(､ᐛ)ﾍ＿/ Nadim's Blog

The myth of isolated lending in DeFi

TLDR: Isolated lending isn't a guarantee of curated vault lending, it can exist within that model but true isolation is rare. Irresponsible curators introduce risks at the protocol level. Risk assessment needs fixing, but isn't enough on its own. Skin in the game is the solution.

Last week's events

Last week was one of those rare ones that contribute disproportionately to the discourse around DeFi's assumptions and hidden risks. Such events are often needed to create the momentum and urgency to improve on weak models and denounce perverse incentives.

In short, we witnessed a confluence of the following: (loosely or closely inter-linked)

DeFi surviving small shocks such as this one, if it manages to learn and adapt, is a sign of robustness. This series of events exposed some flaws:

Let me now elaborate:

Pooled Lending vs. Curated Vaults

Part of the reason why riskier assets are now becoming collateral for loops is the rise of seemingly isolated lending via curated vaults. This model is seen by many as a natural successor to Aave's unified pool-based lending model, bringing the necessary modularity and segregation to accomodate long-tail assets or different liquidity and risk profiles.

The key difference between pool-based lending (Aave, Spark) and seemingly isolated lending (Morpho, Euler, Silo) is the introduction of external curators, responsible for risk assessment of collateral and allocating liquidity from lenders to underlying markets (defined as a collateral and loan asset, as well as a set of risk parameters).

Incumbent money markets such as Aave delegate risk assessment to the DAO and a set of aligned stakeholders (external consultants, risk assessors etc.). All USDC lent out in Aave is receiving the same exposure to underlying markets and utilization: no diversification of risk/return for lenders. Central planning ensures the system stays risk-averse, given that all risk is pooled together.

On paper, curated vaults can improve on this model by creating different risk/return profiles for lenders, packaged in separate vaults. This can enable leverage for riskier or more illiquid assets whereby higher risk lenders are getting appropriately compensated and safe lenders can still selectively maintain exposure to safe markets.

This is more complex in practice, because most curators have no skin in the game whatsoever, and every incentive to maximize yield at all cost while disguising the level of risk.

The existence of high-risk curators within the system cannot be shrugged off. This is because shared markets present a risk of contagion, and safe vaults can and will be affected by risky actions from degen curators. This complexity is not well understood by lenders.

Contagion scenario in isolated vaults with shared markets

Stani, the founder of Aave, denounces shared markets as a core flaw of curated lending when it comes to risk isolation:

The core issue with the curation vault model lies in the illusion of isolation. Curators are meant to manage distinct strategies and segregate risk, yet in practice, they all end up supplying liquidity to the same underlying lending markets. What is designed to promote diversification instead concentrates exposure, turning one curator’s stress into everyone’s problem.

Despite being framed as modular and independent, the model inherently links all vaults through shared borrower pools. Liquidity from multiple curators merges into a single system, so the decisions or withdrawals of one can ripple instantly across all others as we see with recent issues with xUSD and similar assets. Even a cautious curator cannot escape the fallout from a more aggressive participant operating within the same pool.

When confidence falters or withdrawals accelerate, these shared markets seize up. Utilization shoots to 100%, redemptions grind to a halt, and borrowing rates spike to unsustainable levels. A localized liquidity crunch quickly transforms into a protocol-wide freeze, a DeFi version of a bank run essentially.

This creates a design paradox: a system built for isolation that, in reality, amplifies interdependence. Every vault inherits the risk behavior of the weakest curator, making the entire structure vulnerable when liquidity is most needed.

This is a necessary tradeoff and intentional design choice. Fully segregated markets make for incredibly poor user experience and have mostly failed when attempted in the past. Morpho itself has roots in attempting true peer-to-peer lending but realized somewhere along the way that pooling was necessary, and that shared markets meant better capital efficiency.

The above is (for the most part) corroborated by Morpho co-founder's own explanation of the shared market/isolated vault model:

https://x.com/MerlinEgalite/status/1986823500720406949?s=20

He illustrates that true isolation IS still possible within the Morpho model but is not the default state or a guarantee, and somewhat insists that users should be able to evaluate the risk of potential shared markets when depositing via the Morpho UI. Core to the argument is that the safest/largest vaults on Morpho have very little shared risk with higher risk vaults in their current set-up.

While the Morpho team recognizes the need for risk ratings and better explanation of risks to their users, having that function performed by an unrelated third party still falls short of creating a true alignment of incentives. Skin in the game will be necessary to ensure curators behave as they were intended to.

Focus less on isolation, more on skin in the game

Curated money markets cannot pretend their weakest links do not affect the whole, and need to do more to ensure key stakeholders are behaving as intended and aren't adversarial or extractive to lenders. They have a certain responsibility, at the protocol level, to avoid extremely irresponsible curators, since their behaviour does introduce systemic risk that can propagate under the right circumstances. Last week was a glimpse at how a larger liquidity shock could spread through the system.

Curators are incredibly predatory and have no skin in the game whatsoever and every incentive to go as degen as possible. Most teams who have had the (dis)pleasure of working with curators this past year tend to agree with this, and it is accentuated by the amount of incentives and king-making going on. Curators are using incentives by new ecosystems to raise capital, while selecting riskier collaterals to outcompete other incentivized curators. Only a handful of curators receive the bulk of incentives; everyone else is unable to compete if they aren't receiving the same subsidies. This prevents a flight to quality from happening.

We still need segregated risk/return lending to exist in DeFi, pool-based "monolithic" lending can only get so far. It isn't the end all be all. Aave and Spark have an impressive track record and much clearer incentives than your run-of-the-mill Euler or Morpho curator. The system is less opaque, precisely because there are no hidden risks inherited from irresponsible curators via shared markets. The solution cannot be to abolish curated lending, but rather to improve on the current model by ensuring curators behave and risks are correctly evaluated and disclosed.

Morpho, having created a modular system, places the onus on lenders to identify such risks: it's far from straightfoward. The curators are almost adversarial to lenders in this model, risk is repackaged, obfuscated by complexity.

To borrow Nassim Taleb's nomenclature from the Incerto:

No Skin in the Game Skin in the Game Skin in the Game of Others OR Soul in the Game
(Keeps upside, transfers downside to others, owns a hidden option at someone else's expense) (Keeps his own downside, takes his/her own risk) (Takes the downside on behalf of others, or for universal values)
Bureaucrats, Consultants, Large Corpos, Politicians, Bankers, Central Gov Entrepreneurs, Citizens, Speculators, Traders Saints, innovators, real writers, revolutionaries, real journalists

Within these definitions, Morpho has limited skin in the game. Curators have no skin in the game. Only lenders & borrowers have skin in the game. There are mechanisms through which curators can have skin in the game. Implementing such mechanisms does force curators to be more opinionated, since lender and borrower interests are often at odds. It begs the question of whether curators primarily have a duty to lenders, borrowers, protocols or issuers of collateral assets. This has been discussed to some extent but there is no consensus. The Aave DAO itself behaves as a curator and is faced with the same choices, often carefully balancing its own alignment with stakeholder interests.

It will be interesting to see different curators strike a balance between lenders, borrowers, ecosystems, protocols and asset issuers. Skin in the game forces them to put their cards on the table. They can decide to be the first capital into lending vaults, become junior creditors, loop using their own markets or even provide borrow rate insurance. Sharing downside, not only upside, forces them to be more opinionated.

Everyone wants to participate in everyone's upside: it means a lot more to agree to share someone's risk. By doing that, every actor in the system shows exactly where they stand, which perfectly counteracts the obfuscation of risk currently driven by unclear, hidden adversarial interests.

Morpho, while often positioning itself as a perfectly neutral infrastructure layer, cannot exclude itself completely from having skin in the game. It is after all a highly successful, venture-backed entity (or really set of entities) with significant control over curator whitelisting and protocol-defined abstractions. There is no such thing as a neutral agent collecting fees: Morpho participates in the upside and monetizes all activity, and therefore should inherit appropriate downside. It is a question of symmetry, and it cannot exist only at the curator level.

No true isolation: risks of accepting tokenized yield products as collateral

The main argument in favour of Aave's centrally curated, monolithic risk assessment, is that it allows the system to remain extremely risk-averse, with less competitive pressure within the protocol (Aave still does compete with external protocols, but not with itself in the same way curators do).

But is this really true, when they've allowed what is objectively a tokenized hedge fund strategy to be hardcoded as a stablecoin within the protocol, now accounting for a worryingly large portion of all borrowing.

Jordi from Selini makes a good argument here:

https://x.com/gametheorizing/status/1986706547091575124?s=20

The gist of it is that:

The logical conclusion to all of the above is that it is incredibly irresponsible to be hardcoding USDe (and any tokenized yield fund, for that matter) as being worth $1. Stream's near $100m loss likely caused over $500m in contagion, if not more. Ethena's contagion would be orders of magnitude larger.

Aaveethena (the name of the partnership between Aave & Ethena), has proved to be one of Aave's primary growth engines. It involves lending against USDe, sUSDe, but also PT-sUSDe (fixed rate via Pendle). Supply caps and LTV parameters have kept increasing as the DAO has welcomed this partnership's effect on the growth of top-line metrics. As part of all of this, Aave has agreed to operate under the assumption 1 USDe = 1 US Dollar, which exposes its lenders to potential bad debt in order to protect borrowers.

This illustrates Aave's own shift in priorities, away from delivering the safest yields to their lenders, as the DAO places a stronger emphasis on growth and capturing market share. In this case, the interests of partner protocols, asset issuers and borrowers are superseding the interests of lenders. This is a dangerous precedent for Aave to set, especially within its core instance (Aave does achieve cross-chain risk isolation, at the cost of fragmentation, which will be addressed by Aave V4).

Underwriting risk of tokenized funds as collateral: TradFi won't do it, DeFi still needs to evolve to do it

In a recent blog post, DeFi Leverage for Tokenized Assets, I make the case for looping (serious) tokenized funds onchain.

Tokenized funds come in many forms: Ethena is one of them. Stream was one of them. Not all funds are equal in quality and creditworthiness.

DeFi's solution to underwriting tokenized funds as collateral has been to pretend they are "yield-bearing stablecoins" with significant liquidity and no drawdown. This is almost never the case.

TradFi has often favoured internal leverage (ie. leverage within a fund) over accepting fund shares/LPs as collateral (external leverage). DeFi is somewhat inverted: the base "yield-bearing" asset is often low leverage, considered relatively safe, and individual LPs are able to utilize looping to achieve their desired amount of leverage.

This process, however, means liquidity cadence (ie subscription and redemption periods and windows) are now intricately tied to the ability to adjust the leverage amount and ratio. It also introduces a high diversity of borrowers as opposed to lending to the fund itself for internal leverage. TradFi isn't well equipped to perform that kind of external lending against fund shares. DeFi is starting to do it, but isn't equipped to account for the risks and ensure proper unwinding (yet).

Infrastructure will need to be built for accurate, dynamic appraisal of built-in risk and available liquidity, instead of resorting to dangerous shortcuts that amplify the fragility of the entire system.

This will likely need to be the topic of another blog post, as it is a whole other can of worms that I do wish to open in due time. Until then, I am comfortable saying that skin in the game is probably the solution to most of our problems. We used to understand this as an industry (hence the comical obsession with slashing). Maybe a return to form is due.

Enjoyed this blog post? Consider subscribing:


#featured