1. TL;DR

$6.04 billion of lending supply supports $5.40B of outstanding loans across Morpho’s five largest chains, with utilisation at 89.4%. Most of that growth came through apps and issuers that use Morpho as their lending backend: Coinbase for USDC loans, Robinhood Chain, PayPal’s PYUSD and Ripple’s RLUSD.

Net supplied principal rose +89% over the past year, from about $3.09b to $5.83b. Ethereum and Base account for 46.2% and 39.2% of supply.

• Six curator labels manage 86.3% of listed vault assets. Steakhouse, Sentora and Gauntlet alone manage 72.9% and earned $8.1M of the $12.4M in curator fees over the past year.
• Another $402M (6.3%) sits in unlisted vaults, mostly mandates run for integrators such as Grove, Trezor and Safe, often by the same listed curators.
• Rewards worth $49.0M accrued over 12 months, about a third of the $149.8M borrowers paid in interest. The DAO now funds a small share; issuers, chains and Ethena fund most of the rest.
• Utilisation sits at 89.4%, right at the 90% level Morpho’s rate model steers towards.
• Since inception, borrowing against sUSDS has paid a positive spread, sUSDe has roughly broken even and USDe has lost money outside recent reward campaigns. PTs are positive at market prices, but well below what oracle discounts imply.
• Midnight, the fixed-rate layer, holds $3.5M of loans two weeks after Coinbase put it in front of users.

The rest of this study follows the allocation behind those balances, the incentives supporting them and the carry economics of selected borrower cohorts.

2. How Morpho fits together

Morpho Blue markets are defined by five parameters: a loan asset, a collateral asset, an oracle, an interest rate model and a liquidation loan-to-value limit (LLTV). Each market has its own liquidity and credit exposure, so two markets lending USDC can have different risks and withdrawal conditions.

Vaults sit on top of markets. Depositors receive shares in a pooled strategy that spreads capital across several markets. Vaults V1 can only allocate to Morpho markets; Vaults V2 can allocate more widely through adapters. Vaults V2 uses adapters to connect the vault to approved lending strategies. An adapter executes the allocation and reports its value back to the vault, allowing one share price to represent several underlying positions.

Curators configure the risk framework, including permitted exposures and caps. Allocators move capital within those limits and Sentinels can reduce exposure or revoke pending changes. Vaults V2 also support timelocks, optional access gates and mechanisms for exiting into underlying positions when cash liquidity is unavailable. Those controls affect both the strategy’s flexibility and the depositor’s route out.

Morpho Midnight adds fixed-rate, fixed-term loans that can draw available liquidity from Blue markets through callbacks; its mechanics are covered in section 7.

Two points matter for the numbers that follow. Vault assets and market supply overlap because vaults supply capital to the underlying markets, so we never add the two together. Borrowers pay interest to suppliers, while curators can charge performance or management fees at the vault level. Protocol interest fees are separate from those curator fees.

In markets using AdaptiveCurve, rates respond to utilisation around a 90% target. Borrowing becomes progressively more expensive above that level, while the rate curve also adjusts over time to encourage utilisation back towards the target. A lending book near 90% utilisation is therefore consistent with the model’s design, rather than evidence of stress on its own.

3. Growth and curator footprint

Listed vault assets grew from $3.45 billion at the end of January to $5.98 billion at the end of September, and a handful of curators captured over 80% of it. Permissionless vault creation has produced a long tail of strategies, while most listed capital remains concentrated among a small group of allocation teams.

Steakhouse, Sentora and Gauntlet account for 72.9% of listed vault assets in our sample. Behind those labels sit products such as Steakhouse’s USDG vault on Ethereum, its Prime USDC and High Yield USDC vaults on Base, and Sentora’s PYUSD and RLUSD vaults.

Coinbase routes its USDC Earn product through Steakhouse-curated vaults and its borrowing product has approximately $1.66B of active loans, while Sentora curates the main PYUSD and RLUSD vaults with $716M in PYUSD and $466M in RLUSD, with further distribution through Kraken’s DeFi Earn. Morpho markets on Robinhood Chain hold $533.6M of lending supply and $479.3M of outstanding loans on 5 October.

The remaining labels in our selected set are SparkDAO, Sentora/SparkDAO jointly, and Armitage by Wintermute.

3.1 Growth increasingly sits in Vaults V2

77.8% of current listed vault assets sit in Vaults V2. Part of that is migration. Curators have moved existing depositors from V1 to V2 vaults, so V2 growth overstates how much new capital entered Morpho.

The figures include deposits, withdrawals and accrued interest, so they track the size of the curated book, not net inflows.

Steakhouse’s current balance includes $1.91b in V2 and $362m in V1. Gauntlet retains a more mixed footprint, with $559m in V1 and $373m in V2. Sentora’s assets under its standalone label are entirely in V2 in this snapshot.

Gauntlet’s tracked assets fell from $1.33B at the end of October 2025 to $772M at the end of November, against the backdrop of the Stream xUSD and Elixir deUSD crisis. Those failures became a wider confidence and liquidity shock, with independent reporting documenting emergency measures beyond directly exposed Morpho vaults. In the separate Resolv incident in March 2026, OAK Research attributed approximately 96% of the $6.2M of identified Morpho bad debt to Gauntlet vaults. These two losses have caused Gauntlet to experience slower growth compared to other curators.

3.2 The asset split reveals different businesses

$1.31b of Steakhouse’s assets are denominated in USDC, with a further $532m in USDG. Its book also includes USDT, ETH and other assets, giving it a broader mix of loan currencies than curators concentrated in a single issuer’s stablecoin.

Sentora’s standalone book is more concentrated: $716m in PYUSD and $466m in RLUSD. The jointly labelled Sentora/SparkDAO allocation adds another RLUSD position, retained separately in our accounting.

Gauntlet’s largest allocation is also USDC, at $793m, alongside smaller ETH, BTC and USDT balances. SparkDAO’s standalone vault assets are split between USDC and USDT.

A USDC vault lends USDC, but its borrowers may pledge BTC, ETH, a yield-bearing stablecoin or another asset. Assessing the curator’s credit exposure requires looking through the vault to its market allocations, oracles and liquidation limits.

3.3 Twelve months of fee capture

$12.43m of curator fees was recorded between 1 October 2025 and 30 September 2026. That counts on-chain vault fees only, before curators' costs or any offchain deals with integrators. The protocol itself earned nothing, as its fee switch is off.

Steakhouse Financial leads with $4.73m, followed by Sentora with $2.73m. Together, those two labels account for over 60% of measured fees. SparkDAO records $924,000 and Gauntlet $663,000. A further $1.35 million could not be attributed to a curator. Fee figures include unlisted vaults, the asset figures above do not.

Performance fees contribute $10.14 million, or 81.6% of the total. Management fees account for the remaining $2.29 million. The composition also differs between curators: Steakhouse’s recorded fees are performance-based, while management fees contribute $2.09 million of Sentora’s total. Their income therefore responds differently to the size and earnings of the underlying vaults.

The monthly series shows that fee capture did not rise steadily. Recorded charges fell from $1.41m in October 2025 to $583,000 in February 2026, before recovering to $1.57m in September 2026.

The listed rankings also miss capital held in vaults the main Morpho interface does not display.

4. Non-whitelisted vaults

The curator leaderboard captures the most visible part of Morpho’s vault economy. Another $402.4M sits in vaults that were not listed in the main interface at our 5 October snapshot. That is 6.3% of the $6.42B of combined listed and unlisted vault assets in our dataset.

“Non-whitelisted” needs a precise definition here. We use it to mean vaults classified as unlisted by the API, rather than vaults whose depositors require permission. Morpho’s listing policy governs what its interface displays. Separately, Vaults V2 can use gates to restrict particular actions. Listing status and access restrictions therefore need to be checked independently.

The capital outside the listed set is highly concentrated. Base holds $257.4M and Ethereum $143.1M, leaving under $2M on the other three chains. The largest two vaults are Grove x Steakhouse USDC, a V2 vault on Base with $136.0M, and Grove x Steakhouse USDC High Yield, a V1 vault on the same chain with $103.4M.

Beyond those two vaults, the snapshot includes an msETH vault with $22.9M and two Trezor Steakhouse Prime vaults holding a combined $36.3M in USDT and USDC.

Most unlisted capital is not a long tail of anonymous strategies. The largest vaults are mandates run for named integrators, Grove (part of the Sky ecosystem), Trezor, Safe, Yearn and others, and most are run by curators that also appear on the listed leaderboard. Unlisted vaults are mainly a B2B channel.

Vaults V2 account for $248.9M, or 61.9%, of unlisted assets. The historical chart shows how the currently listed and unlisted cohorts developed, but doesn’t reconstruct what the frontend displayed on each historical date.

5. DAO and incentive history

Morpho’s incentive footprint is larger than the rewards funded directly by the DAO. Over the twelve months to September 2026, the recorded Merkl campaigns distributed $49.04M of valued rewards. That figure is based on daily token prices and represents campaign accruals, not necessarily cash already claimed by users. Borrowers paid $149.83M in interest over the same period and curators earned $12.43M in fees. Rewards were equal to about a third of borrower interest and roughly four times curator fees.

Rewards include Merkl campaigns and claims from Morpho’s older Universal Rewards Distributor, valued at each token’s price on the day.

The history also shows how quickly Morpho’s incentive system changed. MIP 86 set the reward framework in late 2024. MIP 92 cut general Ethereum and Base reward rates by 30% from 30 January 2025, and MIP 95 proposed a further 25% cut, taking annual rewards from about 19.89M to 14.92M MORPHO. Together the two cuts came to 47.5%, before asset-specific changes. MIP 111 then moved rewards to Merkl with a starting budget of 624,000 MORPHO a month, and Morpho completed the switch by September 2025.

The DAO has since stepped back. Its MORPHO rewards fell from about $2.6M a month to between $0.03M and $0.32M a month after May 2026, and made up about 3.6% of recorded rewards in September 2026. Most incentives now come from third parties paying to grow demand for their own assets or chains.

The recorded rewards are concentrated among a small number of campaign creators and programme operators:

• DAO $12.64M
• Stablecoin issuers (mostly Sentora’s PYUSD and RLUSD) $14.27M
• Chain programmes (Arbitrum DRIP $6.76M, Monad Foundation $3.72M)
• Ethena $2.68M
• Sky $1.72M
• Morpho-labelled $2.28M

6. Utilisation dynamics

Utilisation is high across the board. Through 4 October 2026, stablecoin and yield-bearing stablecoin markets held approximately $5.69 billion of lending supply and $5.10B of outstanding loans, implying utilisation of 89.6%. Volatile-asset markets are smaller, with around $329M of supply and $284M of loans, but were also highly utilised at 86.4%.

That is by design. Morpho’s rate model targets 90% utilisation: rates rise sharply above it and keep adjusting until borrowing falls back, so most markets settle close to the target. The risk is the remaining 10% buffer, not the level itself.

The utilisation bands show where the pressure is concentrated. Approximately $3.42B of supply sat in markets running between 90% and 95% utilisation, making this the largest band. A further $2.30B was in the 80% to 90% range. Only around $101.8M was below 80%, which means that relatively little of the lending book operated with a wide unused borrowing buffer.

Approximately $192.5M in supplied assets was held in markets with utilisation of at least 99.9%, almost entirely concentrated in the cirBTC/USDC market on Arc. This band therefore reflects near-full utilisation rather than necessarily zero available liquidity, as small residual balances may remain even when utilisation rounds to 100%.

However, high utilisation did not immediately translate into high borrowing rates. The market’s AdaptiveCurveIRM rate at target had declined from an initial 4% in June to approximately its 0.10% APR floor before gradually increasing to around 0.31% by October 5. As a result, borrowing rates were adjusting upwards from a very low starting point despite persistently high utilisation.

The chain-level picture reinforces the same point. Ethereum, Base and Robinhood carried the largest absolute amounts of supply and borrowing, with utilisation close to 88% to 90%. Arc was a much smaller market but was effectively fully utilised at the latest snapshot, while Monad operated at a lower, though still elevated, level of approximately 86%.

Morpho’s TVL overstates the liquidity available for immediate withdrawal. A large part of the supply is actively deployed, and much of it sits in markets where only a limited buffer remains before utilisation becomes restrictive. That makes the lending book productive, but also more sensitive to withdrawals, collateral moves and sudden changes in borrowing demand.

7. The carry trade

High utilisation tells us that Morpho’s lending supply is being used. The carry study asks what borrowers receive in exchange for paying to use it. For a borrower posting sUSDS, sUSDe or a PT as collateral, the trade depends on the yield earned by that asset relative to the cost of the stablecoin loan. Incentives can improve the spread, and additional borrowing can magnify its effect on equity.

Across 30 markets since inception, borrowing stablecoins against sUSDS earned a positive spread net of borrowing costs, sUSDe roughly broke even, and USDe lost money outside recent reward campaigns. PTs were positive at market prices. sUSDS has offered relatively consistent positive carry, sUSDe has moved between positive and negative conditions, and USDe depends on external rewards. For PTs, the yield measure also matters: the discount configured in the collateral oracle can differ substantially from the token’s market discount.

7.1 Yield-bearing stablecoins

For sUSDS and sUSDe, we measure intrinsic yield from daily growth in each token’s exchange rate. The daily spread adds eligible collateral and borrower rewards, then subtracts borrowing costs. USDe has no native yield in this model, so its eligible collateral incentives provide the income against which financing costs are measured.

We follow each selected market from its creation through 4 October 2026. Borrowing costs come from interest accrual history, and the asset charts combine markets using their outstanding debt. This weighting matters because a favourable rate in a small market has less economic significance than the rate paid on a large borrowing position.

sUSDS has the most consistent positive spread of the two yield-bearing stablecoins, and borrowers against it earned about $603k in total net of borrowing costs. Across its four selected markets, the time-weighted spread was approximately +0.90 points, while the spread weighted by debt exposure was +0.57 points. The daily asset series was positive on approximately 84% of observations.

At the snapshot, sUSDS earned approximately 3.60% intrinsically against a debt-weighted borrowing cost of 3.08%, leaving a spread of about +0.52 points.

Over its full history, sUSDe carry roughly broke even: −0.20 pp time-weighted, +0.28 pp weighted by debt, positive on 52% of days, and about +$545k cumulatively. At the snapshot, sUSDe’s intrinsic yield was approximately 5.05%, compared with a debt-weighted borrowing cost of 4.25%, giving a current spread of approximately +0.81 points. That aggregate masks differences between markets: the collateral yield is shared, but borrowing costs differ by loan asset, utilisation and market configuration.

USDe carry has mostly been a cost of farming Ethena points. Across six markets, the spread weighted by debt was about −5.93 points and positive on only 11% of days, roughly −$13.6M cumulatively. Most of that came from 2024 markets, where borrowers were paid in Ethena points that this model does not value. Across its six selected markets, the historical spread including observed rewards was approximately −5.93 points, weighted by debt exposure. Only around 11% of daily observations were positive. More recently, eligible USDe collateral campaigns have made some markets positive. The aggregate snapshot combined approximately 4.75% of collateral rewards with 4.15% of borrowing costs, leaving about +0.60 points.

The model includes Merkl rewards only. Earlier incentives, including Ethena points, are excluded, so the historical USDe figure understates what borrowers received.

7.2 PTs: oracle assumptions against market prices

A Pendle principal token (PT) trades at a discount and redeems at full value on its maturity date, so the discount is its fixed yield.

The PT comparison exposes a large difference between collateral valuation and the return available to a buyer. The first scenario uses each market’s configured annual oracle discount as its yield assumption. The second uses Pendle’s daily market-implied fixed APY, derived from the PT’s quoted discount to redemption in its accounting asset. Both subtract the same Morpho borrowing costs and include the same eligible rewards.

For PT-USD3, the oracle uses a 30% annual discount parameter. Subtracting the historical debt-weighted borrowing cost of approximately 7.75% produces an oracle scenario spread of +22.25 points. Using the daily market-implied yield instead reduces the historical spread to +5.84 points.

The market-price result remains positive, but is substantially smaller. The oracle parameter determines a conservative collateral valuation path; it does not establish that a borrower could buy the PT at a price offering 30% annually.

The same distinction appears in the other PTs. PT-sUSDS falls from +8.07 points under the oracle scenario to +0.27 points using market-implied yields, while PT-sUSDe falls from +2.35 points to +1.37 points. PT-reUSD moves in the opposite direction, from −2.82 points to +2.12 points, but its positive modelled spread did not protect leveraged borrowers from liquidation.

On 25 August 2026, approximately $36.1M of loans were liquidated within fourteen minutes across two Ethereum markets using PT-reUSD-10DEC2026 as collateral. Selling into the Pendle pool moved the price read by the collateral oracle, triggering liquidations of highly leveraged positions. No bad debt in either market was accrued as suppliers were protected, but liquidated borrowers still bore losses.

These comparisons use matched daily quotes for all PT market-days in the selected history. Pendle’s observations are sampled at 00:00 UTC, alongside the existing daily borrowing convention. The resulting spread describes the annualised financing opportunity implied by that day’s quote. It does not track a particular borrower’s entry price, holding period, execution costs or realised sale proceeds.

Integrating daily spreads against debt exposure produces a useful model statistic, but does not reconstruct the profits earned by individual loops.

7.3 Carry by LTV

A positive spread becomes more consequential as a borrower leverages. The equity-return model uses the collateral yield and eligible collateral rewards on the full collateral position, then deducts borrowing costs net of borrower rewards on the debt.

At 50% LTV, collateral exposure is twice the borrower’s equity. At 80%, it is five times equity. If the collateral earns more than financing costs, increasing LTV raises modelled equity returns. At 90% LTV, collateral is ten times equity, and that is where most debt sits. At the 5 October snapshot, a USDe/USDC loop on Base at 90% LTV modelled about 10.4% on equity including rewards, and stays positive while borrowing costs remain below about 5.3%. sUSDS/USDT earned about 10.3% at 92.4% LTV and 15.7% at 95.7%.

The position histogram shows where borrowers actually sit along that curve. LTV is calculated using the market oracle before positions are grouped into buckets. A concentration near LLTV leaves less room for interest accrual or adverse collateral valuation changes, even when the current carry spread is positive.

8. Midnight and fixed rates

The carry study shows how a financing margin can narrow when borrowing rates rise. Midnight adds the option for borrowers and lenders to agree to a fixed rate and a defined maturity when a loan is executed. That gives borrowers certainty over financing costs for the agreed term, although collateral risk and changes in the asset’s yield remain.

Midnight is still small beside Morpho’s variable-rate markets. DefiLlama showed approximately $3.51 million of active loans when checked on 7 October. This measures outstanding debt, keeping the comparison consistent with the borrowing figures used elsewhere in this article.

Its connection to the existing lending book is the more interesting mechanism. Lenders can publish signed offers while their funds continue earning a variable rate in Morpho Blue. When an offer is taken, a callback withdraws the required liquidity and funds the fixed-term loan in the same transaction. The fill depends on that liquidity being available; signing an offer does not reserve it.

Coinbase’s launch announcement on 22 September 2026 brought this structure into its borrowing interface, allowing users to borrow USDC against bitcoin at a fixed rate and repayment date, the same distribution route that grew the variable lending book.

For carry borrowers, fixed financing removes one moving part. Whether a trade remains profitable still depends on collateral income, incentives and the ability to repay or refinance at maturity.

9. What’s next

Three developments will shape how this lending book evolves.

The Midnight vault adapter could allow curators to allocate pooled capital into fixed-term credit. Morpho identified it as a route to deeper liquidity at launch; the next question is how much capital vaults actually deploy through it.

The fee switch would change who captures the economics. Governance can retain up to 25% of borrower interest in Morpho Blue. Any activation would make the division between lender returns, curator fees and protocol revenue more consequential.

Incentive funding is the third. The DAO now funds a small share of rewards, so demand for USDe, PYUSD and RLUSD loans depends on issuers and chains continuing to pay. The OTC-funded Robinhood Chain programme is the first test of how much of that demand stays when campaigns end.

Morpho reached $5.4B of loans by letting curators and apps package its markets. Whether that holds depends on carry staying positive once subsidies shrink.