The Curious Case of On-Chain Private Credit Repo
How loans go slow to fast to slow to fast again (on-chain)
The illiquid, complex, and opaque nature of private credit has traditionally been an awkward fit for crypto-native buyers of tokenized RWAs, who seek high nominal yields but optimize more for liquidity and transparency. In an effort to make private credit more attractive to this subset of investors, issuers and RWA builders have looked to leverage composability within DeFi to help end-investors achieve higher yields (e.g. lever tokenized private credit funds) to compensate for the “illiquidity premium.”
At Steakhouse Financial, we have been innovating on and tracking these developments from multiple angles:
As the biggest vault curator on Morpho
As the leading RWA advisory firm in DeFi, including driving the largest RWA investment efforts to date, by an order of magnitude, at Sky
As the first regulated onchain private credit practitioners ourselves, structuring and managing BlockTower Credit on Ethereum
In the below article we hope to provide some light into our thinking around risk management for off-chain assets, in particular private credit. We focus on primary liquidity and thoughtful LTV parametrization as two key drivers for avoiding bad surprises for DeFi lenders.
Over time, by integrating sound risk management that takes into account the particularities of alternative credit has the potential to transform the industry and give rise to liquid alternative credit as an asset class in its own right.
What is private credit?
Private credit is a catch-all term for all non-bank lending. At its most atomic level, a private credit transaction can be something such as a mortgage on a house, arranged through private negotiation. The traditional view is that most such loans would originate from commercial banks.
The traditional banking system is actually a small subset of the overall credit economy in the real world. When the definition of financial services stretches to include intermediaries that are not regulated like traditional banks, but are nonetheless able to issue loans as a commercial arrangement, the map gets much bigger. This dramatically expands the map of participants in the financial services industry, a chart-view of which was pioneered by Zoltan Poszar in 2010 with “Shadow Banking”1.
This represents a substantial portion of the financial economy. The FSB estimates that the non-bank financial intermediation sector represents almost half of global financial assets2 in a broad definition and still as much as 15%, or $70tn globally on a narrower measure. A relatable example of credit issued by a non-bank financial intermediary would be a Buy Now, Pay Later option at checkout, issued by a company such as Affirm or Klarna. In practice, aggregation of asset-backed credit may include baskets of loans issued both by banks and non-banks alike.
Colloquially in the industry, the term “private credit” is frequently used to describe corporate lending, outside of the traditional bank-syndicated senior secured term loans that primarily lever sponsor capital from Private Equity funds to engage in a LBO. Private credit lenders, such as Ares and Apollo, entered the market by seeking sponsors directly and enticing them, and their corporate targets, with the promise of a single point of contact and more flexible loan terms. More recently, the term has expanded, reflecting the desire for more flexibility across issuers across the asset-backed and corporate landscape.
Why are issuers trying to tokenize baskets of loans?
Securitization has a lot of positive benefits for credit markets. When credit expansion is not distorted by artificially low interest rates, or subsidies and incentives from government policy, it is typically a good lubricant for growth. Small and medium businesses rely on credit to finance inventory turns, expand their services and grow revenues. Households tap into mortgages to facilitate entry into the housing ladder, which stabilizes prospects for family creation and facilitates long-term economic growth and societal wellbeing.
Securitization is the process of standardizing baskets of very different types of loans so that it becomes easier for borrowers to access loans and for lenders to tap into borrowing demand. In practice, it means pooling, for e.g., a basket of mortgages into a single fixed income instrument, so that the risk of individual mortgage loans is socialized across a larger pool and so that the underwriting of the instrument can be made easier for the ultimate buyer.
Steakhouse Chef Kevin Miao3 discussed this topic at length on Not Boring a few years ago (borrowed at length, original below):
The core objective of every lender is to borrow money at X% and loan money at X+Y%, capturing a spread in the form of ongoing net interest margin (Y%) or through a one-time “gain on sale” of the underlying loan (roughly Y% * the effective duration of the loan). The cheaper the cost of borrowing (X%), the more profit you stand to make (until efficient markets drive down Y%). With that incentive, who would ever ask for a higher X%?
From the first modern clearinghouses in Liverpool, to the emergence of fractional reserve banking, to the proliferation of AI/ML in modern credit underwriting, driving down the cost of capital has long been the financial market’s north star. To this end, it’s hard to think of a more successful (or notorious) innovation than securitizations. Securitizations, archetypically mortgage-backed securities, are instruments which pool financial assets (loans) into tranches of debt that is subsequently purchased by investors.
Any mortgage-backed security can be thought of as a traditional company. Unlike traditional companies, however, where assets are factories, IP, or human capital; here, the “company’s” assets are mortgages, whose value derives from the contractual obligation of the underlying borrowers to make payments according to a fixed schedule. The company, in the form of a bankruptcy remote SPV, purchases these assets through a sale of liabilities to investors. These liabilities, or the tranches of the MBS, exhibit similar characteristics to any corporate capital structure: there are senior claims (tranches), junior tranches, equity claims (residuals), and so on.
Thus, a basic capital structure of a MBS could look like this:
What’s the point of all this financial engineering? At a high level, securitizations ratchet down the cost of credit through 3 main principles:
1. Law of Large Numbers
If you had to bet $1,000 dollars on coin flips, would you feel more comfortable betting $1,000 on one coin flip or $1 per coin flip repeated a thousand times? For risk-averse investors (i.e., debt investors), the latter is clearly preferred because one would expect the realized results to trend towards the true odds, or 50/50.
The same is true in lending. Even if I knew the probability that a loan would default, the idiosyncratic risks associated with a single loan are impossible to price. What if the borrower loses their job? What if an asteroid hit the town? Pooling these loans into a securitization dulls the impact of these exogenous, binary risks and increases an investor’s confidence that they will, in aggregate, perform according to the underwritten expectations. As a rule of thumb, the higher an investor’s degree of confidence in the outcome, the lower the returns one will demand.
2. Standardization (“Narrow Waist”)
Securitization is a form of standardization that enables very distinct forms of debt, ranging from 1-month credit card receivables to 30-year mortgages, to be purchased by and traded amongst investors. The “narrow waist” in securitizations is a 9-character identifier (CUSIP), a unit of ownership somewhat analogous to a unit of stock ownership in the equity markets.
Imagine if you could only trade shares of Tesla for other auto manufacturers; rational investors would require a greater discount to purchase that asset due to the relative illiquidity. Tranches of securitizations, standardized as CUSIPs (with consensus around cashflow waterfalls, legal structures, etc.), allow investors to trade very different forms of debt for each other, increasing the liquidity of their investments and lowering the rate of return they demand.
3. Product-Market Fit
People are different, and so are investors. Conservative investors, like banks or insurance companies, optimize for principal preservation. Others, like hedge funds, seek higher risk-adjusted returns. By creating senior and junior tranches, whereby higher-yielding junior tranches incur losses before lower-yielding senior tranches, you can cut up an asset into the distinct profiles that investors want. More competition, for both tranches, decreases investor leverage and often results in a lower aggregate cost of financing.
Pooling, standardizing, and tranching a pool of loans via securitization allows lenders to borrow at a lower-rate (X%). Say a lender has originated 100 $1 million loans, each yielding 10%. In securitized form, the senior debt may be a $70 million tranche yielding 4% and the junior debt a $30 million tranche yielding 15% (the standard target for hedge funds). As discussed above, investors are happy to pay this because they get a diversified pool of assets, with greater liquidity, and a better match for their preferences. But those pools of assets are now financed, by banks and hedge funds, at a weighted average cost-of-capital of 7.3%. This is the lower X%!
The question is, though, who gets the 2.7% (or Y%)? Part of this is fair compensation for lenders, but make no mistake: a significant portion of this spread is a symptom of our archaic securitization infrastructure and third-party fee extraction.
After the process of securitization is complete, credit funds invest into portfolios of securitized assets and often seek to further enhance returns by entering into repurchase transactions (i.e. repo). Repo is a form of short-term, collateralized lending that allows asset managers, or any other ultimate lenders, to avoid selling a perfectly solvent basket of loan securities if they have an immediate liquidity requirement or investment opportunity elsewhere. Therefore, the ability to access repo at attractive terms on a wide variety of assets is a pillar of most credit investing strategies.
How does DeFi repo work?
DeFi actually does collateralized repo extremely well and has the potential to offer a lot of cost savings and efficiency to traditional finance repo. The need for escrow or trilateral agreements that depend on court enforcement is completely sidestepped as the terms of a loan are set by rules that are known and programmed ahead of time and executed or enforced automatically.
TradFi issuers, like Apollo and Blackrock, have long been exploring tokenization as a pillar of their future distribution strategy, given existing channels are relatively tapped out (e.g. pension capital) or too expensive to serve (e.g. global wealth and retail). As tokenization moves from curiosity to strategic imperative, however, these issuers are seeking to prove to internal and external stakeholders that investments into tokenization will pay medium-term dividends. Given their primary business imperative is to raise and assess fees on invested capital, DeFi capital is the first “killer app.”
The largest such market today is Aave, with over $20bn of liquidity and collateral locked in the platform. For this article, we walk through an example repo transaction on Morpho. We are more familiar with the architecture and its simplicity makes it easier to explain. Morpho’s protocol enables two parties to coordinate a peer-to-peer repo transaction, so long as those two parties agree on five specific parameters4: Collateral Asset, Loan Asset, Liquidation Loan-To-Value (or LLTV), Oracle, and Interest Rate Model (or IRM).
In an oversimplified example, let’s say I have $100 wstETH:
I am seeking to borrow USDC against an asset I hold (wstETH)
I deposit $100 wstETH into a wstETH/USDC market with an 86% LLTV market with a Chainlink oracle
For simplicity, let’s say that the oracle is deriving prices from a composite of DEX prices (where actual transactions are being executed)
The IRM is the standard Adaptive Curve IRM5
I borrow $80 USDC, collateralized by my wstETH, and begin to accrue PIK interest on this position
If the market value of my wstETH position drops towards $86 (or towards 86% LLTV), my wstETH (Collateral Asset) may be liquidated into USDC (Loan Asset) and the lender, who lent USDC against my wstETH to generate yield, will be made whole
In this construct, the primary risk to a lender’s principal is if the wstETH position can be liquidated into USDC to pay off the lender’s position at par. If not, there will be an unrealized or realized loss on the lender’s position (e.g. I receive <$80 of USDC principal back). This problem can also be triggered by an oracle’s failure to submit accurate prices into the market (e.g. oracle reports a wstETH market value of $40 vs. $90, which can erroneously trigger a liquidation and result in an unnecessary realized loss).
The secondary risk to a lender’s principal is time to liquidation: The lender cannot force the borrower to repay because they may not even know who the borrower is. The longer this standoff persists, the more exposure a lender has to volatility in the underlying collateral asset and a potential loss of principal. Note: A liquidation is not guaranteed and occurs if independent third parties choose to purchase the Collateral Asset once its liquidatable. This primarily depends on the size and the certainty any liquidator has to realize the arbitrage.
For an asset like wstETH, with robust liquidity in atomically composable markets (i.e. DEX), onchain repo markets function extremely well. The oracles are often reading real and executable market prices, which means a lender has a fairly strong guarantee that the collateral asset can be liquidated and they will recover par on their loan.
Why is NAV important for tokenized private credit?
When it comes to offchain private credit RWA like ACRED, for e.g., however, these guarantees are difficult to come by. For a multitude of regulatory and economic reasons, there is effectively zero secondary liquidity for most tokenized private credit RWA. Instead, the “price” that an oracle propagates is not real or executable in size - it is the Net Asset Value (NAV) of the fund’s portfolio, which is the best estimate that any issuer can provide given the relatively illiquid nature of the underlying investments (some of which may never trade in their entire lifecycle). Marking on NAV is not nefarious in any way - it is simply the Issuer’s best attempt to provide transparency at frequent intervals, especially when the underlying assets do not have a market or exchange price (due to their relative illiquidity).
Let’s use BlockTower Credit (BTCR) as an example, which invests in more liquid but similar underlying assets (e.g. IG consumer asset backed securities). Every month, we followed the same structured process:
Estimate current market value on every position owned by the fund(s)
Seek multiple indicative marks from market participants who traffic in the underlying securities (e.g. broker dealers)
Send our positions, internal marks, and external indicative marks to an independent, third-party valuation agent (e.g. Houlihan Lokey)
Third-party valuation agents would then run their own independent analysis to come up with a final composite mark on every underlying position
That composite mark would then be applied to our current holdings, which form the NAV of BTCR
This NAV was critical for our LPs - it determined the mark-to-market performance of their investment but, more importantly, served as the valuation at which new LPs would enter the fund and redeeming LPs could exit. If a NAV was artificially or erroneously high, redeeming LPs would benefit by taking out a higher pro-rata share of real market value than LPs who stayed in the fund. Whether or not a NAV is calculated monthly, daily, hourly, or block-by-block, building trust around the valuation process for NAV is paramount.
Apollo has a track record for running an incredibly tight and highly surveilled process. Forbidden question: Can we say this about every potential private credit RWA asset issuer in the crypto ecosystem?
What are the risks for DeFi lenders?
Ultimately, the NAV of a fund is an estimate, not an instantaneously executable price (akin to wstETH on a DEX). For ACRED investors, the NAV is the basis of a narrow promise to “repurchase no less than five percent of its outstanding shares at net asset value (NAV) on a quarterly basis in accordance with the repurchase schedule.” For those who are lending USDC against ACRED on Morpho (or any DeFi repo market), however, this narrow promise of primary redemption - coupled with a complete lack of secondary liquidity - means that their USDC loan position lacks some of the functional protections that you receive when lending USDC against a liquid, crypto-native asset like wstETH.
To be specific, if I have lent $70 USDC against $100 of ACRED (instead of $100 of wstETH), what should I be asking myself that’s different than lending against an atomic, onchain native asset?
ACRED’s onchain price is its NAV. How comfortable should I be lending against this NAV vs. a real market-derived price?
ACRED, which tracks the Apollo’s NASDAQ-listed Diversified Credit Fund Class I shares (ticker: CRDIX), ranks much better than typical private credit funds in this regard
When I seek to withdraw my $70, how long will it take to unwind my position?
The preferred method would be for the borrower to simply pay back the $70 USDC, which the borrower can source from cash-on-hand or by selling liquid assets
If the borrower has no liquid assets, the borrower must monetize their ACRED position. The only way to do this is to submit a repurchase request to the Fund, which guarantees at least 5% of investors will be able to get out, quarterly and pro-rata, at NAV. In the worst case scenario, it could take the borrower (i.e. ACRED holder) over 5 years to fully exit their position. That said, this is highly unlikely to happen as not all ACRED holders will look to redeem simultaneously
As aforementioned, there is unlikely to be any meaningful secondary liquidity for ACRED in the near future. This puts additional onus on the primary redemption channel: The borrower cannot sell ACRED, only redeem quarterly. This increases the time to liquidation, exposing the lender to potential volatility and/or increasing the likelihood that their position becomes undercollateralized (unrealized loss) or liquidated (realized loss).
If the position became potentially liquidatable, what market participant would work to liquidate it? Traditionally, liquidators are seeking to arb the price at which they can purchase liquidatable collateral (wstETH) and sell it on the open market. The magnitude of the potential price arb gets market participants excited, but the certainty of realizing that arb depends on selling the position they purchase via liquidation. What discount rate would a market maker apply to liquidating this position, if their only way out is through primary redemption? Do they have any certainty that they can even enter the primary redemption queue (as they were not the initial holder)?
These questions are much bigger than just ACRED: They must be answered for the entire tokenized RWA industry to fulfill its potential.
How does Steakhouse mitigate risk?
The house view is that there are at least two design choices that can narrow the redemption-mismatch that plagues NAV-based RWAs in DeFi: (1) a prefunded liquidity facility, and (2) disciplined liquidation-loan-to-value (LLTV) parameters.
1. Liquidity facility
Issuers are gradually realizing that in order for the asset to keep pace with the 24/7 nature of DeFi and actually be useful onchain, it needs to be supported by a frictionless redemption route. Without a convenient redemption route, repo lenders and secondary buyers are likely to impose a large haircut on illiquid RWAs - no matter how pristine the underlying credit - undermining the asset’s effectiveness as on-chain collateral.
BlackRock’s BUIDL treasury fund shows the model in production. At launch, Circle parked USDC 100 million in a dedicated Ethereum wallet that a permissioned contract pulls from whenever an investor redeems BUIDL 1-for-1. The buffer is replenished as needed, letting the token drift no more than basis points from NAV despite zero traditional market-makers.
RWA issuers have several sourcing options for liquidity. Some choose to earmark a portion of the portfolio uninvested (which admittedly creates a cash drag) to keep in stablecoins or even invest in BUIDL itself (which creates less of a cash drag and still provides atomic liquidity). Others fund a dedicated facility out of the issuers’ own balance sheet. Riskier strategies might involve tapping a revolving credit line against the portfolio. Whatever the method, it is becoming clear that on-chain assets require decent exit ramps if illiquid, hard-to-price assets like private credit are to effectively rival crypto-native collateral in DeFi repo markets.
2. LLTV discipline
Even with a redemption buffer, lenders still bear mark-to-model risk. Setting LLTVs well below the fund’s historical draw-down and liquidity facility size adds a second layer of protection in lending markets, ensuring that collateral can be liquidated, or redeemed with a limited haircut before lenders suffer principal loss.
Together, a visible redemption buffer and conservative LLTVs give illiquid, mark-to-model RWAs a better chance to compete with crypto-native collateral in DeFi repo markets.