observed so far. On October 10, 2025, the futures market experienced the largest liquidation in history, but this shock did not lead to a rapid contraction of collateral loan balances.

If this trend continues, the burden will fall on the on-chain loan market itself. As of April 7, 2026, the yield on stablecoin deposits in major lending apps is 2.61% for Aave USDC and 2.55% for Compound USDC, which is lower than the 3.14% applied by Interactive Brokers to idle cash on the same day.

On the other hand, the overall market’s weighted borrowing rate for stablecoins confirmed after the end of the quarter is 3.88%, which is 38 basis points higher than the lower limit of the federal funds target range at 3.50%. Although the reference points for these two figures are different, the direction is the same. Those depositing funds receive lower yields than traditional financial accounts, while borrowers bear higher costs than short-term policy interest rates.

[Interpretation] In the current interest rate environment, there is little incentive for funds to quickly return to the on-chain loan market. There is little reason to receive lower interest than traditional financial accounts while taking on additional technical and protocol risks, and from the borrower’s perspective, the incentive to borrow while incurring higher costs than short-term policy interest rates is limited. However, this judgment is based on figures from some major apps and specific points in time, and there is no data available showing the average deposit yield of the entire on-chain market.

[Interpretation] For the funds that have exited to flow back in, two major conditions are needed. One is a reduction in policy interest rates that narrows the gap with on-chain borrowing costs, and the other is the re-establishment of yields on-chain that exceed those of traditional finance. However, on July 29, the Federal Reserve decided to keep rates unchanged by a vote of 9 to 3, with the three dissenting members advocating for a rate increase. Therefore, it is currently difficult to predict when a reduction in policy interest rates will occur.

[11] Strategy Completes $1.5 Billion Debt Repurchase and Achieves BTC Yield of 13.3% YTD [1st] — Strategy, 2026-05-26 https://www.strategy.com/press/strategy-completes-1-5-billion-debt-repurchase-and-achieves-btc-yield-of-13-3-ytd-now-holds-843738-btc_05-26-2026

[12] Bitplanet Semiannual Report (2026.06). Check the borrowing notes and collateral asset notes directly by searching the company name in the electronic disclosure system, submitted on 2026-08-14 [1st] — Electronic Disclosure System (DART), 2026-08-14 https://dart.fss.or.kr

[13] DeFi Lending Protocols Rankings (as of 2026-08-24, screen values, manually secured) — DeFiLlama, updated regularly https://defillama.com/protocols/lending

Appendix: Data Limitations and Calculation Basis

① Limitations of Analysis and Data

Limitations of Aggregation Agency Uniqueness: The time series of balances by category relies on a single aggregation from Galaxy. The original text revealed the possibility of double counting and specified that some figures are unverified third-party contributions [1]. The DeFi section was contrasted with the independent aggregation from DeFiLlama at the scale level. As of the inquiry on August 24, 2026, the total borrowing balance of the top 10 protocols in the Lending category is approximately $24.3 billion (self-aggregated) [13], which aligns with Galaxy’s aggregated figure of $21.94 billion as of July 21 and does not contradict the rebound direction. The two aggregations differ in the range of included protocols, so value alignment cannot be expected. A comparison of the end-of-quarter (June 30, 2026) point values was not performed.

Understanding Source Interests: Galaxy is a lender in the market being aggregated (as specified in Section 5 of the text).

Limitations of Data Revision: The peak figure for Q3 2025 was reported as $73.59 billion at the time of announcement [4], but the current report uses a basis of $78.69 billion. This is presumed to be due to aggregation revisions. The circumstances of the revision could not be confirmed in the original text.

Criteria for Adopting Conflicting Figures: The original text indicated a decrease rate of 40.13% from the peak, but there is no footnote for the calculation. Dividing the decrease amount of $22.53 billion by the end-of-quarter balance of $56.16 billion yields approximately 40.1%, suggesting that the original value was calculated with the end-of-quarter balance as the denominator. The conventional decrease rate based on the peak of $78.69 billion is about 28.6%. This report uses the latter in the text and notes the estimation of the original value (Sections 1 and Appendix ②).

Nature of Remaining Leverage: The Aave E-mode figure in Section 2.1 is a snapshot from August 7, 2026, and is not based on the end-of-quarter. It is the aggregated value filtered by a minimum debt of $100 and a health indicator below 50 from the Aave V3 core ledger. It does not represent the entire on-chain lending.

Limitations of Driver Separation: The extent to which the three methods of balance reduction (voluntary repayment, liquidation and proactive repayment due to price decline, and reduction in the displayed amount of borrowed balance in coins) contributed cannot be separated with public data (as specified in Section 1).

Observation Range of Deposit Yields: 2.61% (Aave USDC), 2.55% (Compound USDC), 1.84% (Aave USDT), and 3.14% (IBKR) are single-point figures from individual apps and assets as of April 7, 2026. An average deposit yield indicator for the entire on-chain market has not been secured.

Price Precision and Aggregation Source Differences: The all-time high price of Bitcoin and the price on August 17 are based on CoinGecko aggregation. Values may vary by aggregation source. The description on August 17 is a snapshot at a specific point (approximately $64,500), and the decline rate of about 49% is also an approximation based on the snapshot.

Figure Decimal Places: The market share of 22.70% is a value standardized to match the decimal places of 40.93% and 36.37% from the original notation of 22.7%. The value itself is the same as the original, and no new information has been added through decimal standardization.

② Self-Calculated Formulas

{#rps-14}

③ Methodology

• Priority of Aggregation Agencies: Original data publishing agencies are prioritized, and re-quoted reports are used only for comparison.
• Primary and Secondary Distinction: Only materials directly verified from the original text are given a primary badge.
• Reference Date and Inquiry Date: Time series figures indicate the reference date, while dashboard screen values indicate the inquiry date.
• Conflicting Figures: Both are noted, and the basis for adoption is stated.
• Self-Calculation: This is noted in the text, and formulas are listed 1:1 in Appendix ②.
• Legal and Regulatory Interpretation: Descriptions are made only within the scope of confirmed stages and wording in official documents.
• Disclosure of Conflicts of Interest: The same disclosure wording is published in a fixed position at the top of all reports.

Published by Bitplanet Research Lab · Written by Kim Tae-won · Reviewed by Kim Soo-young