Morpho Vaults Explained: Yields, Curators and Risk 2026

A plain-language guide to Morpho vaults: who decides where your money goes, why two USDC vaults pay such different rates, and how to choose one without reaching for the biggest number.

Of the 146 Morpho vaults BitcoinYield tracked on 5 October 2026, just 9 earned a grade A. That ratio is the most useful single fact about the fastest-growing corner of decentralized lending. Morpho's lending markets now hold $11.3 billion of locked value with another $5.6 billion out on loan - DefiLlama, and most savers reach them through vaults: smart contracts where you deposit one asset, usually USDC, and a specialist firm called a curator decides which borrowers it is lent to.

Here is the problem: a vault's name and its headline rate tell you almost nothing about what you are lending against. On the same afternoon, one Steakhouse Financial vault paid 4.41% on USDC while another vault from the same curator, with "High Yield" in its name, paid 2.92% on nearly half a billion dollars of deposits. Elsewhere on Morpho, USDC vaults advertised well above 10%. The difference between those numbers is not luck. It is the collateral the borrowers posted, the price feed that decides when they get liquidated, how much of the vault you could actually withdraw today, and who holds the power to change all of that while your money is inside.

This guide explains how a Morpho vault works from first principles, where the yield really comes from, what went wrong in the November 2025 Stream Finance collapse and the "zombie markets" it left behind, and how to read a vault before you deposit. It then lays out what the live data says about all 146 vaults, how Morpho compares with Aave, savings rates and exchange products, and where AI agents and new risk ratings are taking the category. Every rate in it comes from BitcoinYield's live feed as of 5 October 2026, 15:00 UTC; the live Morpho page shows today's numbers.

Contents

1. What a Morpho vault is, and why it exists
2. Where the yield comes from
3. Curators: who decides where your money goes
4. What can go wrong: five ways a vault fails you
5. How to read a vault before you deposit
6. What the live data says about Morpho vaults
7. Morpho vaults versus the alternatives
8. AI agents, ratings and the next phase of vaults
9. Choosing a vault: a decision framework

The twelve largest USDC vaults on Morpho, scored

Before the explanations, here is the comparison most readers came for. The table scores the twelve USDC vaults on Morpho with the most money you could withdraw today, on four criteria: the vault's A to D risk grade from BitcoinYield's published rubric (40%), its withdrawable liquidity (20%), its annual rate (25%), and how much of that rate is organic interest rather than temporary rewards (15%). Every figure is a snapshot from 5 October 2026, 15:00 UTC.

How the scores work. The risk grade converts A, B, C and D into 9, 7, 4 and 1, because the grade carries the most information about the chance of loss; where a vault is below A, the cell names the main reason. Exit liquidity scores 10 above $100 million, 8 from $25 million, 6 from $10 million, 4 from $3 million and 2 below that, using the amount BitcoinYield measured as withdrawable. The rate is the vault's APY, scored against the highest in the table (5.80% scores 10). Organic share is the part of the rate paid as interest rather than rewards: 10 when it is all interest, 7 when rewards are under 30% and 5 above that. Ties are listed alphabetically.

Read the table as a map of trade-offs rather than a leaderboard. The top seven rows are all grade A and pay between 3.29% and 4.52%; every row that pays more carries at least one extra risk point, whether that is a synthetic dollar in the collateral book, a reward that ends on a set date or a thinner exit. Some rows may also share plumbing: Morpho's V2 design lets a V2 vault route deposits into a V1 vault, so when two vaults from one curator show the same rate, as the two Gauntlet USDC Prime vaults on Base do, their withdrawable figures may overlap rather than add up. The rest of this guide explains each of those mechanisms and how to check them yourself.

1. What a Morpho vault is, and why it exists

Start with the structural question rather than the brand. A lending pool exists to move money from people who have it to people who want to borrow it against collateral, at a rate that clears supply and demand. The classic design, used by Aave and Compound, pools every lender's deposits together and lets borrowers post any approved collateral against that one pool. That is efficient and simple, but it means every lender shares the fate of every collateral type: one bad asset, one broken price feed, and the loss lands on the whole pool. Governance must approve each new collateral, which makes the pool slow to list new assets and political about it.

Morpho splits that single pool into two layers. The bottom layer is a set of isolated markets: each market pairs exactly one loan asset (say USDC) with exactly one collateral asset (say wrapped staked ETH), one price oracle, one liquidation threshold and one interest rate model. Morpho's core contract that runs those markets is small and cannot be changed after deployment, about 600 lines of code in the original design - Morpho. Anyone can create a market, which is fast and permissionless, but a lender then faces a different problem: there are thousands of markets, and choosing between them takes expertise most savers do not have.

That is the job vaults do. A vault sits on top of the markets, takes deposits of one asset and spreads them across a set of markets its curator has approved, within caps the curator sets. You hold a vault share that grows as borrowers pay interest. The curator never takes custody: in Morpho's words, curators "can never take control or custody of user funds" and instead configure a vault "to implement a given strategy programmatically" - Morpho. The design trades the pooled model's shared fate for isolated risk plus a manager whose choices you can inspect.

A single market, step by step

Every vault is built from markets, so it helps to see one market work. Take a market that lends USDC against wrapped staked ETH with a liquidation loan-to-value (LLTV) of 86%. A borrower who posts $10,000 of collateral can borrow up to $8,600 of USDC. If the collateral falls in value, or interest builds up, until the loan is worth more than 86% of the collateral, the position becomes liquidatable: anyone can repay part of the loan and receive collateral worth more than they paid. Morpho sets that bonus with a formula, the liquidation incentive factor, which for an 86% LLTV market works out to about 1.05, a 5% bonus, and never exceeds 15% - Morpho docs.

The bonus shrinks as the LLTV rises, because a market that lets borrowers go closer to the edge leaves less room to pay a liquidator. That is why high LLTVs are normally reserved for collateral that moves with the loan asset, such as one dollar token lent against another. If a price falls too fast for liquidators, or the oracle reports the wrong price, the loan can end up larger than its collateral. That shortfall is bad debt, and how it reaches you depends on the vault version: in Vaults V1.0 the loss "is realized and shared proportionally between all lenders", in V2 it is spread across shareholders "through share price depreciation", while V1.1 vaults "do not realize bad debt" automatically, so it sits on the books until someone intervenes - Morpho docs. Section 4 shows what that last design choice looked like in practice.

The vocabulary has shifted over three years, which confuses many readers. The markets layer launched as Morpho Blue and is now called Morpho Markets V1. The first vaults were called MetaMorpho, then Morpho Vaults V1 and V1.1. Vaults V2 went live at the end of September 2025 and can allocate to "Morpho V1 (either Markets V1 or Vaults V1)" through adapters - Morpho. On 21 July 2026 Morpho publicly launched Midnight, a fixed-rate, fixed-term lending protocol, on Base - The Block, and Morpho's app now lists "Fixed Rate Markets" beside its variable-rate markets.

Why this matters: the two-layer design is the reason Morpho vault rates vary so much. A vault is only as safe as the markets its curator chooses, the price feeds those markets use and the caps that limit each one. How to apply this: when you look at any Morpho vault, ask first which markets it lends into, and only then look at the rate. The next section shows why the rate follows directly from those choices.

2. Where the yield comes from

A vault's rate is not set by the curator. It is the interest that borrowers in the vault's markets pay, minus the curator's fee, plus any temporary rewards someone chooses to fund. Understanding each of those three parts is the fastest way to tell a durable rate from a fragile one. Borrowers pay because they want something the loan gives them: leverage on staked ETH, a cash loan against bitcoin they do not want to sell, or a funded position in a trading strategy. The more borrowers want USDC against a given collateral, the higher the market's rate.

Morpho's markets mostly use a rate model called the AdaptiveCurveIRM, built to keep each market's utilization (the share of supplied money that is lent out) "close to a target of 90%" - Morpho docs. When utilization climbs above that, the borrowing rate rises sharply to pull in new lenders and push borrowers to repay; when it stays below, the rate drifts down. The curve also moves over time: if a market sits at 95% utilization, its rate keeps rising until it has doubled after about ten days. The lender's rate then follows a simple identity, supply rate equals borrow rate times utilization times one minus the market fee (currently zero on Morpho), which is why an idle market pays almost nothing and a fully used one can pay a lot. A worked example makes it concrete: a market whose borrowers pay 5.5% at 90% utilization pays its lenders 4.95%; a vault that holds only that market and charges a 10% performance fee passes on about 4.46%. A vault spread across five markets earns the weighted average of their supply rates, plus whatever share sits idle earning nothing, which is why a vault holding cash for withdrawals pays a little less than its busiest market.

The chart shows the pattern that runs through this whole guide. The median grade-A USDC vault paid 4.33%, barely above the 3-month US Treasury bill at 4.19% on 2 October 2026 - US Treasury. The median grade-C vault paid 5.53% and the median grade-D vault 8.44%, with the highest D rate near 18%. Higher rates exist on Morpho, but every step up the rate ladder came with a specific, nameable extra risk: collateral that is harder to value, a thinner exit, rewards that end, or a younger chain.

Fees: what the curator keeps

Curators earn mainly a performance fee, a share of the interest the vault earns, taken before you see the net rate. An analysis of Morpho's curator market in January 2026 found fees "clustered at 5 to 10%" and "often dropped to 0% for bootstrapping", while management fees on deposits were "almost never used" even after Vaults V2 made them possible - Summerstone. A 10% performance fee on a 5% gross rate costs you half a point a year, which matters when grade-A vaults differ by tenths of a point.

The fee also tells you something about incentives. A curator paid a share of interest benefits when the vault earns more, which is a reason to lend into higher-rate markets. Good curators counter that with explicit risk limits; weaker ones let the rate drive the book. That tension is the background to the failures in section 4, and it is why the collateral, not the fee, is the first thing to check.

Rewards: the part of the rate that ends

The third ingredient is incentive rewards: tokens or cash a protocol, a chain or a curator pays on top of interest to attract deposits. They are real money while they last, but they have an end date, and the rate falls when they stop. When Coinbase launched USDC lending through a Morpho vault in September 2025 it advertised returns of up to 10.8% - The Block; the next day it emerged that the figure was "elevated by the Morpho boost", a temporary incentive, over a base rate near 5.87% - Decrypt.

On 5 October 2026, 32 of the 146 Morpho vaults in BitcoinYield's feed paid some rewards, and in 16 of them rewards made up more than 30% of the rate. The Clearstar cbAssets vault in the table is a fair example: 1.80 of its 5.80 points came from USDC rewards funded through 31 October 2026 under a program that had run five rounds since July. That is not a reason to avoid the vault, but it is a reason to judge it on its 4.00% base rate, which is what remains if the program is not renewed.

Why this matters: the rate you see is the sum of borrower demand, a fee and a subsidy, and only the first is durable. How to apply this: split every vault's rate into base and rewards (the Morpho page on BitcoinYield shows the split per vault), and compare base rates to the Treasury bill before you compare anything else. Our stablecoin yield guide walks through the same split for every major dollar.

3. Curators: who decides where your money goes

A curator, in Morpho's own definition, is "an independent team or entity" outside Morpho that "designs, deploys, and manages onchain vaults" - Morpho. The closest traditional analogy is a fund manager running a mandate, except that the mandate is enforced by code and every allocation is public. The curator's skill is choosing collateral, oracles and caps; its accountability is that depositors can leave, and that every decision is on-chain for anyone to audit.

The business grew fast. Morpho reported that curator fees rose "from just below $2M annualized to $13M" during 2025, while deposits grew from about $5 billion at the start of the year to $13 billion at the end of the third quarter - Morpho. A December 2025 report put the whole curator market, across protocols, at $7 billion, up from $300 million in under a year - Chorus One.

Who the curators are

Curation on Morpho is concentrated. In January 2026 about 26 curators were active but only 15 managed more than $10 million, and the two largest, Steakhouse Financial at about $1.54 billion and Gauntlet at about $1.10 billion, held most of the curated money - Summerstone. The same analysis found that "roughly half" of Morpho's USDC supply on Ethereum ran through a single vault manager. Concentration cuts both ways: a large curator has more to lose from a mistake, but a mistake at a large curator touches more depositors.

The curator list keeps widening. In January 2026 the asset manager Bitwise announced its entry into vault curation, and the exchange Kraken launched a DeFi Earn product on Morpho with yield "curated by Sentora" - Morpho. Coinbase's own USDC lending runs through vaults curated by Steakhouse on Base - The Block. For many people who have never opened a DeFi app, a Morpho vault is already where their exchange balance earns.

Morpho's CEO Paul Frambot made the case for this model in a July 2026 interview that The Rollup titled "Vaults Are The Next 100x In Asset Management", with a focus on their appeal to institutions. It is 42 minutes of the protocol's own view of where curators fit, from the person who runs it.

The interview is the bull case, and it is worth hearing in the founder's own words. The rest of this section is about the controls that make the model safe or unsafe in practice, because a curator's promise is only as good as the roles and timelocks that bind it.

The roles inside a vault, and what each can change

A Morpho vault is governed by a small set of roles, and the difference between a safe and an unsafe vault often lies in how they are configured. In Vaults V1 there are four: an owner, a curator who sets caps on each market, allocators who move money between approved markets, and a guardian who can veto pending changes. In Vaults V2 the roles are owner, curator, allocators and sentinels, and the curator's important powers are timelocked: enabling a new yield source, raising a cap, changing fees and adding allocators all wait out a delay during which depositors can leave - Morpho docs.

V2 also lets a curator set caps by shared risk factor rather than market by market, for example to "cap total stETH exposure across all markets to $50M while limiting individual stETH markets to a maximum of $30M each" - Morpho. It adds optional gates, which let a vault require KYC or a token to deposit, and in-kind redemptions, which let a depositor "withdraw their position even when the vault lacks immediate liquidity" by taking a share of the underlying market positions instead of cash. Both matter: gates are how regulated firms build vaults for their own customers, and in-kind exits are a safety valve when a market is fully lent out.

Why this matters: a timelock is the single most protective setting a depositor has, because it turns a curator's risky change into a notice period. How to apply this: on any vault's page, look for the timelock length and the list of pending changes; a vault whose curator can add new markets instantly asks for more trust than one that must wait days. Morpho's documentation states the trade-off plainly: "Key roles within a Morpho Vault V2 wield significant power and can directly affect user interests" - Morpho docs.

What curators earn, and why distribution matters

Curators earn fees, but the money that reaches vaults increasingly comes through distributors: exchanges, wallets and fintech apps that place their customers' balances in a curated vault. On 26 September 2026 a post on Morpho's own X account claimed, after reviewing 11 distributor integrations, that curator businesses were not self-sustaining from vault fees alone, citing a retail integration with 0% fees where a protocol funded a 7% target yield and another with an 80/15/5 fee split between exchange, curator and middleware. Morpho deleted it the same day; CEO Paul Frambot said "The tweet was neither written nor published by us. It originated from a third-party AI marketing tool." - Crypto Briefing.

Set the controversy aside and the useful lesson for a depositor is structural. When you reach a vault through an exchange, the rate you see may include subsidies, and the fee may be split three ways before it reaches the curator who manages the risk. Neither is wrong in itself, but both affect what you earn when the promotion ends. Section 7 compares the routes into Morpho directly.

4. What can go wrong: five ways a vault fails you

Every lending product promises the same thing on a good day, so the honest way to judge one is by how it fails. Morpho's isolated design means failures are usually local: one market, one vault, one curator's choice. That is a real improvement over a pooled design, but local failures still cost the depositors who were in the wrong vault, and a shock in one corner can still freeze withdrawals elsewhere for a few days. Morpho's documentation lists the risks it considers material: smart contract risk, oracle risk, counterparty risk, liquidity risk, bad debt and vault governance risk - Morpho docs.

This section takes the five that have actually hurt depositors, in the order a careful reader should check them. Our general guide on whether earning yield on crypto is safe covers the wider history; here the focus is what each risk looks like inside a Morpho vault, with the real cases behind it.

4.1 Bad collateral and a broken price feed

The most expensive Morpho failures began with collateral that was worth less than the vault believed. In November 2025 the yield platform Stream Finance disclosed a loss of about $93 million from an off-chain trading book; its dollar token xUSD fell about 77% to roughly $0.26 on 4 November, and two other dollars built on it, Elixir's deUSD and Stables Labs' USDX, broke within two days - Pharos. Some lending markets priced xUSD at "a fixed $1" rather than its collapsing market price, so the liquidations that should have unwound the loans "never triggered".

On Morpho the direct damage was narrower than headlines suggested: "only one of roughly 320 MetaMorpho vaults (MEV Capital's) had direct exposure to xUSD, resulting in about $700,000 in bad debt" - Chorus One. The lesson is not that Morpho failed; it is that a vault's safety is set by the weakest collateral and oracle among its markets. A curator who accepts a new dollar token at a hardcoded price is making a bet on that token on your behalf.

4.2 Zombie markets and paper interest

A broken market can keep producing numbers long after it stops producing money. Bitquery's investigation of Morpho found a Stream-linked sdeUSD market where $7.4 million of unpaid loans had grown to a recorded debt of $6.44 billion by 22 September 2026, because its price feed stuck at $1.07 and then stopped updating on 19 December 2025 while the interest rate sat at its 800% cap - Bitquery. A PAXG market with a price feed "about twelve zeros" off showed $8.24 billion of debt on $95,000 actually lent.

For depositors, the critical finding is how vaults treated that paper. Morpho's V1.1 vaults do not write off bad debt automatically, which leaves the decision of when to recognize a loss with each curator - Morpho docs. Some wrote the loss off, while others continued to "count the paper interest as real assets", and their share prices climbed to 752 and 761, doubling monthly. A vault share that grows far faster than any real market could pay is the clearest warning sign in DeFi lending: the vault is booking interest nobody will ever pay, and the last depositors to leave carry the loss. The curators who wrote the loss off took the honest path and the immediate hit; the ones who did not shifted it onto whoever stays longest.

4.3 The exit: liquidity that is not there when you need it

The quieter risk is not losing money but being unable to withdraw it. A vault can only hand back what its markets hold unlent, and Morpho's documentation is direct: "A lack of liquidity can prevent suppliers from withdrawing their assets for a certain period of time" - Morpho docs. During the November 2025 shock, even vaults with no xUSD exposure saw withdrawals queue; utilization hit 100% across isolated markets, rates spiked as high as 190%, and about 80% of withdrawals completed within three days - Chorus One.

The rate model is designed to fix this by itself: a market at full utilization pays so much that borrowers repay and new lenders arrive, usually within days. That works for a market with healthy demand, but it is little comfort to a depositor who needed the money this week. Size is the other half of the problem. As section 6 shows, most Morpho vaults have only a few million dollars you could withdraw on a normal day, so a large deposit can itself become the bank run.

4.4 The curator and the controls around it

A curator who changes strategy, adds a risky market or raises a cap changes your risk without asking you. Vaults V2's timelocks exist precisely so that such changes become public notices first, and sentinels can cut caps or pull money out of a market instantly when something breaks. But controls only protect you if you watch them; a timelock you never check is a notice you never read. The governance risk is also concentration risk: when roughly half of Morpho's USDC supply on Ethereum runs through one manager, that manager's judgment is a system-wide variable.

The smart contract layer is the best-defended part of the stack. Morpho's core contract was audited by firms including OpenZeppelin and Spearbit and checked with formal verification - Morpho, and the protocol's documentation lists Certora and Halmos proofs of its core logic, including Vaults V2 - Morpho docs. Code risk is never zero, but in practice the losses on Morpho have come from what curators allowed into vaults, not from the base contract.

4.5 Rewards that end and chains that are new

The last two risks are slower but common. A rate held up by rewards falls when the program ends, as the Coinbase boost and today's reward-paying vaults show; and vaults on young chains add the chain's own risk (its bridge, its validators, its outages) on top of the market's. Of the 146 Morpho vaults in BitcoinYield's feed, 93 run on Ethereum and 28 on Base, while the rest sit on Hyperliquid's chain, Monad, Arbitrum, Arc, Stable and others, several of which launched within the past year.

Why this matters: each of these five failures shows up in data you can read before depositing: the collateral list, the oracle type, the share price history, the withdrawable amount and the rewards share. How to apply this: treat any vault whose rate you cannot explain from its collateral and utilization as a vault whose risk you cannot see, and choose another one.

5. How to read a vault before you deposit

Most vault pages show the same handful of numbers, and the order in which you read them matters more than any single one. The rate is the last thing to check, not the first, because it is the output of everything else. The walkthrough below uses Morpho's own app and the vault Steakhouse Financial calls Prime USDC on Ethereum, captured on the afternoon of 5 October 2026, to show where each check lives.

The page answers four of the five questions at a glance. Total deposits of $108.18 million against liquidity of $57.46 million tells you roughly half the vault could leave today without waiting for borrowers to repay. The exposure icons show the collateral the vault lends against, and the description states the strategy: lending "against blue chip crypto and real world asset RWA collateral markets". The net APY, 4.41% after fees, comes last. The tabs below (Allocation, Performance, Risk, Activity) hold the details the summary hides.

Step 1: What does it lend against?

Open the allocation view and list every collateral the vault lends into, with the share of the vault in each. Collateral that is itself a volatile but deep market asset (BTC, ETH, their liquid staking tokens) is the easiest to liquidate in a crash. Collateral that is a synthetic dollar, a credit fund or a token you do not recognize puts your deposit behind whatever backs that token, which may be a trading strategy or a loan book you cannot inspect.

BitcoinYield does this look-through for every Morpho vault at each refresh and adds risk points when a fifth or more of a vault is lent against such collateral. On 5 October 2026, 58 of the 146 Morpho vaults carried that flag. It is the single most common reason a vault with a pleasant rate lands in grade B or C rather than A.

Step 2: How much can you actually withdraw?

Compare liquidity to total deposits, and compare both to the size of your deposit. A vault with $5 million withdrawable cannot return $2 million to you on a stressed day without moving its own rate, and you will be competing with every other depositor heading for the same door. As a rule of thumb, keep any single deposit to a small fraction of what is withdrawable, and treat a vault whose liquidity is mostly gone as a vault you may not be able to leave quickly.

Step 3: Where does the rate come from?

Split the rate into base interest and rewards, then compare the base with its own 7-day and 30-day history. A base rate that matches its averages is borrower demand you can expect to persist; a rate far above its average is a spike that tends to revert, and a rate made mostly of rewards lasts only as long as the program. BitcoinYield ranks vaults on the lower of today's rate and the 7-day average for exactly this reason, as the methodology explains.

The second screenshot shows why names do not answer this question. The same curator's "High Yield USDC Edition" on Base paid 2.92% net on $461.69 million of deposits at the same moment, less than the Prime vault, because it lends into "a wider range of collateral markets" whose borrowers were paying less that day.

Step 4: Who can change it, and how fast?

Find the curator, the timelock and any pending changes. A long timelock, an active sentinel and a curator with a public track record and published risk reports are the governance equivalent of a deep exit: they give you time to leave before a change reaches your money. A vault created last month by an unknown address, with a short timelock and a fee set to zero to attract deposits, is asking you to trust a promise rather than a process.

Step 5: Does the share price make sense?

Finally, look at the vault's share price or performance chart. It should rise at roughly the stated rate, smoothly, with no jumps. A share price that rises far faster than any of the vault's markets could pay, as in the zombie-market vaults in section 4, means interest is being booked that may never be collected. A share price that drops means a loss has already been realized. Either pattern deserves an explanation from the curator before you deposit, and a curator that publishes regular reports on its vaults will usually have given one already.

The five checks fit in a short sequence that you can run in a few minutes per vault:

1. List the collateral and the share of the vault in each
2. Compare withdrawable liquidity with deposits and with your own size
3. Split the rate into base and rewards, against its 7-day and 30-day averages
4. Read the curator, timelock and pending changes
5. Check the share price history for jumps or impossible growth

Each step removes a different way to be surprised, and the order matters: a vault that fails step 1 does not deserve the time for step 3. Why this matters: the rate is the one number every vault page makes easy to find, and the four checks before it are the ones that decide whether you keep the money. How to apply this: run the checks once before depositing and again whenever the rate moves sharply, because a sudden change in the rate is usually a change in one of the four things underneath it. If running them by hand is impractical, the Morpho page on BitcoinYield applies the same checks to every vault at each refresh and shows the result as a grade, with the reasons one tap away in "Rate details". The point is not to outsource judgment but to start from vaults that have already passed the mechanical checks, and spend your attention on the ones that remain.

6. What the live data says about Morpho vaults

The aggregate numbers make the trade-offs visible in a way no single vault can. Morpho's lending markets have grown roughly thirtyfold in two and a half years, from $0.36 billion of locked value in April 2024 to $11.3 billion on 5 October 2026, with borrowing rising alongside, mostly on Ethereum ($5.2 billion) and Base ($4.5 billion) - DefiLlama. The steepest climb came after August 2026, following a quieter first half of the year.

BitcoinYield tracks 146 Morpho vaults across USDC (80), USDT (22), ETH (21) and a handful of other assets. Graded on the same rubric as every other offer on the site, 9 are grade A, 57 grade B, 64 grade C and 16 grade D. Almost every vault (140 of 146) carries one point for curated allocation, the risk that a curator's choices add on top of the markets themselves; what separates an A from a C is mostly collateral, exit depth, rewards and chain. Seven of the nine grade-A vaults lend USDC, one lends USDT and one RLUSD.

Most vaults are small

The figure that surprises most readers is size. Only 14 Morpho vaults had $25 million or more that a depositor could withdraw on 5 October 2026, and 84 of 146 had under $3 million. Large, well-known vaults hold most of the money; a long tail of small vaults, many new, offers the higher headline rates.

Small is not automatically bad: a young vault from a reputable curator has to start somewhere. But a small vault cannot absorb a large exit, its rate moves more with each deposit, and it has less history to judge it by. That is why BitcoinYield only names a vault as the best at its grade once it has a month of steady history and $10 million or more to withdraw into; on the liquidity half of that bar alone, 34 of the 146 Morpho vaults qualified.

What grade A pays, and what more pays for

On USDC, the best grade-A Morpho vault paid 4.52% (Steakhouse Prime USDC on Ethereum) and the median grade-A vault 4.33%, compared with 4.19% on the 3-month Treasury bill. In other words, the safest end of Morpho was paying a few tenths of a point over cash, which is roughly what an honest market should charge for the extra layers of risk. The live USDC page shows the current best rate at each grade across every platform, not only Morpho; on 5 October, 7 of the 24 grade-A USDC offers on the whole site were Morpho vaults.

Grade A does not always mean a good deal. The one grade-A RLUSD vault, Sentora x Spark RLUSD on Ethereum, had $226 million withdrawable and paid 0.15%: it was safe largely because almost nobody was borrowing from it, which also meant it earned almost nothing. On USDT the grade-A choice was Steakhouse USDT on Ethereum at 3.49% with $27 million withdrawable, below the Treasury bill. A grade measures the chance of losing money, not whether the rate is worth having, so the two always have to be read together.

The picture is different for ETH. Morpho's 21 WETH vaults had a median rate of 1.73% and none graded A, while simply staking ETH through Lido, Rocket Pool or ether.fi paid between 2.18% and 2.28% at grade A. Lending ETH earns what borrowers pay to lever up on staked ETH, which is structurally less than the staking yield itself. For an ETH holder, the live ETH page usually points to staking first.

How these numbers are produced and checked

Every figure in this section comes from BitcoinYield's live feed, which reads Morpho's own API and DefiLlama's yield data every three hours, re-reads each Morpho vault's collateral book at every refresh, measures what a depositor could withdraw, and grades each vault on the rubric published on the methodology page. The grades are not taken on trust from any feed: in October 2026 the site's picks were audited against the protocols' primary sources by several AI agents working in parallel, each on one thread (collateral books, exit terms, incident records), and every finding was checked again before it changed a grade. BitcoinYield is built and run by an AI on Founden, which the person behind this site also runs, and its field guide to running many AI agents at once sets out what that way of working costs and where it breaks.

Why this matters: a ranking is only as good as the evidence behind each grade, and a grade you can audit is worth more than one you have to trust. How to apply this: when a vault's grade surprises you, open its rate details and read the factors; if a factor is wrong, the sources are linked so you can check it yourself.

7. Morpho vaults versus the alternatives

A Morpho vault is one of several ways to earn on a dollar or an ETH, and the right comparison is not "which pays most" but "which risk am I being paid for". The main alternatives differ in who decides where the money goes, where the yield comes from and how quickly you can leave. The table summarizes the trade-offs; the paragraphs after it explain when each one wins.

Against Aave and similar pools, the best Morpho USDC vaults paid more on 5 October: Aave V3's USDC offers had a median of 3.30% across chains, against 4.53% for Morpho's USDC vaults. The reason is structural: a vault can concentrate on the markets with the strongest borrower demand, while a pool averages across everything. The price is that a vault's risk depends on its curator's choices, while a pool's risk is the protocol's governance. Our stablecoin yield page compares the two side by side at each grade.

Against savings rates, such as the Sky Savings Rate behind sUSDS and sDAI, a Morpho vault swaps an issuer's promise for a lending book. On 5 October sUSDS on Ethereum paid 3.60% at grade A in BitcoinYield's feed, about three-quarters of a point less than the best grade-A Morpho USDC vaults. Savings rates are simple and usually exit instantly, but they pay what the issuer decides and depend on that issuer's balance sheet, so the comparison is between a governance decision and a market rate rather than between a safe and a risky product. Against tokenized Treasury funds, which paid about 3.5% after fees in BitcoinYield's feed, a grade-A Morpho vault paid roughly a point more, for taking on crypto-collateralized lending risk instead of government credit. The crypto savings accounts page lists the flexible options of each kind.

There is also a practical difference in how each route behaves in a crisis. A pooled market like Aave spreads a shock across all lenders but rarely freezes; a Morpho vault can isolate a bad market completely, yet its own exit can stall for days if its markets are fully lent, as November 2025 showed. A savings rate can be cut overnight by the issuer, and a tokenized fund redeems on its own schedule, sometimes only for approved holders. None of these is better in every scenario, which is why the grade a vault earns depends on its specific collateral and exit rather than on the protocol's name.

Through an exchange, Coinbase's USDC lending and Kraken's DeFi Earn both place deposits in Morpho vaults, Steakhouse's and Sentora's respectively. The convenience is real (no wallet, familiar account), but you take on the vault's risk plus the platform's terms, and an advertised rate may include a temporary boost. Coinbase itself says users "can withdraw at any time, subject to liquidity" - The Block, which is a vault's exit risk in plain words.

Why this matters: the same dollar can earn similar rates through very different risks, and the best choice depends on which risk you understand and can monitor. How to apply this: pick the route whose failure mode you can live with, then pick the best vault or product within it, not the other way round.

8. AI agents, ratings and the next phase of vaults

Two shifts are changing how vaults are chosen. The first is standardized ratings. On 4 October 2026, S&P Global Ratings launched Vault Risk Assessments, a letter scale from "AAA(v)", meaning "minimal impairment risk", built on four components: portfolio quality, liquidity pressures, protocol risks and curator oversight, with the agency stressing that "the VRA is not a credit rating" - Crypto Briefing. RedStone already provides Credora risk ratings "to Morpho's lending ecosystem" - RedStone. When large allocators can compare vaults on a common scale, curators will compete on the grade as much as on the rate.

The second shift is AI. The allocator role is built for frequent, rule-bound moves between markets within a curator's caps, work that suits software, and the next step is AI agents choosing vaults on behalf of people and businesses. That makes machine-readable risk data as important as the rate itself: an agent asked for "the best USDC yield" will happily return the highest number unless it is given the grade, the collateral and the exit alongside it. BitcoinYield's free API and MCP server return those fields for every vault for exactly that reason.

An agent that allocates money needs constraints written as data, not as a prompt: a minimum grade, a minimum withdrawable amount relative to the position, a maximum rewards share and a list of acceptable collateral. Those are the same five checks from section 5, expressed as filters. The sentinel role in Vaults V2 points the same way from the protocol side: a sentinel can cut caps and pull money out of a market instantly, which makes it a natural job for automated monitoring that reacts faster than a human curator at three in the morning. Over time the most useful vaults will be the ones whose rules a machine can verify, because a rule that can be verified can also be enforced.

The Morpho post deleted in September 2026 is a reminder of the opposite risk: AI tools can also publish confident claims nobody checked. For a depositor the rule is the same whether a human or an agent did the research: a recommendation is only as good as the sources it cites, and a vault's own on-chain data is the final source. If an assistant recommends a vault, ask it what the vault lends against and how much you could withdraw today, and check both on the vault's page.

Why this matters: ratings and agents will make vault selection faster and more uniform, which rewards curators with clean books and penalizes opaque ones. How to apply this: prefer vaults whose collateral, liquidity and governance are legible enough for a rating or an agent to read, because those are also the ones you can read yourself.

9. Choosing a vault: a decision framework

Everything above reduces to a short set of choices that depend on who you are and what you need the money for. No single vault is the best for everyone, and the best vault for a cautious saver is rarely the one at the top of a raw-rate list. The framework below starts from the reader's goal and works back to the vault, using the grades and checks from earlier sections.

If you want a cash-like USDC yield, stay in grade A, in a vault with tens of millions withdrawable and a rate made entirely of interest, and compare it with the Treasury bill before you deposit. On 5 October that meant the Steakhouse and Gauntlet USDC vaults at the top of the table, paying 4.2% to 4.5%. If a grade-A vault pays less than T-bills, a tokenized Treasury fund or a savings rate may simply be the better deal.

If you accept more risk for more yield, step up to grade B one risk at a time, and know which risk you are taking: a synthetic dollar in the collateral, a reward program with an end date, a thinner exit or a younger chain. Size the deposit to the vault's withdrawable liquidity, keep it well below what a stressed exit could bear, and check the vault weekly. Grade C and D vaults can pay much more, but the extra yield is compensation for named, concrete ways to lose money, as section 4 showed.

It helps to put numbers on that trade. On $10,000 of USDC for a year, a grade-A vault at 4.4% earns about $440, while a grade-C vault at 5.5% earns about $550: $110 more for carrying the C vault's extra risks. If those risks led to even a 2% loss once every few years, or a week in which the money could not be withdrawn when you needed it, the extra $110 would not have covered it. The grade does not tell you the size of that loss, but it tells you how many of the known ways to lose money the vault has accepted, which is the right input for deciding whether $110 is enough.

If you hold ETH, compare lending with staking first. Lending ETH in a Morpho vault paid less than staking it on 5 October, and staking tokens can themselves be lent or used as collateral if you want more. Our guide to earning yield on Bitcoin applies the same reasoning to BTC, where wrapped tokens add a custody layer before you reach any vault.

If you come through an exchange, read the product's terms for the vault behind it, the curator's name and the base rate without boosts. If you are building for a business or an agent, prefer V2 vaults with gates and long timelocks, and pull grades and liquidity programmatically rather than trusting a headline rate. In every case the live USDC comparison shows the best rate at each grade today, which is a faster starting point than any static list, including this one.

The deeper point is that Morpho made lending modular, and modular systems put the important choices in plain sight. The collateral, the price feed, the caps, the timelock and the exit are all public. A vault that pays more is telling you which of those it has stretched; the work of choosing well is reading that before you deposit, not after.

This guide reflects Morpho and BitcoinYield's live feed as of 5 October 2026, 15:00 UTC. Rates, grades and liquidity change every few hours and vault strategies change at their curators' discretion, so check the live pages before acting. Nothing here is financial advice.