Galaxy Research: Crypto lending has contracted for three consecutive quarters, and the market is orderly deleveraging

Author: Zack Pokorny, Research Assistant at Galaxy Research

Compiled by: Jiahua, ChainCatcher

Introduction

The second quarter of 2026 marks the first quarter since the fourth quarter of 2022 where the collateralization of crypto assets in CeFi, DeFi, and collateralized debt positions (CDP) stablecoins has simultaneously declined, continuing the previous trend of deleveraging in the market.

Compared to the last bear market cycle, the most notable difference is that outstanding loans are decreasing in a stable, stepwise manner rather than collapsing suddenly.

In the second quarter of 2022, the market size of crypto asset collateralized lending plummeted by over 55%, followed by declines of 9% and 29% in the third and fourth quarters, respectively. In contrast, during the recent deleveraging cycle, the market has seen declines of only 10%, 5%, and 17% over three consecutive quarters.

In our view, this relatively mild pace indicates that the current deleveraging process is much healthier, primarily driven by a voluntary market contraction rather than forced liquidations or institutional defaults.

If lending activity continues to contract over the next few quarters, we expect this stepwise decline to persist rather than repeating the severe and cascading losses seen during the 2022 deleveraging process.

There has also been some degree of deleveraging in corporate treasury, mainly due to Strategy's repurchase of $1.5 billion in debt in May 2026. This has reduced the debt used by corporations to supplement digital asset treasury strategies to $16.1 billion, roughly returning to the debt levels these companies had in July 2025.

In the futures market, open interest (OI) at the end of the second quarter was basically flat, decreasing by 3.08% quarter-over-quarter to $103.2 billion.

However, this relatively mild overall decline masks more pronounced changes within the market: BTC open interest fell by 6.24% to $45.04 billion; ETH open interest saw a larger decline of 26.31%, reducing to $21.99 billion. By the end of the quarter, BTC and ETH together accounted for 65% of the total futures open interest.

Notably, this relative stability in the futures market did not persist. By the end of July, open interest had rebounded to about $114 billion, with BTC open interest bouncing back to around $48 billion and ETH recovering to $25.74 billion, both moving away from their second-quarter lows.

Key Points

• In the second quarter of 2026, the overall size of crypto asset collateralized lending decreased by $11.33 billion, a decline of 16.78%, to $56.16 billion. This represents a 40.13% drop from the peak of $78.69 billion in the third quarter of 2025.

• In the second quarter, USD-denominated outstanding loans in DeFi lending applications fell for the third consecutive quarter, decreasing by $7.79 billion, a decline of 27.61%, to $20.43 billion.

• Galaxy Research currently tracks approximately $16.1 billion in outstanding debt, which corporations use to directly purchase digital assets or to fund their digital asset treasury strategies through leveraged holdings.

• Including perpetual contracts, futures open interest decreased by 3.08% quarter-over-quarter to $103.2 billion.

Crypto Asset Collateralized Lending

The market landscape below showcases some major historical and current participants in the CeFi and DeFi crypto lending markets. As crypto asset prices plummeted and market liquidity dried up, some of the largest CeFi lending institutions went bankrupt in 2022 and 2023. These institutions are marked with red warning dots in the chart below.

Figure: Major participants in the CeFi and DeFi crypto lending markets, along with key CeFi lending institutions that went bankrupt from 2022 to 2023

CeFi

The table below compares the CeFi crypto lending institutions covered in our market analysis. Some companies provide multiple services to investors. For example, Coinbase's core business is as an exchange, but it also offers credit to investors through over-the-counter crypto loans and margin financing. However, this analysis only accounts for the size of these companies' crypto asset collateralized loans.

Figure: Major CeFi crypto lending institutions and their crypto asset collateralized loan sizes

As of June 30, Galaxy Research tracked a CeFi outstanding loan size of $22.98 billion, a quarter-over-quarter decrease of 9.62%, or a reduction of $2.45 billion. Compared to the bear market low of $6.8 billion in the fourth quarter of 2023, this represents an increase of $16.14 billion, a growth rate of 235.94%. However, the current CeFi outstanding loan size is still 37.16% lower than the historical peak of $36.58 billion in the first quarter of 2022.

In the second quarter, the overall loan size of CeFi lending institutions contracted, but this decline was mainly due to a reduction in Tether's outstanding secured loans. Galaxy, Coinbase, Ledn, Arch, Sygnum, and Milo all saw increases in their loan sizes in the second quarter.

Figure: Changes in loan sizes of major CeFi lending institutions in the second quarter

In our analysis, Tether remains the dominant lending institution, accounting for 58.54% of the CeFi lending market, a decrease of 371 basis points from the previous quarter. Adding Maple, which has a market share of 8.91% (an increase of 52 basis points), and Nexo, with a market share of 7.51% (an increase of 49 basis points), the top three CeFi lending institutions we track together control 74.96% of the market, down 270 basis points from the previous quarter.

When comparing the market shares of various institutions, it is important to note that there are significant differences among CeFi lending institutions. Some institutions only offer specific types of loans, such as accepting only BTC collateral, providing altcoin collateral products, or issuing fiat loans instead of stablecoin loans; some institutions only serve specific types of clients, such as institutional or retail investors; and some operate only in specific jurisdictions. These factors collectively determine the ability of different lending institutions to scale their businesses.

Figure: Market shares of major CeFi lending institutions

The table below details the sources of data Galaxy Research obtained for each CeFi lending institution and the methods used to calculate their loan sizes. DeFi and on-chain CeFi lending data can be directly obtained from the blockchain, with data being transparent and easily accessible, while obtaining CeFi data is much more difficult. This is mainly because different CeFi lending institutions have inconsistent methods for counting outstanding loans, the frequency of publicly available information varies, and such data is often difficult to obtain.

Note: Data from private third-party lending institutions has not been formally verified by Galaxy Research.

Figure: Data sources and loan size calculation methods for various CeFi lending institutions by Galaxy Research

CeFi vs. DeFi Lending

In the second quarter, USD-denominated outstanding loans in DeFi lending applications fell for the third consecutive quarter, decreasing by $7.79 billion, a decline of 27.61%, to $20.43 billion. When combining DeFi applications with CeFi lending platforms, the total amount of crypto asset collateralized outstanding loans at the end of the quarter was $43.41 billion, a quarter-over-quarter decrease of $10.24 billion, or 19.08%, primarily due to the contraction in on-chain loan sizes.

Notably, this is the first time since the third quarter of 2023 that the size of outstanding loans in CeFi has surpassed that of DeFi lending applications.

Note: There may be double counting between the total loan size of CeFi and DeFi borrowing, as some CeFi institutions may use DeFi applications to provide loans to off-chain clients. For example, suppose a CeFi lending institution uses idle BTC as collateral to borrow USDC on-chain and then lends those USDC to an off-chain borrower.

In this case, the on-chain borrowing of this CeFi institution would be counted as both DeFi outstanding borrowing and appear in the institution's financial statements as outstanding loans issued to clients.

Due to a lack of sufficient disclosure or on-chain address attribution data, it is difficult to completely eliminate this double counting.

Figure: Comparison of outstanding crypto asset collateralized loans between CeFi and DeFi

Since the quarter-over-quarter decline in outstanding loans in DeFi lending applications significantly exceeded that of CeFi platforms, DeFi's previous lead over CeFi disappeared in the second quarter. By the end of the second quarter of 2026, the market share of DeFi lending applications dropped to 47.05%, a quarter-over-quarter decrease of 555 basis points, down from 52.6% at the end of the first quarter of 2026.

Figure: Changes in market shares of CeFi and DeFi lending

The third component, the portion of the CDP stablecoin supply generated by crypto asset collateral, decreased by $1.09 billion, a decline of 7.86%. Similarly, there may be double counting between the total loan size of CeFi and the CDP stablecoin supply, as some CeFi institutions may mint CDP stablecoins by collateralizing crypto assets to finance loans provided to off-chain clients.

Overall, in the second quarter of 2026, the size of crypto asset collateralized lending decreased by $11.33 billion, a decline of 16.78%, to $56.16 billion. This represents a 40.13% drop from the peak of $78.69 billion in the third quarter of 2025.

Figure: Total size of crypto asset collateralized lending and changes in different components

As of the end of the second quarter of 2026, DeFi lending applications accounted for 36.37% of the entire crypto asset collateralized lending market, down 544 basis points from the first quarter of 2026; CeFi platforms accounted for 40.93%, an increase of 324 basis points; and the portion generated by crypto asset collateral in the CDP stablecoin supply accounted for 22.7%, an increase of 220 basis points.

If DeFi lending applications and CDP stablecoins are combined, on-chain lending platforms still account for 59.07% of the entire market, down 324 basis points from the first quarter of 2026.

Figure: Shares of DeFi, CeFi, and CDP stablecoins in the crypto asset collateralized lending market

Other Perspectives on DeFi Lending

Since reaching a historical peak of $47.13 billion on September 19, 2025, outstanding loans in DeFi lending applications have significantly shrunk. As of July 21, 2026, the on-chain loan size was $21.94 billion, a decrease of $25.19 billion from the peak, representing a decline of 53.45%.

Figure: Changes in outstanding loans in DeFi lending applications since the 2025 peak

Since the end of the first quarter of 2026, the withdrawal of outstanding loans in DeFi has further intensified, but recently it has begun to ease slightly.

Figure: Recent changes in the magnitude of DeFi outstanding loan withdrawals

Stablecoins

In the second quarter, the weighted average borrowing rate for stablecoins increased. According to the seven-day moving average, from March 31 to June 30, the stablecoin borrowing rate rose by 27 basis points. After the end of the quarter, the stablecoin rate continued to rise to 3.88%.

This indicator combines the borrowing costs in lending protocols with the costs of minting CDP stablecoins, weighted by the size of outstanding loans.

Figure: Trends in the weighted average borrowing rate for stablecoins

The following chart further breaks down the costs of borrowing stablecoins through lending applications and the costs of minting CDP stablecoins using crypto asset collateral. The trends of the two types of rates are very close, although the CDP stablecoin minting rates typically fluctuate less because these rates are manually adjusted periodically and do not fully follow market changes. Over the past 21 months, both types of rates have used the federal funds rate as a lower limit.

Figure: Comparison of DeFi stablecoin borrowing rates, CDP stablecoin minting rates, and federal funds rates

Throughout the second quarter, the benchmark over-the-counter (OTC) borrowing rate for USDC fluctuated between 4.25% and 5%. By the end of the quarter, this rate was 4.25% and remained so until August 3.

Figure: Trends in USDC on-chain and OTC borrowing rates

The following chart tracks the same rate indicators, but for USDT. Similar to the OTC rate for USDC, the relevant rates for USDT also fluctuated between 4.25% and 5%.

Figure: Trends in USDT on-chain and OTC borrowing rates

Bitcoin

The following chart shows the weighted borrowing rates for Wrapped Bitcoin (WBTC) across multiple lending applications and blockchain networks. The cost of borrowing WBTC on-chain is typically lower because wrapped Bitcoin is primarily used as collateral in on-chain markets, and the demand for borrowing WBTC itself is not high. Unlike stablecoins, the cost of on-chain BTC borrowing is relatively stable because users borrow and repay BTC less frequently.

In the second quarter, the on-chain BTC borrowing rate fluctuated between 0.44% and 0.5%.

Figure: Trends in the on-chain weighted borrowing rate for BTC

Throughout the second quarter, the long-standing gap between on-chain and off-chain OTC borrowing rates for BTC remained. In the OTC market, the demand for BTC borrowing primarily comes from two sources: 1) the demand to short BTC; 2) borrowing stablecoins or cash using BTC as collateral. The former demand is not common in the on-chain lending market, thus creating a spread between the borrowing costs of BTC on-chain and in the OTC market.

Throughout the second quarter, the OTC borrowing rate for BTC remained unchanged at 1%.

Figure: Comparison of on-chain and OTC borrowing rates for BTC

ETH and stETH

The following chart shows the weighted borrowing rates for ETH and stETH, which is the token obtained by staking ETH through the Lido protocol, across multiple lending applications and blockchain networks. Historically, the borrowing cost for ETH is usually higher than that for stETH because the market demand for borrowing ETH itself is greater. Users often borrow ETH to leverage their exposure to the Ethereum network's staking APY, using stETH obtained through Lido staking as collateral.

Thus, under normal market conditions, the borrowing cost for ETH typically fluctuates within a range of about 50 basis points above and below the Ethereum network's staking APY. Once the borrowing cost exceeds the staking yield, this strategy becomes uneconomical, so the ETH borrowing APR rarely stays above the staking APY for long.

Similar to WBTC, the borrowing cost for stETH is usually lower because this asset is primarily used as collateral.

Figure: Trends in the on-chain weighted borrowing rates for ETH and stETH

Users can use yield-generating liquid staking tokens (LST) or liquid re-staking tokens (LRT) as collateral to borrow ETH at lower rates, or even at negative net costs. This cost advantage drives a leveraged strategy: users repeatedly deposit LST and LRT as collateral, borrow un-staked ETH, stake it, and then use the newly acquired LST and LRT to continue borrowing more ETH, thereby amplifying their exposure to ETH staking APY.

This strategy is only viable when the borrowing cost for ETH is lower than the staking APY obtained from LST and LRT. Except for a few special periods, users are mostly able to execute this strategy smoothly.

Figure: Relationship between borrowing costs of ETH in leveraged staking strategies and staking yields

ETH OTC Borrowing Rates

Similar to Bitcoin, borrowing ETH through on-chain lending applications has historically been cheaper than borrowing through the OTC market. This is mainly due to two factors. First, similar to BTC, there is demand for ETH borrowing from short sellers in the off-chain market, which is not common on-chain.

Second, the Ethereum staking APY constitutes the lower limit for off-chain ETH borrowing rates. If the yield is lower than what can be earned by directly staking ETH, asset providers have little incentive to deposit ETH into off-chain platforms, and off-chain platforms lack the motivation to lend out at rates below staking yields. In contrast, in the on-chain market, the Ethereum staking APY often constitutes the upper limit for ETH borrowing rates.

Figure: Comparison of on-chain and OTC borrowing rates for ETH and Ethereum staking APY

Aave Lending Book Perspective

The following is a detailed analysis of the filtered Aave V3 Core lending book. Aave V3 Core is currently the largest on-chain lending market. We applied the following filtering criteria:

• Minimum debt size ($100): We excluded positions with debts below this threshold from the main aggregated data. This helps eliminate dust positions while making the statistics more biased towards larger loans.

• Health factor (HF) statistical upper limit (HF ≤ 50): We also excluded positions with snapshot health factors above 50 from the main statistical sample. These highly over-collateralized loans are usually small in size and have limited significance for risk analysis. In loans below this upper limit, the debt-weighted average HF and HF percentile statistics only include loans with snapshot health factors greater than or equal to 1 and less than or equal to 50.

  Loans with health factors below 1 are not included in these HF statistics, while loans with health factors equal to 1 are included. The higher the health factor, the safer the loan; a health factor below 1 means the position is already in liquidation territory. The formula for calculating the health factor is: (Total collateral value × Weighted average liquidation threshold) ÷ Total loan value.

• Debt-to-equity ratio (D/E): We only calculate this metric for loans where the collateral value exceeds the debt, meaning the equity is positive. Positions with non-positive equity are excluded from the D/E distribution and do not enter the debt-weighted average D/E. This rule is independent of the minimum debt size filter, meaning a loan may be excluded from D/E-related statistics even if the debt exceeds $100, as long as the equity is not positive.

As of the snapshot on August 7, 2026, after applying the above filtering criteria, there were a total of 19,073 outstanding loans. Although the number of Aave's "Efficiency Mode" (e-mode) loans accounts for only 8.91% of all outstanding positions, the outstanding debt size corresponding to e-mode and regular loans is roughly equal. In e-mode loans, the borrowing assets and collateral assets have a high correlation, such as ETH and WETH.

Compared to Galaxy Research's last analysis of the Aave lending book on April 22, 2026, this ratio has decreased, as e-mode and regular loans' debt sizes were about 60/40 at that time. The change mainly comes from the decline in outstanding debt size for e-mode loans.

Figure: Distribution of positions and debt sizes for e-mode and regular loans in Aave V3 Core

The table below summarizes the risk metrics calculated on a debt-weighted basis from the filtered lending book, listing all positions, e-mode positions, and regular positions separately.

On average, e-mode borrowers exhibit significantly higher leverage. Their debt-weighted LTV is approximately 90%, with a debt-weighted health factor of about 1.06 and a debt-to-equity ratio of about 10.7. This means that even a small shock to the collateral asset prices could quickly put a large number of such loans under pressure.

In contrast, non-e-mode loans have a significantly larger safety cushion, with a debt-weighted LTV of about 49%, an HF of about 1.79, and a D/E of about 1.07. This also somewhat compensates for the risks arising from the lack of price correlation between borrowing assets and collateral assets, such as borrowing USDC using cbBTC as collateral.

The formula for calculating the debt-weighted average is:

D/E = Σi (Di × (Di ÷ (Ci − Di))) ÷ Σi Di

The calculation scope only includes loans where Ci > Di, where Di represents the debt of a single position and Ci represents its collateral value.

Figure: Debt-weighted risk metrics for all positions, e-mode, and regular loans in Aave V3 Core

The next table ranks all enabled collateral assets in Aave V3 Core according to their corresponding dollar values in the statistical sample. Collateral assets related to ETH dominate: WETH, Etherfi's wrapped re-staked ETH (weETH), and Lido's wrapped stETH (wstETH) account for approximately 24%, 16%, and 14%, respectively, collectively making up 54.6% of all available collateral. WBTC also holds a significant share, approximately 14%.

Thus, a few assets actually carry most of the collateral in the entire lending book. Additionally, stablecoins and other yield-bearing tokens also hold smaller but still meaningful shares.

Figure: Dollar value and market share of various collateral assets in Aave V3 Core

The table below ranks borrowing assets by dollar size and their proportion of the total borrowing amount in the sample. WETH dominates the liability side, accounting for just over 37%. Considering the prevalence of strategies that leverage ETH-related assets as collateral in the market, this result is expected.

The borrowing size of stablecoins is also substantial, with USDT and USDC together accounting for about half of the total borrowing, with USDT accounting for about 28% and USDC about 22%, while the shares of other assets are significantly smaller.

Compared to our last analysis of the Aave V3 Core lending book, the share of WETH in outstanding liabilities has significantly decreased from 51.1%, which also aligns with the previously mentioned reduction in e-mode loan sizes.

Figure: Dollar size and proportion of various borrowing assets in Aave V3 Core

E-mode Perspective

In e-mode loans, collateral is highly concentrated in liquid staking and re-staked wrapped assets. The weETH alone accounts for about 42% of this category of collateral. When combined with rsETH and wstETH, these assets collectively account for about 66.2% of e-mode collateral.

Therefore, from a risk perspective, e-mode is less about having "diversified collateral" and more about being essentially a highly concentrated bet on Ethereum staking basis.

Figure: Composition of various collateral assets in Aave e-mode, where ETH staking and re-staking assets dominate

In e-mode borrowing, liabilities are almost entirely dominated by WETH. WETH alone accounts for about 73% of the total debt in e-mode. This outcome is entirely expected when users use assets highly correlated with ETH as collateral and then borrow ETH for leveraged strategies.

Stablecoin borrowing still holds a certain scale, with USDT, USDe, and USDC collectively accounting for about a few percent of e-mode borrowing.

Figure: Composition of various borrowing assets in Aave e-mode, with WETH holding absolute dominance

The table below ranks e-mode positions involving different collateral assets, displaying debt-weighted risk metrics and, where applicable, the implied cycles calculated from sub-samples of single assets with collateral concentration not less than 99%. Thus, unlike the relatively neutral market cap ranking above, this table resembles a "which asset is leveraged the most" leaderboard.

Liquid re-staked and liquid staked ETH wrapped assets appear at the top of the rankings. These assets have high debt-weighted LTVs, D/E ratios typically in the high single digits to tens, and health factors not far from 1. This aligns with high-leverage, repeated ETH Beta cycle strategies.

The implied cycles are calculated for each loan using the following formula:

Per-loan Ni = ln((1 − (D/E)i(1 − Li)) / Li) / ln(L_i)

Where the current LTV of each loan, Li, and the debt-to-equity ratio are used, and only sub-samples where a single asset constitutes at least 99% of the collateral are calculated. The final reported value is the average of each loan Ni weighted by debt size.

If a loan's Li does not fall within the (0,1) range, the logarithmic parameter is non-positive, meaning leverage has exceeded the cycle limit corresponding to Li/(1−L_i), or the calculation result is not finite, it will be excluded.

Values will only be displayed when there is at least one qualifying loan among liquid staking ETH, liquid re-staked ETH, yield-bearing stablecoins, or Pendle PT assets.

Figure: Leverage levels, health factors, and implied cycles for different e-mode collateral assets

Corporate Debt Strategy

We currently track $16.1 billion in outstanding debt, which corporations use to directly purchase digital assets or to fund their digital asset treasury strategies.

Due to limitations in Bloomberg's tracking of Strategy's preferred shares, the timing of the increase in the issuance of STRC shares may have biases in the time series. However, the total debt issuance size still represents the company's outstanding liabilities well.

In the second quarter, the outstanding debt issued by digital asset treasury companies (DAT) decreased by $1.5 billion, primarily due to Strategy completing a $1.5 billion debt repurchase in May.

Figure: Outstanding debt issued by digital asset treasury companies used for purchasing or supporting digital asset treasury strategies

The following chart details the actual interest payable each quarter on the debt issued by DAT.

It is important to note that the STRC dividends only need to be paid when the board formally announces them and the company has legally usable funds. However, any unpaid dividends will continue to accumulate and must be prioritized for payment before distributions to lower-tier securities.

Thus, the payment of STRC dividends may not be uniform and will not occur at fixed intervals.

Figure: Effective quarterly interest expenses corresponding to the debt issued by digital asset treasury companies

Including the debt borne by DAT, the total outstanding crypto-related debt decreased by 15.08% in the second quarter. After reaching a historical peak in the third quarter of 2025, the total outstanding debt formed through on-chain and off-chain channels fell to $73.2 billion by the end of the second quarter of 2026, marking a decline for the third consecutive quarter.

Figure: Changes in the total scale of on-chain, off-chain, and crypto debt related to digital asset treasury

Futures Market

Including perpetual contracts, futures open interest decreased by 3.08% to $103.2 billion. During July, open interest began to rise again, reaching about $114 billion by the end of the month.

It is important to note that the total scale of futures open interest cannot be directly viewed as the absolute amount of actual market leverage. This is because some futures positions may hedge against spot longs, giving traders a delta-neutral exposure to the underlying asset. Therefore, open interest alone cannot directly observe the true overall leverage ratio of the entire market.

Figure: Total scale of crypto futures open interest, including perpetual contracts

Throughout the second quarter, BTC futures open interest fluctuated between $44 billion and $62 billion. At the beginning of the quarter, BTC open interest was $48.04 billion, which decreased by 6.24% to $45.04 billion by June 30. Subsequently, BTC open interest slightly rebounded, reaching about $48 billion again by early August.

Figure: BTC futures open interest trends in the second quarter and since July

The decline in ETH open interest in the second quarter was significantly greater than that of BTC. At the beginning of the quarter, ETH open interest was $29.84 billion, which fell to $21.99 billion by June 30, a decline of 26.31%.

However, since the end of the quarter, ETH open interest has rebounded to $25.74 billion.

Figure: ETH futures open interest trends in the second quarter and since July

As of the end of the second quarter, the total open interest for BTC and ETH futures reached $67.07 billion, accounting for 65% of the entire futures market.

Conclusion

In our view, the second quarter of 2026 further demonstrates that after the significant drop in the futures market on October 10, 2025, leverage in the crypto market is gradually being absorbed.

The lending market is "walking down the stairs rather than falling down the elevator." The market has contracted at a relatively controllable rate for three consecutive quarters, rather than experiencing large-scale cliff-like declines in a single quarter as seen in the 2022 bear market.

Corporate treasury debt and futures open interest are also showing similar trends. The market is undergoing a controlled withdrawal rather than forced deleveraging. Meanwhile, data from early July also began to indicate that open interest and DeFi borrowing sizes may have gradually approached a bottom.

If this trend continues, even if the market contracts further, its capacity to withstand pressure may be stronger than in the previous cycle, thus avoiding a repeat of the market turmoil caused by cascading liquidations and institutional defaults at that time. As it stands, the market's deleveraging is still ongoing, but the overall process remains relatively smooth.