Crypto leverage continued to come out of the system in the second quarter of 2026, but the decline remained measured rather than disorderly, according to Galaxy Research. The firm said Q2 was the first quarter since Q4 2022 in which on-chain lending activity fell across every major segment at the same time, covering centralized finance, decentralized finance, and the crypto-backed portion of CDP stablecoins.
Galaxy contrasted the current cycle with the 2022 downturn. In Q2 2022, crypto-collateralized lending fell by more than 55% in a single quarter, then dropped another 9% and 29% in Q3 and Q4 of that year. In the current deleveraging phase, the last three quarterly declines came in at 10%, 5%, and 17%. Galaxy said that pattern points to a healthier process driven by market participants reducing risk step by step instead of large-scale forced liquidations or counterparty failures.
Total crypto-collateralized lending fell to $56.16 billion
Across the market, crypto-collateralized lending shrank by $11.33 billion in Q2, down 16.78% from the prior quarter to $56.16 billion. That left the market 40.13% below the Q3 2025 peak of $78.69 billion.
Galaxy broke the market into three segments: CeFi lending platforms, DeFi lending applications, and the crypto-backed share of CDP stablecoins. At the end of the quarter, the market split was:
- DeFi lending applications: 36.37%, down 544 basis points quarter over quarter
- CeFi lending platforms: 40.93%, up 324 basis points
- Crypto-backed CDP stablecoins: 22.7%, up 220 basis points
When DeFi lending and CDP stablecoins are combined as the on-chain lending market, the category accounted for 59.07% of total market share, down 324 basis points from the previous quarter.
CeFi loan books declined, but remained larger than DeFi by quarter-end
Galaxy estimated CeFi outstanding loans at $22.98 billion as of June 30, a quarterly decline of $2.45 billion, or 9.62%. Even after that pullback, the sector was still up $16.14 billion, or 235.94%, from the bear-market low of $6.8 billion in Q4 2023. It remained 37.16% below the historical high of $36.58 billion recorded in Q1 2022.
The quarter’s contraction in CeFi lending was driven mainly by a decline in Tether’s secured outstanding loans. At the same time, Galaxy, Coinbase, Ledn, Arch, Sygnum, and Milo all posted growth in loan book size during the quarter.
Tether remained the dominant CeFi lender with a 58.54% market share, though that was down 371 basis points from the prior quarter. Maple ranked next with 8.91%, up 52 basis points, and Nexo followed with 7.51%, up 49 basis points. Together, those three firms accounted for 74.96% of the market tracked in Galaxy’s sample, down 270 basis points quarter over quarter.
Galaxy noted that direct comparisons across CeFi lenders need caution. Some lenders only accept specific collateral types, such as BTC-only collateral or selected altcoins. Others issue fiat rather than stablecoin loans, target either institutions or retail users, or face jurisdictional restrictions. Those differences affect addressable demand and growth potential across platforms.
The report also said CeFi data remains much harder to gather than DeFi or on-chain CeFi data, which can usually be pulled from public blockchain records. Accounting standards for outstanding loans differ across CeFi firms, and disclosure frequency is inconsistent. Galaxy added that figures supplied by private third-party sources in its dataset have not yet been formally verified by the research team.
DeFi lending posted a third straight quarterly decline
Dollar-denominated outstanding loans on DeFi lending apps fell for a third consecutive quarter, dropping by $7.79 billion, or 27.61%, to $20.43 billion in Q2.
Combined with CeFi platforms, total outstanding crypto-collateralized borrowing in those two categories stood at $43.41 billion at the end of the quarter, down $10.24 billion, or 19.08%, from the previous quarter. Galaxy said the contraction was driven mainly by on-chain borrowing. It also marked the first time since Q3 2023 that CeFi outstanding loans exceeded DeFi lending applications.
The firm flagged the possibility of double counting between CeFi balance sheets and DeFi borrowing figures. A CeFi lender can use a DeFi protocol to source on-chain capital and then relend those funds to an off-chain borrower. In Galaxy’s example, a CeFi firm could post idle BTC as collateral, borrow USDC on-chain, and then lend that USDC to an off-chain client. In that case, the position may appear both in DeFi outstanding borrow data and on the CeFi lender’s financial statements. Because disclosures are limited and wallet identities are often unclear, removing those overlaps is difficult.
DeFi contracted faster than CeFi in the quarter, erasing its former scale advantage. By the end of Q2 2026, DeFi lending applications represented 47.05% of the combined CeFi-DeFi market, down 555 basis points from 52.6% at the end of Q1.
Crypto-backed CDP stablecoins also moved lower
The third segment, the crypto-backed portion of CDP stablecoins, declined by $1.09 billion in Q2, or 7.86% quarter over quarter. Galaxy said this category can also overlap with CeFi activity because some CeFi lenders mint CDP stablecoins to raise capital before lending it to off-chain clients.
DeFi balances are down more than 53% from the peak
Galaxy said outstanding DeFi lending hit an all-time high of $47.13 billion on Sept. 19, 2025. By July 21, 2026, that figure had fallen to $21.94 billion, a decline of $25.19 billion, or 53.45%, from the peak.
The report said the drawdown intensified from the end of Q1 2026, though recent data showed signs of a modest easing in the pace of decline.
Borrowing costs diverged across stablecoins, BTC, and ETH
Stablecoin rates moved higher
Using a 7-day moving average, Galaxy said the weighted average stablecoin borrowing rate rose by 27 basis points between March 31 and June 30. After the quarter ended, the rate continued to climb to 3.88%.
The measure combines borrowing costs on lending protocols and minting fees for CDP stablecoins, weighted by outstanding borrow balances. Galaxy separated those costs into two buckets: borrowing stablecoins through lending protocols and minting CDP stablecoins against crypto collateral. The paths were broadly aligned, though CDP minting rates were less volatile because they are set manually on a periodic basis instead of adjusting in real time with the market. For more than 21 months, both rate series have used the U.S. federal funds rate as a floor.
In over-the-counter markets, the benchmark USDC lending rate stayed in a 4.25% to 5% range during Q2. It ended the quarter at 4.25% and remained there through Aug. 3. OTC USDT borrowing rates also traded in a 4.25% to 5% band.
BTC borrowing stayed cheap on-chain
Galaxy’s charts on wrapped bitcoin, or WBTC, showed that on-chain BTC borrowing demand remained limited because WBTC is used primarily as collateral rather than as a borrowed asset. That kept rates low and stable. During Q2, on-chain BTC borrowing rates ranged from 0.44% to 0.5%.
The gap between on-chain and OTC BTC borrowing rates persisted through the quarter. Galaxy attributed OTC demand mainly to two uses: borrowing BTC to short it and posting BTC as collateral to raise stablecoins or fiat. The short-selling component is far less common in on-chain lending markets, which helps explain the spread. OTC BTC lending rates held steady at 1% in Q2.
ETH borrowing remained tied to staking economics
Galaxy also examined weighted borrowing rates for ETH and stETH across major lending markets and chains. Historically, ETH borrowing costs have exceeded stETH rates because demand to borrow ETH is stronger. Users often borrow ETH as part of looped leverage strategies tied to staking yield: they post stETH as collateral, borrow ETH, restake it, and repeat.
Under normal market conditions, ETH borrowing costs tend to trade within roughly 50 basis points of Ethereum staking APY. If borrowing costs move above staking yield, the strategy stops making economic sense, which limits how long ETH lending APY can stay above staking returns. Like WBTC, stETH is used mostly as collateral, so borrowing stETH is typically cheap.
The report added that users can pledge liquid staking tokens, or LSTs, and liquid restaking tokens, or LRTs, as collateral and borrow ETH at very low, or even negative, net funding cost. That enables a classic loop: deposit an LST or LRT, borrow unstaked ETH, stake the ETH, receive a new LST or LRT, and borrow more ETH again. The strategy works as long as ETH borrowing costs remain below the staking APY earned by the LST or LRT. Galaxy said that condition has generally held outside a few unusual periods.
Off-chain ETH borrowing rates, like BTC, were generally higher than on-chain rates. Galaxy gave two reasons. First, demand from short sellers is much more visible in off-chain markets than on-chain. Second, Ethereum staking yield acts as a floor in OTC markets because lenders are usually unwilling to lend ETH below the return they could earn by staking it. On-chain, by contrast, staking yield often acts more like an upper bound on ETH lending rates.
Aave V3 core market: fewer e-mode positions, but nearly half the debt
Galaxy carried out a filtered analysis of the Aave V3 core instance, which it described as the largest on-chain lending market. The sample used an Aug. 7, 2026 snapshot and applied three filters.
- A minimum debt threshold of $100, which removed dust positions but biased the dataset toward larger loans.
- A health factor reporting cap of HF ≤ 50. Positions above that level were excluded from the core analysis. Debt-weighted health factor averages and percentile statistics only included positions with 1 ≤ HF ≤ 50. Positions with HF below 1 were excluded from those health-factor calculations, while HF equal to 1 was included.
- A debt-to-equity, or D/E, filter that included only positions with positive net equity, meaning collateral value exceeded debt. Any loan without positive net equity was excluded from the D/E distribution and debt-weighted average D/E calculation, even if debt was greater than $100.
Galaxy defined health factor as total collateral value multiplied by the weighted average liquidation threshold, divided by total borrowed value. Higher health factors indicate safer positions, while a value below 1 means the position has entered liquidation territory.
After applying the filters, Galaxy counted 19,073 valid open loans. Efficiency mode, or e-mode, represented only 8.91% of total positions by count, yet accounted for nearly half of outstanding debt, with e-mode and standard mode each close to a 50% share. In Galaxy’s previous tally on April 22, e-mode debt was closer to a 60:40 split, and the shift came from a decline in e-mode balances.
Risk metrics showed a sharp difference between the two groups. E-mode borrowers carried much higher leverage, with debt-weighted loan-to-value around 90%, a debt-weighted health factor of just 1.06, and debt-to-equity around 10.7. That leaves many positions vulnerable to even modest collateral price shocks. Standard-mode loans had a much thicker cushion, with debt-weighted LTV near 49%, health factor at 1.79, and D/E around 1.07. Galaxy said those positions are more typical in use cases where borrowed assets and collateral are not closely linked in price, such as borrowing USDC against cbBTC collateral.
The report gave the debt-weighted D/E formula as D/E = Σi (Di × (Di ÷ (Ci − Di))) ÷ Σi Di, where Di is debt for a single position and Ci is collateral value. The measure only applies to positions where Ci exceeds Di.
Collateral on Aave remained concentrated in ETH-linked assets
On the collateral side of Aave V3 core, ETH-linked assets dominated. WETH accounted for about 24% of all collateral value, weETH for 16%, and wstETH for 14%, bringing those three assets to 54.6% combined. WBTC contributed about 14%. In practice, a handful of assets carried most of the collateral base, with the rest made up of stablecoins and other yield-bearing tokens.
The liability side was also concentrated. WETH made up a little more than 37% of total debt, a clear sign of widespread ETH collateral looping strategies. Stablecoin borrowing was also large, with USDT at roughly 28% and USDC at around 22%. Together they represented about half of all borrowing, while other assets made up smaller shares.
Compared with Galaxy’s previous review, WETH’s share of outstanding liabilities dropped sharply from 51.1%, in line with the broader decline in e-mode debt discussed earlier.
E-mode risk was heavily tied to ETH staking and restaking wrappers
A closer look at e-mode showed the collateral base was highly concentrated in ETH staking and restaking wrappers. WeETH alone represented 42% of e-mode collateral, and when combined with rsETH and wstETH, the three assets made up 66.2% of the segment. Galaxy said that means e-mode risk is not broadly diversified across collateral types. It is largely a concentrated bet on Ethereum staking fundamentals.
On the borrowing side, WETH accounted for 73% of e-mode debt, matching the behavior of users who repeatedly borrow ETH against ETH-linked collateral. Stablecoins still had a presence, with USDT, USDe, and USDC together making up a share in the teens of total e-mode borrowing.
Galaxy also calculated debt-weighted risk metrics for e-mode positions by collateral asset. For sub-samples where 99% of collateral was concentrated in a single asset, the report estimated implied loop counts as a way to rank which assets were being used most aggressively in recursive leverage strategies. Liquid staking and restaking ETH wrappers ranked near the top. Those positions showed high debt-weighted LTVs, D/E ratios in the high single digits to the teens, and health factors only slightly above 1, pointing to dense clusters of looped ETH leverage.
The formula for implied loop count on a single position was given as N_i = ln((1 − (D/E)_i(1 − Li)) / Li) / ln(Li), where Li is the position’s LTV. Galaxy applied it only to sub-samples with single-asset collateral concentration of at least 99%, and excluded positions where L was not in the interval (0,1), where the logarithm input was non-positive, or where the result was not finite. The report only presented values for liquid staking ETH, liquid restaking ETH, yield-bearing stablecoins, and Pendle PT tokens.
Corporate debt tied to digital asset treasury strategies fell to $16.1 billion
Galaxy said it is now tracking $16.1 billion in outstanding debt used by corporations to directly purchase or add to digital asset treasury strategies. The report noted that Bloomberg’s limitations in capturing preferred equity statistics for Strategy created timing mismatches in the time series for increases in STRC shares outstanding, but said total debt still reflects the company’s actual liabilities.
Strategy completed a $1.5 billion debt repurchase in May, reducing outstanding debt among digital asset treasury companies by the same amount during the quarter. That brought debt used to support those treasury strategies back to roughly the level seen in July 2025.
Galaxy also included a table showing actual quarterly interest obligations on debt issued by digital asset treasury companies. It noted that dividends on Strategy’s STRC require board approval and must be paid from legally available funds. Any unpaid dividends accumulate and must be satisfied before distributions can be made to junior securities, which means STRC dividend timing is irregular and payment amounts may be uneven over time.
Including digital asset treasury company debt, total crypto-related outstanding debt across the industry fell 15.08% quarter over quarter. After hitting an all-time high in Q3 2025, aggregate crypto-linked debt across on-chain and off-chain channels declined for a third straight quarter, reaching $73.2 billion at the end of Q2.
Futures open interest slipped modestly in Q2, then rebounded in July
Futures open interest, including perpetual swaps, fell 3.08% quarter over quarter to $103.2 billion in Q2. Galaxy stressed that open interest is not the same as absolute leverage because some positions are hedged with spot holdings and create delta-neutral exposure. OI alone does not reveal the market’s true net leverage.
BTC futures open interest traded in a $44 billion to $62 billion range during the quarter. It started Q2 at $48.04 billion and ended June 30 at $45.04 billion, down 6.24%. By early August, it had recovered to around $48 billion.
ETH futures open interest fell more sharply. It began the quarter at $29.84 billion and dropped to $21.99 billion by June 30, a decline of 26.31%. After quarter-end, it rebounded to $25.74 billion.
At the end of Q2, BTC and ETH together accounted for $67.07 billion in open interest, or 65% of the entire futures market. By the end of July, total futures open interest had climbed back to roughly $114 billion, with BTC near $48 billion and ETH at $25.74 billion, both above their Q2 lows.
Galaxy’s view: deleveraging remains gradual, not disorderly
In its conclusion, Galaxy said the second quarter added to the evidence that crypto leverage has been unwinding gradually since the sharp futures-market drop on Oct. 10, 2025. Lending markets, in the firm’s view, are moving down in steps rather than collapsing all at once, with three straight quarters of controlled contraction instead of the cliff-like declines seen in the 2022 bear market.
Galaxy said corporate treasury debt and futures open interest show the same broad pattern: a managed pullback rather than forced deleveraging. Data from early July, it added, suggests futures open interest and DeFi borrowing may already be approaching a bottom range. If that pattern continues, the market may be better positioned to absorb further shrinkage without repeating the chain liquidations and cascading counterparty failures of the last cycle.
The original report was written by Zack Pokorny of Galaxy. The Odaily version identified Saoirse of Foresight News as the translator.