TokenIntel Research
CROSS-ASSET · LENDING ARCHITECTURE
A TokenIntel Cross-Asset Report

Judgment
Is the Real
Concentration

Two lending protocols can hold the same percentage of capital in their top vaults and have completely different risk. Kamino concentrates around incentives. Morpho concentrates around underwriting. The second is harder to see and harder to escape.

Drafted June 17, 2026
Tokens KMNO · MORPHO
Lens Concentration Pattern
tokenintel.org/reports June 17, 2026
The argument, stated first

DeFi's next concentration crisis will not arrive through too much capital in too few vaults. It will arrive through too much underwriting trust placed in too few risk managers. Kamino and Morpho both show top-heavy capital on a risk dashboard. The shape of the concentration determines the shape of the failure, and the two shapes have almost nothing in common.

Live data via DefiLlama; full sources in the foot. Jump to sources.
Kamino-lend total TVL
$2.07B
live, DefiLlama
Morpho-blue total TVL
$10.81B
live, DefiLlama
Largest Kamino single market
$276M
USDE on Ethena Market
Top-8 Morpho curators aggregate
~$7.27B
TI curator-layer research, Apr 2026
Capital can diversify across vaults. It cannot diversify away from a shared risk model.

The shape beneath the heatmap

A risk dashboard that says "Kamino: 49 percent of TVL in the top four markets" and "Morpho: 65 percent of TVL across eight large curators" triggers the same alert color. The colors are the same because the percentages are the same. The shapes are different. The first is a concentration of capital around where the incentives point. The second is a concentration of capital around who decides where it goes.

Those two designs respond to stress in opposite ways. The first shape contracts on a known calendar when the incentive program ends. The second shape moves when a single risk manager re-prices a piece of collateral and every vault they run re-prices with them. Same percentage, different counterfactual when one number breaks.

WHAT WE CLAIM A concentration ratio in a vault market is not a complete risk metric. The right second question is whether the concentration is sitting in incentives or in underwriting. Those two answers point to different failure modes and different hedges.

Kamino: concentration around incentives

Live DefiLlama yields shows the top four Kamino-lend markets aggregate to about $736M, with the single largest market (USDE on the Ethena Market market) at $276M. These are the markets where the protocol is paying yield enhancements through external incentive programs. The depositor decision to sit there is rational, observable, and time-bounded.

Incentive concentration has a calendar. The depositor knows the program's emission schedule, the program's funding source, and the spread between subsidized and unsubsidized yield. When the calendar runs out, the spread compresses and capital migrates. That migration is what risk dashboards label as "outflow risk," and it is real, but it is not the kind of risk that surprises an attentive market.

The market can hedge incentive concentration. It can time the cliff, exit ahead of it, or simply not enter. The loss, when it lands, is bounded by the spread that disappears.

Morpho: concentration around underwriting

Morpho concentration sits one layer up from the vault. About $7.27B of capital is concentrated across roughly eight large curators (per the curator-economy analysis shared at the April 2026 Vault Summit). The top of the distribution is heavy: Steakhouse and Gauntlet have been trading the top spot at roughly $1.3B to $1.9B each across their vault networks. Multiple sources have surfaced specific concentration ratios; the underlying shape is consistent across them.

Each curator runs many isolated Blue markets through one shared methodology. That methodology decides which collateral is acceptable, which oracle is acceptable, which LLTV is correct, and which liquidation parameters survive a fast move. Depositors trust the methodology indirectly, by trusting the curator's track record. They do not, in practice, run an independent re-rating of the methodology itself.

The honest counter is that curator risk is not invisible. Vault holdings are public on chain. Curator allocations are public. Methodologies are documented. Parameter changes hit governance forums before they hit production. The right read is not that curator risk is opaque, it is that it is harder to translate into a position-sizing decision than incentive risk. A depositor can read Steakhouse's published methodology in full and still not know what the model misses. The data are visible; the blind spot in the model is not.

When a curator's methodology has a blind spot, that blind spot expresses across every vault the curator runs at the same time. The blind spot does not respect the protocol's market-level isolation. The architecture isolated the markets. The portfolio did not.

AN EMPIRICAL EXAMPLE, NOT A PROOF In April 2026, the KelpDAO bridge exploit minted ~$292M of unbacked rsETH and deposited it on Aave V3, producing $177M-$230M in estimated bad debt. Morpho had minor exposure through curator vaults that had taken on rsETH collateral. Per-market isolation contained the protocol-level cascade. The curators who allocated to that market still took the loss on every vault that held it. One incident is evidence of the shape, not proof of the magnitude. The shape recurs; the magnitude depends on what the next event is.

Which concentration would you rather own?

Our base case is that incentive concentration is the easier risk to underwrite as a depositor. The trigger is contractual, the timeline is published, and the worst-case loss is a contracted vault, not a lost vault. A depositor who reads the emission schedule can price the cliff and exit ahead of it.

The tail-risk read on underwriting concentration is harder. The trigger is a curator's judgment call that has not yet been falsified by a market event. There is no public schedule, no visible cliff, and the depositor cannot stress-test the methodology from outside. When the methodology fails, every depositor at the same curator discovers it together on the same day.

That is not a prediction that any specific curator will fail. The dominant case is that Steakhouse, Gauntlet, Re7, and MEV Capital continue to be correct about most of their underwriting decisions most of the time. They are professional risk teams with audit trails and a strong financial incentive to stay right. The asymmetric tail is that when one of them is wrong, the depositors who chose that curator are wrong together.

Incentive concentration is visible. Judgment concentration is hidden. The market can price the first. It struggles to price the second.
DimensionKamino-shape (incentive concentration)Morpho-shape (underwriting concentration)
Where the concentration sitsMarkets (per-pool incentive programs)Curators (shared underwriting models)
Failure triggerIncentive program ends or spread compressesCurator methodology has a blind spot that lands
Timeline of stressPredictable, calendar-basedUnpredictable, market-revealed
Cascade scopeOne market, containedEvery vault the curator allocates to
Depositor cost when it firesSpread that disappearsCorrelated loss across the curator network
Hedgeable from outside?Yes (read the calendar, exit ahead of the cliff)No (depositor cannot stress-test the methodology)

Same dashboard color, opposite failure mechanics.

The pre-2008 anchor, and why the parallel is exact

The global credit market before 2008 outsourced underwriting judgment to three firms: Moody's, S&P, and Fitch. Their stamps were the inputs every institution used to size positions, set collateral haircuts, and meet regulatory capital. The methodology behind each stamp was opaque to the consumer. The consumer was paying for the brand and the track record, not for the reasoning.

When the methodology miscalibrated on subprime mortgage CDOs, the error did not stay inside one firm. It propagated through every institution that had built positions sized to AAA. The capital was not concentrated in any single instrument. The judgment was. That was the contagion vector, and that was the part the regulators had not previously thought to measure.

Curator concentration in modular DeFi lending has the same shape. The depositor is not exposed to a single market or a single asset. The depositor is exposed to a methodology that has not yet been falsified by a market event. The Stream Finance xUSD vault loss in late 2025 was the first real test, and it landed on the depositors who had chased the highest curated yield. If curator AUM continues to grow at its recent pace, the next event of that shape could land on a materially larger capital base.

The question for a depositor in the next stress event will not be "was my vault concentrated?" It will be "was my curator wrong about the same risk as the other big curators?" That is the question worth being able to answer in advance.

When everyone outsources underwriting to the same minds, mistakes stop being individual and start being systemic. That is the lesson the credit-rating era taught the bond market. Modular DeFi lending is rediscovering it from first principles.