Aave, Morpho, and Compound: The Margin That Isn't on the Table

Between July 14 and August 12, 2026, borrowers on Aave V3, Morpho Blue, and Compound V3 paid $48.26 million, yet only 3.99 ended up in a protocol's treasury. The three charge in structurally different ways. Aave retains a percentage set by governance for each reserve, averaging 13.3%. Compound accumulates the spread between what it charges and what it pays as reserves within each contract—7.6%—and these reserves serve as a loss buffer before they are considered profit. Morpho retains exactly zero: its fee switch, a contract parameter capped at 25% of the borrower's interest, has never been activated, and the margin is taken by vault curators, third parties that do not appear in its revenue. This article breaks down the three mechanisms using daily series, calculates the rate each borrower actually pays, and puts a number on the third pocket: the curators.

How much did Aave, Morpho, and Compound charge, and how much did each keep?

Each reports three amounts per day: fees, everything the borrower pays; revenue, the portion retained by the protocol; and the supply-side share, what goes to those who provided the capital. Revenue divided by fees is the take rate.

Of the $48.26 million charged, 44.27 went to the supply side: 91.7%. The order by size is undisputed; what disrupts it is the destination of the money. Morpho charges nearly ten times more than Compound yet retains less than it does, and over twelve months, the gap remains the same: Aave V3 accumulated $857.35 million in fees and $110.95 retained, a 12.94% rate consistent with the short window; Morpho Blue has $222.39 million in fees and zero retained, as has been the case throughout its history; Compound V3 shows $29.37 million and $271,000, a 0.92% rate that only measures since its adapter began calculating the margin.

Why an aligned window and not DeFiLlama's standard thirty days?

Because the DeFiLlama adapters for Aave, Morpho, and Compound do not close the day at the same time, and a window counted backward from today punishes laggards with empty days. On August 12, all three had data for that day, so the aligned window is clean; two days prior it was not, and Aave lagged three days behind, each uncounted. That flaw, extended across the thirteen largest credit markets, is what reconstructs the on-chain credit ranking by who keeps the margin.

There is a second reason to date the window, and this one can be verified. That ranking left a dated prediction: that starting August 12, with the liquidation peak of July 12 falling out of the thirty-day window, the ratio between Morpho's fees and Aave's would drop to a range of 55% to 62%. Morpho's series moved between 0.48 and 0.66 million daily throughout most of July except for three days—the 7th (2.647 million), the 8th (2.706), and the 12th (3.620)—and that last day alone accounted for 17% of the window the ranking measured. Today the peak has exited: over the aligned window, Morpho charges 17.425 million compared to Aave's 29.065, a 60.0% ratio. The prediction holds, and in the process, the asterisk that forced Morpho's rate to be published with and without that day two days ago disappears.

How does Aave retain its 13.3%?

Aave does not have a single dial, but rather a reserve factor per asset and per market, set by governance, which determines what portion of the interest is diverted to the Aave Collector instead of going to the depositor. The 13.3% aggregate is the sum of dozens of these parameters weighted by volume, and it moves when the portfolio composition changes even if no one touches a thing. The breakdown by chain shows this with a three-to-one range within the same protocol.

Ethereum accounts for 78.2% of fees and 81.3% of retained revenue, so its 13.8% drags the entire average. Polygon retains 27.0% on a flow a hundred times smaller, and Plasma—the fastest-growing deployment, now second by fees—keeps 8.7%: a new destination with aggressive parameters lowers the protocol average as it grows, without anyone touching the legacy chains.

Aave's fees are also not just interest: its adapter also counts flash loans—uncollateralized and returned within the same transaction—liquidation penalties, the portion of swap fees coming from Paraswap, and MEV recapture via Chainlink SVR (Smart Value Recapture) on Ethereum liquidations, 100% of which has gone to the Aave Collector since April 2025. Added to this is GHO: its stablecoin is minted against collateral, there is no depositor to share with, and the interest goes entirely to the treasury. Every dollar of GHO growth raises the aggregate take rate without any reserve factor changing.

Why Morpho's zero doesn't mean no one is keeping the margin?

Morpho Blue's zero is literal and intentional. The contract includes a fee switch capped at 25% of the borrower's interest; governance can activate it but has never done so. The reason has been on record since a February 2025 proposal and is not competitive: the legal and tax work for activation is not finished, and remains unfinished eighteen months later. The extreme case came with the project's first commercial license—a low seven-figure sum paid by Berachain to use Morpho Blue's code—which governance voted to send to the Morpho Association, a French non-profit, rather than the DAO treasury. We detailed this in the analysis of Morpho following its unicorn valuation.

That the protocol does not retain does not mean the lender keeps 100% of the interest. In Morpho, the work that governance does in Aave—what collateral is accepted, with which oracle, at what liquidation ratio, and with what ceiling—is done by a curator, a third party that sets up a vault and charges to manage it. Morpho Vaults V2 set two fees and their caps: a performance fee of up to 50% of the interest generated by the vault and a management fee of up to 5% annually on total assets. These are set by the curator, collected by minting new shares in their favor, and none appear in protocol revenue because it is not protocol money.

This third pocket is measurable, and DeFiLlama tracks it in a category separate from the protocol. DeFiLlama maintains a distinct category, Risk Curators, with 39 entries that between July 14 and August 12, 2026, totaled $31.394 million in yield generated in the vaults they manage, of which curators retained 2.733 million: a take rate of 8.7%.

The limitation of this table must be stated: the category does not publish a breakdown by protocol, so it is impossible to separate what portion of those 2.733 million comes from vaults on Morpho versus Euler or other hosts. That total curated yield exceeds Morpho Blue's fees in the same window—31.39 million versus 17.43—confirms that curators operate in multiple venues. What is established is the rate: a curator keeps an average of 8.7% of what they manage, more than Compound V3's 7.6% and less than Aave V3's 13.3%, and very unevenly among them, from Gauntlet's 1.3% to Sentora's 11.4%.

For those depositing in Morpho, the "0% protocol fee" describes an accounting line, not a net yield. The margin hasn't disappeared: instead of being set by a public vote, it is set by the curator of the specific vault. The risk this introduces is covered in the DeFi vault risk taxonomy, and what happens when the choice goes wrong, in the Resolv USR episode.

What does Compound retain and why is its one-year take rate useless?

Compound V3 does not divert a portion of interest to a treasury: it accumulates the spread between what it charges and what it pays as reserves within each Comet contract, market by market. Compared to Aave, the difference lies in the nature of the money, not the percentage. These reserves are a first-loss buffer: when an account falls below the threshold, absorption transfers its debt to the protocol and consumes reserves to settle it. Below the target set by governance, anyone can buy the seized collateral at a discount until they are replenished; once the target is reached, that purchase is no longer possible and only then, by vote, can the DAO withdraw reserves. Compound's 7.6% is margin trapped in a contract that may have to spend it on bad debt before anyone votes on it; Aave's is in an available treasury. In our liquidations explainer, we discuss why sometimes this reserve is not enough.

There is also a measurement detail that invalidates any historical comparison, and it is worth explaining precisely because it is easily confused with a governance decision. The Compound V3 fee series has the full 548 days, all with values; the revenue series has only 79 days of data, from May 26 to August 12, 2026, and zero for the previous four hundred and sixty-nine. The four hundred and sixty-nine zeros are an artifact of the code, not a contract reading: the adapter returned a fixed zero based on a note assuming Compound did not take interest to treasury, until a commit on May 26, 2026 changed it to the actual margin calculation, market by market. The 0.92% annual rate therefore comes from dividing seventy-nine days of margin by three hundred and sixty-five days of interest, measuring only the adapter's reach. The only usable number is the 7.6% from the short window, and even that is an upper bound: the code counts as zero, rather than negative, markets that pay interest out of those same reserves.

What interest rate does the borrower actually pay in Aave, Morpho, and Compound?

Dividing annualized fees by outstanding debt yields the implicit average rate borne by the collective of borrowers, the figure comparable to that of a bank. Supplied capital is TVL plus debt, because DeFiLlama counts deposited funds that are not lent out as TVL.

The average Morpho Blue borrower pays 4.78% compared to 3.17% for Aave V3: 51% more. Behind this is a different portfolio: a large part of Aave's eleven billion in outstanding debt consists of leverage operations on liquid staking derivatives at very low rates. Morpho, with isolated markets and specialized vaults, accepts collateral that no conservative parameter would allow and charges accordingly. What makes it attractive for the lender is what makes it expensive for the borrower.

The implicit net lender yield averages all deposited capital, including idle funds, so it falls below what an interface might advertise for a specific market with high utilization. What it measures well is overall efficiency: with 44.0% utilization, Aave makes its capital work harder than Morpho (35.7%) and Compound (33.0%) and still charges less. The Morpho lender theoretically receives 100% of the interest; applying the 8.7% average curator fee, that theoretical 1.71% drops to around 1.56%.

What changes for the depositor and for the AAVE, MORPHO, and COMP holder?

For every $100 a borrower pays, the Compound V3 depositor receives $92.36; the Aave V3 depositor, $86.74; and the Morpho Blue depositor, $100 minus whatever their curator decides. That isolated ranking suggests Compound is the best place to deposit, but it isn't: the distribution is one of three variables and not the largest. The second is the gross rate being distributed, because 92.36% of 3.87% yields less than 86.74% of a higher rate. The third is what that margin protects: in Compound, it is reserved to cover bad debt before being distributed; in Aave, it funds a balance sheet that in April 2026 was used to absorb the loss from the rsETH bridge incident, about 25,000 ETH; in Morpho, it doesn't exist, and the one who absorbs the loss is the depositor of the affected vault. A low take rate has a deferred cost: the absence of insurance that is only paid for when needed. The details of each architecture are in our technical comparison, and the definitions of TVL, fees, and revenue are in the DeFi metrics guide.

For the token holder, the reading is different. AAVE is the only one whose protocol has converted margin into token buybacks: $42.94 million since the program started on April 9, 2025, of which 25.98 were in the last twelve months. The detail lost in annual aggregates is that the AAVE buyback series has been at zero since the last purchase on June 24, 2026, for $576,537. The monthly pace was around 3.89 million in August 2025 and dropped to $413,512 in February 2026; the spike to 6.16 million in March was a single purchase on the 3rd. The explanation is public and dated: the DAO paused buybacks on April 19, 2026, the day after the rsETH bridge incident, and formalized the pause in a proposal on April 22 to preserve balance sheet capacity while the loss allocation was unclear; in March, it had already cut the annual budget from 50 to 30 million. After that date, the adapter only records four residual entries—$20,188 on April 30, $135,368 on May 26, eleven dollars on June 16, and $576,537 on June 24—and since that day, the series has not moved. The protocol continues to retain 3.85 million every thirty days: what has been interrupted is the portion reaching the holder, as told in the piece on Aave Will Win.

MORPHO is the clean case: zero retained, zero distributed, and a quantifiable latent option. On an annual flow of $222.39 million, activating Morpho's fee switch at the 25% cap would divert up to $55.6 million per year to the protocol. Since the declared blockage is legal and tax-related, the signal to watch for is a formal governance proposal. COMP, on the other hand, is the hardest to read from a dashboard: its reserves are first-loss capital and only reach the treasury if governance withdraws them, and only above the target. The honest comparison: Compound retains little and in a place from which it is difficult to extract. How the three stand against other tokens with active distribution is in our comparison of fee switches and buybacks.

What signals would indicate that the margin distribution has changed?

Between August 2025 and August 2026, distribution was moved by risk and paperwork: Aave stopped distributing due to an exploit and Morpho hasn't started due to a tax file, and neither has anything to do with attracting depositors. The only figure that changed by design is that of Aave V4, which has retained margin since its first month live. These are the four observations that would break the current picture.

• A formal proposal on the Morpho forum to activate the switch. The cap is 25% of the borrower's interest and the magnitude, based on current flow, is around $55 million annually.
• Non-zero values in the AAVE buyback series. Paused since April 19, 2026, and with no record since June 24.
• A Compound vote to withdraw reserves to the treasury. This is the only act that converts the 7.6% retained into available money.
• Aave's aggregate take rate falling outside the 12% to 15% range. With Ethereum at 13.8% and Plasma at 8.7%, the average moves on its own; moving outside this range would indicate a change in parameters rather than mix.

Before comparing two lending markets by their take rate, it is worth making three checks: what the adapter measures, since when it has measured it, and if the missing margin is held by a third party outside the table. In these three protocols, all three have changed the result: they have turned 0.92% into 7.6%, explained a zero that looked like an accounting error, and put a figure on $2.733 million in curator fees tracked outside the Morpho profile.

Sources and links: DeFiLlama — Aave V3 daily fee series (Aug-12-2026 query) · DeFiLlama — Aave V3 profile and methodology · DeFiLlama — Morpho Blue profile and methodology · DeFiLlama — Compound V3 profile and methodology · DeFiLlama — Risk Curators category · DefiLlama — May-26-2026 commit starting Compound V3 margin calculation · Morpho Docs — curator performance and management fees · Morpho Docs — governance organization · Compound III Docs — reserves, absorption, and collateral purchase · Aave Governance — ARFC to pause buybacks (Apr-22-2026) · Morpho Forum — MIP-94, Berachain license destination · Aave Docs — the AAVE token and revenue destination