Market StructureJune 2026
On Chain Lending Needs a Market for Its Credit Risk
On chain lending now holds tens of billions and takes real credit losses, yet there is no open price for the risk underneath. Traditional finance built that market with the credit default swap and kept it closed. On chain, the same market can be built in the open.
Part 1: The Market Is Large, and It Loses Money
On chain lending is no longer small. Aave holds about $12.2 billion. Morpho holds about $6.8 billion. Spark holds $3.4 billion, Maple $2.1 billion, Compound $1.1 billion. The slice that is actively curated, where a firm picks the loans and sets the terms, is about $4.1 billion on Morpho alone, spread across 41 curators.
And it takes real losses. In 2022 Maple lost about $54 million when trading firms defaulted on uncollateralized loans, $36 million to Orthogonal Trading alone. In November 2025 the collapse of Stream Finance and its xUSD token put roughly $285 million of credit at risk across lenders and wrote down nine figures of it. In 2026 the Resolv and Kelp failures added more. Today roughly $80 million of bad debt sits unrecovered on Morpho.
This is credit risk, not a hack. A hack is the venue's code breaking, a one time fault that an audit can find and close. Credit risk is the lending losing money on a loan. It does not get patched, and it grows with the book. Hacks have taken more dollars on chain so far, but they are one time events. Credit risk is the permanent one.
| Loss | Year | Size | Cause | Type |
|---|
| MakerDAO, Black Thursday | 2020 | ~$4M | Liquidations could not clear in a crash | Credit |
| Maple, Orthogonal and others | 2022 | ~$54M | Borrowers defaulted on uncollateralized loans | Credit |
| Aave, CRV | 2022 | ~$1.6M | Illiquid collateral, the liquidation gapped | Credit |
| Stream Finance, xUSD | 2025 | ~$93M | The backing behind the collateral failed | Credit |
| Gauntlet, Resolv | 2026 | ~$6M | A stolen key minted unbacked collateral | Hack |
| Aave, Kelp rsETH | 2026 | ~$180M | A bridge exploit minted fake collateral | Hack |
Credit losses come from the lending. Hacks come from the code. Amounts approximate. The 2025 stablecoin wave also hit Elixir and Usual.
These losses are not theoretical. They are realized, on chain, and recent. The risk is here. The price for it is not.
Sources: DefiLlama protocol TVL (June 2026); Morpho API curator AUM and bad debt (June 2026); on chain default records and CoinDesk on Maple 2022; reporting on the Stream Finance xUSD collapse (November 2025).
Part 2: The Risk Is Set by Curators, and Their Record Is Invisible
On chain credit is underwritten by curators. A curator is the entity that sets the loan terms, which borrowers, what collateral, what loan to value, which oracle. On Morpho a curator is a firm like Steakhouse or Gauntlet running an isolated vault. On Aave the curator is the governance vote of the token holders. Centralized or decentralized, the curator is whoever sets the terms.
The terms do not even need a manager behind them. A market can run on a fixed set of parameters, a loan to value, a borrow cap, an oracle, with no one adjusting it by hand, and the credit risk does not go away. MakerDAO in March 2020 is the clearest case. The ether price fell by half in a day, the network clogged, the liquidations could not clear, and a few liquidators bought the collateral for almost nothing. The protocol was left with about $4 million of bad debt and had to cover it by minting and selling its own token. No firm curated it. No contract was hacked. The parameters and the liquidation mechanism produced the credit loss on their own. So the risk a market prices is the underwriting, whether a firm, a vote, or a fixed rule set it.
The curator earns fees and carries little or no first loss. The depositors take the loss. So the curator is paid much the same whether the credit is good or bad. Even when a curator posts first loss capital, it is usually raised from investors rather than the curator's own pocket, and it is a static buffer that does not reprice as the book weakens.
Worse, the record is hidden by survivorship. A curator that takes a loss tends to shrink and exit, so a snapshot of the live book reads clean even when the history is not.
K3 Capital was a major curator before the 2025 losses and holds about $22,000 today. Gauntlet lost about $6 million in the Resolv failure in March 2026, and a reading of its book three months later showed zero. Maple shows zero on a book that lost $54 million in 2022. The live number says nothing about the record.
So nobody can see, in a single number, how good a curator's underwriting actually is. Two curators showing the same clean snapshot can have completely different histories. The price of underwriting skill does not exist.
Part 3: TradFi Built This Market, and Kept It Closed
Traditional finance has a market for exactly this. The credit default swap, or CDS, lets one party pay a running premium and be paid if a borrower defaults. The premium is the price of the credit risk. The Federal Reserve describes CDS quotes as relied upon indicators of investors' perceptions of credit risk, and describes buying protection as akin to shorting the borrower's debt.
It is a large market. About $9 trillion of notional was outstanding in late 2024, down from a peak above $60 trillion in 2007. It is the established way to take a view on credit without owning the bond.
But it is closed. Single name CDS trade over the counter, priced by a handful of dealer banks. The largest dealers intermediate around 90 percent of the market. The European Systemic Risk Board found that roughly 80 percent of single name CDS on the largest European banks sits uncleared and outside public disclosure, accessible to only a handful of market participants.
So the one market that prices credit risk is private, dealer quoted, and walled off from everyone but large institutions. You cannot read the price, and you cannot take a side unless you are a dealer's counterparty.
Sources: Federal Reserve, Finance and Economics Discussion Series on CDS (2022); ISDA, OTC derivatives market size, second half 2024; BIS Quarterly Review on single name CDS clearing; European Systemic Risk Board report on credit default swaps (2025).
Part 4: On Chain, the Same Market Can Be Open
On chain, the inputs that TradFi keeps private are public. Every loan, every rate, every piece of collateral, and every loss is written to the chain and can be read by anyone. The opacity that defines the CDS market is not a feature of credit risk. It is a feature of where that market was built.
That changes two things. The reference can be read, not asserted. There is no dealer to ask for a quote. The realized loss is read off the chain after the workout, by a fixed rule, and published.
And the market can be open to anyone. A CDS is not insurance, because the buyer does not need to own the underlying debt. The same holds here. Anyone can take a position on a curator's underwriting whether or not they lend on the book. That is what brings two sided flow, and two sided flow is what makes a price mean something.
The data is already on chain. What is missing is the market on top of it.
Part 5: What the Market Looks Like, and Where the Line Sits
The unit of the market is the curator. A market exists for each curator whose book is large and diversified enough to price. Go long the curator's underwriting if you trust it, and earn the credit spread while the book stays clean. Go short it if you do not, pay a small premium, and be paid the realized loss if one lands.
The market sets the price. The reference, the value weighted spread the loans pay over the risk free rate, is published as a starting point, not a binding price. Where the market trades above the reference, it judges the credit riskier than the curator underwrote it. Below the reference, it judges the credit safer.
Settlement is on realized loss, read off the chain after recovery, through a bonded challenge and a determination panel so a payout cannot be faked. And the line between what counts and what does not is the heart of it.
A credit loss is covered. A hack of the protocol's own contracts is not, because that is the venue breaking, not the credit going bad. The losses table above already sorts the recent events by that line, and the market prices the underwriting, not the exploit.
The Case
On chain lending is going to get larger. Aave on its own is already bigger than many bank loan books, and the curated market beside it is growing. As the lending grows, the credit risk underneath grows with it, and today that risk has no price until it breaks.
TradFi met the same need with the credit default swap and built it behind a dealer wall. On chain, the loans, the rates, and the losses are public. So the market can be built the other way, in the open, where anyone can read the price and anyone can take a side. That is the market on chain lending needs, and it is the first place the market for credit risk can actually be transparent.
Sources
DefiLlama, protocol TVL for Aave, Morpho, Spark, Maple, Compound (June 2026)
Morpho API, curator AUM and on chain bad debt (June 2026)
On chain default records and CoinDesk, Maple 2022 defaults (Orthogonal $36M, Babel, Maven11)
Bank for International Settlements, Bulletin 57, on credit loss despite overcollateralization; MakerDAO Black Thursday forensics (Whiterabbit, Glassnode, The Block), March 2020
Reporting on the Stream Finance xUSD collapse (November 2025); OAK Research and Rekt News on the Resolv USR exploit (March 2026); reporting on the Kelp rsETH bridge exploit (2026)
Federal Reserve, Finance and Economics Discussion Series on credit default swaps (2022)
ISDA, OTC derivatives market size, second half 2024; BIS Quarterly Review on single name CDS clearing
European Systemic Risk Board, report on credit default swaps (2025)