A yield gap inside Morpho’s stablecoin markets is opening up even when the same curator sits behind the products being compared.
In an analysis published by MarsBit and written by Daii, Steakhouse Financial’s most conservative stablecoin vault, Steakhouse USDC, delivered a 7-day average APY of 3.96%. A higher-risk vault run by the same curator, Steakhouse High Yield USDC, posted 5.09% over the same period. The single market receiving nearly $10 million from that vault was generating an implied 7-day average supply yield of about 7.7%. That puts the spread between the conservative vault and the direct pool route at roughly 3.74 percentage points.
Where the extra yield comes from
The piece states plainly that the return is not being produced by token emissions, points programs, or airdrop expectations. It is borrower-paid interest. In the author’s framing, that makes it one of the cleaner forms of DeFi yield, but also means it is not free money.
The article breaks the stablecoin lender’s choices on permissionless credit rails into three paths.
The first is the default route: depositing into a conservative curated vault managed by a professional team. That team selects markets, spreads exposure, monitors conditions, and can move positions if something goes wrong. The depositor receives blended yield, reduced by performance fees and by the drag created from keeping some idle liquidity available for withdrawals.
The second path is less visible. The same curator may also operate a higher-risk vault that is still monitored and diversified, but allowed to allocate across a wider set of markets. In the Morpho vault list, searching for Steakhouse shows the same curator name across products while APY ranges from 15.74% down to 3.71%, which the article says reflects differences in what each vault is permitted to hold rather than differences in manager skill.
The third path is direct supply into one specific market. There is no intermediary layer, no performance fee, and no cash drag. The depositor takes whatever interest the market is paying. The catch, according to the article, is that Morpho’s default interface does not guide users toward that route even though the protocol itself leaves it fully open.
The author’s core point is that the spread is not evidence that someone is skimming hidden profits. It is the price paid for delegating decision-making and offloading the portion of collateral risk that conservative vaults do not want to bear.
Displayed yield is not the same as realized yield
The article spends considerable time on rate presentation because it considers this one of the easiest places for users to misread what they can actually earn.
On Morpho’s interface, the same market at the same moment can show three different figures side by side: an instantaneous rate, a 24-hour average, and a 7-day average. For the market used in the example, the borrowing side was showing 18.25% instantaneously, 10.51% on a 24-hour average basis, and 8.19% on a 7-day average basis. The article uses only the 7-day average throughout, arguing that it is the cleanest figure to verify after the fact.
The reason these numbers can differ so sharply lies in utilization-based rate mechanics. As utilization rises, rates rise, and once usage moves beyond the target line they increase steeply. But when a lender supplies fresh capital to the market, total supply expands immediately, utilization falls, and the rate drops with it. In other words, the attractive figure displayed on the list belongs to the market before the depositor enters. The larger the deposit, the stronger the downward impact.
The author describes a common trap: a market can show a four-digit annualized rate while offering less than $1 in withdrawable liquidity. The sign is not technically false, but it was never relevant to the next sizable depositor.
That is why, in the article’s method, users should model their own deposit size against the interest-rate curve rather than anchor on the posted yield.
How the spread could close
The analysis says this kind of structural arbitrage has no hard expiry date, but it will get competed away. It points to three likely channels.
- First, crowding. More direct suppliers dilute utilization and pull yields on high-rate markets closer to curated-vault levels.
- Second, productization. If Morpho’s default interface adds a direct entry point, or if outside tools reduce the friction, much of the spread could disappear quickly.
- Third, fee compression. Custodians or managers may cut performance fees to retain assets, which would erase part of the service spread.
The author says the proper review cadence is weekly, not daily. The bigger danger is not that the spread vanishes tomorrow, but that users continue to wear the same risks after the premium has already been compressed away.
The three risks the article says users must accept first
The first is the loss of diversification and active oversight. Once a depositor leaves the curated route and enters a single market directly, no professional team is handling routine monitoring, diversification, or emergency reallocation.
The second is exit risk. In the market highlighted by the article, total size stood at about $84.31 million, while immediately withdrawable liquidity was only $4.16 million, around 5% of total supply. If borrowers do not repay and others are also trying to leave, users may need to wait in line.
The third is collateral risk. The article points to maturity risk in dated structured assets, depeg risk in yield-bearing stable assets, and risks in wrapped cross-chain assets. It also notes that oracles can fail in stressed conditions. If bad debt emerges, losses are socialized across suppliers.
The piece says explicitly that it is not investment advice, does not provide a safety endorsement, and does not promise any return.
Protocol details and the author’s conflict disclosure
The protocol discussed is Morpho. The article says Morpho’s markets can be created permissionlessly, which leads to wide variation in market quality. Morpho’s interface steers depositors toward curated vaults, while direct supply to individual markets remains open at the protocol layer but is not surfaced through the default user flow.
The author also discloses that the tool referenced in the piece, Vane, was built and deployed by the author at https://vane.cryptodaii.org/ and collects a small tip on each transaction. The article treats that as a conflict of interest and says the operational explanation is written so readers can reproduce the process without touching the tool.
All figures in the article were taken from Morpho’s official interface on 2026-09-08, with screenshots timestamped between 06:59 and 07:20 UTC.
The three-step yield ladder in numbers
The first layer is Steakhouse USDC, an Ethereum v1 vault with contract address 0xBEEF01735c132Ada46AA9aA4c54623cAA92A64CB. The article lists total deposits of $67.79 million, a 5% performance fee, a 7-day average APY of 3.96%, a 30-day figure of 3.93%, and Steakhouse Financial as curator.
The second layer is Steakhouse High Yield USDC, an Ethereum v2 vault with contract address 0xbeeff2C5bF38f90e3482a8b19F12E5a6D2FCa757. It had $43.21 million in deposits, a 5% performance fee, a 0% management fee, a 7-day average APY of 5.09%, a 30-day figure of 5.29%, and the same curator, Steakhouse Financial.
The third layer is direct supply to the Ethereum market USDC / PT-reUSD-10DEC2026. The market ID is 0x1e9d614631a7df0ec07fb05b2c8cb2491575fd1a63a33bf187a6afb295a4fc64. The article lists a liquidation threshold of 91.5%, creation date 2026-06-16, oracle address 0x217d6DdCDB95112C51657F6270e8C079CFDB51f0, interest-rate model 0x870aC11D48B15DB9a138Cf899d20F13F79Ba00BC, target utilization of 90%, current utilization of 94.01%, total size of $84.31 million, total borrows of $78.82 million, withdrawable liquidity of $4.16 million, a liquidation penalty of 2.61%, and realized bad debt of 0.00.
Borrow-side rates in that market were 18.25% instantaneously, 10.51% over 24 hours, and 8.19% over 7 days. Because the interface did not print the 7-day average supply rate directly, the author multiplied the 8.19% borrow rate by 94.01% utilization and arrived at a supply-side 7-day average of about 7.7%.
That leaves the three-step comparison at 3.96%, 5.09%, and roughly 7.7%.
The screening method: how many verified curators are willing to allocate
The article says the author uses a process of elimination rather than a scoring framework. The most useful signal, in that method, is how many verified curator vaults have actually allocated capital to the market being considered.
For the target market, the first page of supplying vaults showed Pendle Ecosystem USDC with $31.45 million from Armitage by Wintermute; Steakhouse High Yield USDC with $9.99 million from Steakhouse Financial; RockawayX USDC Yield with $9.97 million from RockawayX; Wintermute USDC Select with $6.46 million from Armitage by Wintermute; and Hyperithm USDC Apex with $4.99 million from Hyperithm. Later pages included Re Ecosystem Vault, Smokehouse USDC, and Keyrock USDC.
The article says this answers two questions at once. First, at least six verified curators independently assessed the market and were willing to allocate capital. Second, the same Steakhouse Financial brand appears both behind the conservative vault yielding 3.96% and behind the vehicle that placed $9.99 million into the market yielding roughly 7.7%.
The author contrasts that with markets in the same list showing four-digit annualized rates, only one icon in the Trusted By field, less than $1 in withdrawable liquidity, and borrow totals in the low double digits. In the author’s view, several extra points of yield often correspond to the absence of multiple professional teams willing to back the market with real money.
A finer signal is the curator-imposed cap. The article says the high-yield Steakhouse vault had set an absolute cap of $10 million on this market and had already allocated $9.99 million, effectively full. Even so, the market accounted for only 23% of that vault’s total position. The author says a team willing to allocate but unwilling to go beyond a clearly defined cap is more informative than one making an all-in bet.
The article also lists hard rejection factors: a single-source or upgradeable oracle, collateral lacking enough on-chain depth for liquidation, borrowing so small that it suggests no real demand, and withdrawable liquidity that is too low relative to total supply. The target market itself is marked down on the last of those counts because its 5% liquidity cushion is thin.
There is also a technical note on data collection. Morpho currently has two generations of vaults, and many newer deployments store curator data in a different field. Checking only the legacy field can create the false impression that a market has no curator backing. The author says an earlier version of the Vane tool made exactly that mistake before it was fixed.
Step-by-step process and on-chain verification
In the operational section, the article says users should first choose an asset they understand, ideally a major stablecoin for a first attempt, then run candidate markets through the elimination framework. Before sending a transaction, five parameters should be read directly from chain and cross-checked against the interface: lend asset, collateral, oracle, rate model, and liquidation threshold.
The article warns that any front-end display can be stale or mismatched. Without this step, a user could supply funds to a market that is not actually the one they intended. The author says Vane forces a fresh on-chain read before approval and deposit, and refuses execution if the values do not match.
The next step is calculating the landing rate rather than staring at the headline number. Users should plug in the size of their planned deposit, estimate the market’s new utilization after their entry, and then estimate the corresponding rate.
The personal threshold used by the author is simple: if, on a 7-day average basis, the landing rate improves by less than 2 percentage points relative to the conservative vault, the trade is not worth taking. In the featured case, moving from 3.96% to about 7.7% improves the number by around 3.7 percentage points, which clears that hurdle. The article still says that being large enough on paper does not automatically make it worthwhile.
The landing-rate estimate, the author adds, is an instantaneous extrapolation from the public rate curve. It does not model the path of rates over time and is not an official protocol figure.
Execution requires one approval transaction and one deposit transaction. The article says the approval should target the protocol’s official adapter contract and should be limited to the precise transaction amount, not left as an unlimited approval. If a user later moves funds between markets, that action requires a broader one-time approval covering all positions on the chain, which deserves extra scrutiny. Morpho’s official contract address list is published at docs.morpho.org/get-started/resources/addresses.
The piece identifies three common loss vectors at this stage: unlimited approvals, approvals granted to unknown contracts, and signing on the wrong chain.
Net value breakdown on a $10,000 position
The article then runs the economics on a $10,000 principal amount.
It first fixes the holding period. The collateral expires on 2026-12-10, and the author later argues that users should exit at least two weeks before maturity. That leaves a usable window of roughly two and a half months, but the calculation uses two months to keep some buffer.
On the nominal side, a 7.7% annualized return over two months would generate about $128.3. The same two months in the conservative vault at 3.96% would produce about $66.0. The gross difference is therefore $62.3.
Costs are broken out into four categories, though one is already embedded in the comparison and not counted again:
- Gas. The full trip requires three on-chain transactions: entry approval, entry deposit, and exit withdrawal. On Ethereum mainnet on 2026-09-08, Etherscan showed gas at 0.051 gwei and ETH at about $2,470. Based on actual execution experience, the author estimates about $0.3 per transaction, or roughly $0.9 total.
- Tool tip. Vane charges a fixed proportion of gas cost with a minimum amount. At the gas level used in the example, the total stays under roughly $0.3 per action.
- Slippage. There is no swap in this flow, so slippage is 0.
- Opportunity cost. This is already captured in the comparison against the conservative vault and is not counted separately.
The result is an estimated all-in cost of about $1.5 and net excess return of roughly $61.
The article then computes the minimum viable position. The annualized excess is 3.74 percentage points, which equates to about 0.62% of principal over two months. Dividing costs by that ratio gives a breakeven principal.
At the observed fee level, with the round trip costing about $1, the threshold is around $160. If gas rises back to 10 gwei, the threshold climbs to about $1,800. At 30 gwei, it rises to about $5,400. The article says that means the barrier is determined by gas conditions at the time of execution rather than by the chain in the abstract. In the market conditions of September 2026, the old assumption that mainnet is necessarily too expensive does not hold, though that can change again if gas moves.
The author’s takeaway here is direct: under current fees, cost is no longer the main obstacle. Risk is.
The trigger lines that can break the trade
The article identifies several specific trigger conditions.
First is collateral maturity. The collateral in the featured market is a principal token with a clear expiry date embedded in its name: 2026-12-10. As maturity approaches, pricing behavior and discount dynamics can shift, so the author recommends exiting at least two weeks beforehand.
Second is collateral depeg. The oracle was pricing one unit of collateral at 0.973 USDC, already at a discount, while the liquidation threshold stood at 91.5%. The article notes that this is not hypothetical. On Aug. 25, the market saw a batch of real liquidations.
One liquidation shown in the screenshot repaid $6.83 million, and every line in that set showed realized bad debt of 0.00. The article reads that two ways: the liquidation system did work cleanly at multi-million-dollar scale, but it also proved that the collateral can in fact reach liquidation levels in batches.
Third is utilization saturation. Current utilization was 94.01%, already above the 90% target, while withdrawable liquidity was only about 5% of total supply. The article says the strong posted yield and the difficulty of exiting are two sides of the same condition.
Fourth is oracle failure or manipulation. A single-source price feed, an upgradeable oracle contract, or unclear control rights are all treated as outright disqualifiers. In the author’s wording, that category can produce full-loss outcomes, so it needs to be screened out at the market-selection stage.
Exit signals the author says should be written down before entry
The article closes with five pre-defined exit rules.
- Exit unconditionally two weeks before collateral maturity.
- Start withdrawing if withdrawable liquidity as a share of total supply falls below the threshold set in advance. The author’s own line is 5%, and the market in the example is already close to it.
- Leave if the improvement in landing yield over the conservative vault narrows to 2 percentage points or less on a 7-day average basis.
- Watch for a drop in the number of verified curator vaults supplying the market, or for large withdrawals by those funds.
- Leave if the user’s own position has become large enough to push the landing rate down close to the curated vault level.
All data and screenshots in the article were taken from Morpho’s official app, app.morpho.org, with capture times from 2026-09-08 06:59 to 07:20 UTC. The return metric used throughout is the 7-day average, while instantaneous rates appear only for comparison. The article also references Morpho’s official contract list at docs.morpho.org/get-started/resources/addresses and discloses the author-built tool at vane.cryptodaii.org.